GenAI Daily News Briefing: Ratings

Get the Ratings by Email:

GenAI Daily News Briefing: Ratings

Date Rating Title Rationale
02/20/2024 Important The Shift from Models to Compound AI Systems The evolution towards more complex, integrated AI systems is a key trend in the field, underscoring the importance of understanding and leveraging multiple models for improved AI applications.
02/20/2024 Important Air Canada must pay refund promised by AI chatbot, tribunal rules Highlights the legal and ethical implications of AI promises, serving as a cautionary tale for businesses on the importance of aligning AI actions with company policies and obligations. AI leaders will be held accountable for fails like this.
02/20/2024 Important Groq AI model goes viral Introduces advancements in AI processing speed and efficiency aided by new hardware that could compete with GPUs at some point, potentially impacting a wide range of AI applications and industries. While not essential at the moment, it's an important development to keep an eye on for future implications.
02/20/2024 Optional Glass Consult is our NEW clinical reference chatbot Offers advancements in clinical decision support for healthcare professionals. Its specialized nature makes it of limited relevance to a broader audience.
02/20/2024 Optional NBA Shows Off "NB-AI" Generative AI Tool to Personalize Basketball Game Viewing Showcases potential advancements in sports media personalization. Some of us were impressed by the capabilities - John was not :)
02/20/2024 Optional How Westpac is using Generative AI to speed up software development Demonstrates the power of GenAI in coding assistance but that use case is quickly becoming tables stakes, so not an essential development (though if your developers are not yet using such tools, get on it.)
02/21/2024 Essential You should be playing with GPTs at work If you aren't yet building personal GPTs to use at work you're missing out on more productivity and thus more coffee breaks :) This article shares 20 GPTs real people have built for themselves. Some of our faves included using a GPT to create technical documentation, fully cloning yourself, and a product copy tool that let's engineers and others who might not be natural writers create good copy.
02/21/2024 Important Quantum-Enhanced Generative AI Generates Viable Cancer Drug Candidates While the researchers admit they did not test whether their computation could be done as fast or well using classical computing, the potential for hybrid computing using AI and quantum is one for AI leaders to keep track of as it could become a viable option soon.
02/21/2024 Optional Adobe Acrobat adds generative AI to ‘easily chat with documents’ Our reaction: it's about time Adobe offered this obvious PDF-assistance capability (which shows how far we've come with GenAI - two years ago we would have been blown away by this use case). Adobe offers this for free now, but plans to charge in the future - will people be willing to pay? We're skeptical given other tools people are paying for can do this.
02/21/2024 Optional Bioptimus raises $35 million seed round to develop AI foundational model focused on biology The substantial seed funding and the focus on a foundational AI model specifically for biology reflect the shift towards specialized AI models in life sciences. TBD how effective the model will be but important for those in life sciences to track.
02/21/2024 Optional Scale AI to set the Pentagon's path for testing and evaluating large language models This is only about testing and creating standards for LLMs for the military, so not relevant to most. But we're surprised the DoD would only look to one vendor (and why ScaleAI?) for help with this - perhaps there are others working on it that haven't been announced.
02/22/2024 Important Microsoft and Intel strike a custom chip deal that could be worth billions Marks a significant partnership in the chip industry, potentially reshaping market dynamics and highlighting strategic collaboration between two tech giants.
02/22/2024 Important Nvidia's Q4 revenues hit $22.1B, up 265% from a year ago Reflects the tremendous demand for AI compute that's showing no signs of slowing down. The question is: When will we see the massive chips investments yield equally massive revenue gains? The pressure is on you to deliver, AI leaders.
02/22/2024 Important AI Beyond LLMs: Benchmarking the Complex Reasoning Performance of GPT-4 and EC AI Focuses on the evolution of AI technology beyond LLMs alone, emphasizing the importance of complex reasoning capabilities in AI's development and combining models to achieve superior results in complicated tasks like planning.
02/22/2024 Important ChatGPT bug leads to unpredictable outputs The incident highlights the need for reliability in AI applications, important for developers and businesses relying on AI models like GPT-4. Reminder that these models are still unstable.
02/22/2024 Optional Adtech pioneers launch AI startup to empower publishers and brands at ˜crucial moment™ Optional because this offering is not yet available, but could significantly transform the web ad space. Plans to enable companies to embed agents in third-party web content that consumers can interact with. Could become the new "ad."
02/22/2024 Optional Google launches two new open LLMs Yet another announcement of a somewhat open source model (weights released, but training data not and usage limited). However, of note - Hugging Face leaderboard shows this model outperforming other 7B ones like Llama.
02/22/2024 Optional China's Moonshot AI zooms to $2.5B valuation, raising $1B for an LLM focused on long context The significant investment underscores China's commitment to AI development, yet its immediate impact is seen as more relevant to those deeply invested in AI advancements, particularly within China.
02/22/2024 Optional RESEARCH: LongRoPE: Extending LLM Context Window Beyond 2 Million Tokens Represents an interesting development in AI research to significantly expand context windows, but it's only research at this point.
02/23/2024 Essential Google pauses AI tool Gemini's ability to generate images of people after historical inaccuracies Underscores the necessity of rigorous testing and the potential repercussions of deploying AI tools without it. Disappointing to see Google again making a mistake.
02/23/2024 Essential Pfizer is building a new generative AI platform for pharma marketing Pfizer's deployment of an AI platform for marketing purposes illustrates the growing acceptance and integration of AI technologies in sectors where accuracy and compliance are paramount. Includes system to flag sensitive content for extra review. Also interesting to see Publicis, a company threatened by GenAI content creation tools, now building these tools for clients as part of their business model - kudos to them!
02/23/2024 Optional Stable Diffusion 3 The introduction of Stable Diffusion 3 represents ongoing innovation in AI-generated imagery. However, we rated as optional due to its limited release status.
02/23/2024 Optional ChatGrid: A new generative AI tool for power grid visualization Great to see development for another use case in a sensitive industry, but unsure how likely it is that grid operators will grab this off of GitHub and put it to use given such sensitivity.
02/23/2024 Optional The Justice Department gets a chief AI officer The government is following through on the plans laid out in the AI Executive Order, but this isn't a critical read. However, if you're an AI leader looking for a change, note that the government is hiring many CAIOs right now.
02/26/2024 Important Salesforce Unveils Tableau Pulse for Faster Decision Making with GenAI This upgrade integrates generative AI into decision-making tools, enhancing efficiency and accessibility for Salesforce's extensive user base. Integrations into tools like this is how most people will interact with GenAI in the coming months and years.
02/27/2024 Essential Your Organization Isn't Designed to Work with GenAI Crucial for AI leaders to understand organizational adaptation for effective GenAI integration. You need to focus on enabling a collaborative dialogue between humans and AI as opposed to simply taking a capability out of human hands and putting it into machines.
02/28/2024 Important GitHub Copilot Enterprise Hits General Availability GitHub Copilot Enterprise includes some useful new capabilities like internal search and ability to incorporate an organization's own code and knowledge base.
03/11/2024 Essential Introducing Devin, the first AI software engineer While much of this functionality has been available for some time, this tool makes code generation and debugging fully autonomous, marking a step forward. Software engineers will still be needed and, in fact, their skill level must be very high to be able to guide this tool when it messes up. Companies are going to be able to do a lot more software development and upgrading going forward, removing current bottlenecks.
03/11/2024 Optional Amazon, Google Quietly Tamp Down Generative AI Expectations We usually like The Information's reporting, but this piece is off the mark. Yes, revenue for cloud providers won't be as significant as expected, but we know companies are using GenAI in areas like customer service (Walmart, Wells Fargo, AT&T to name just a few) where this article claims they aren't.
03/11/2024 Optional Building Meta's GenAI Infrastructure We already knew Meta's plans for building up compute capacity to 600,000 Nvidia H100s. Perhaps they published this info to reiterate their dedication to openness and attract scarce talent.
03/11/2024 Optional Midjourney is testing a highly requested “consistent characters” feature Midjourney's development of a feature for consistent character generation was discussed as an important advancement for those in creative industries such as marketing and entertainment. The ability to create entire videos or marketing campaigns with recurring, synthetic characters is growing near. However, it's not available yet.
03/11/2024 Optional Generative AI video startup Tavus raises $18M to bring face and voice cloning to any app This is confirmation of a funding round from last year for a tool with many competitors, though notable that this one will be able to live in your own CRM/environment.
03/11/2024 Optional Perplexity brings Yelp data to its chatbot Part of an ongoing trend of integrating various real-time data sources into AI systems and evolution of search.
03/11/2024 Important Insilico Medicine unveils first AI-generated and AI-discovered drug in new paper Massive speed-up and cost efficiencies achieved here in terms of finding a drug candidate and getting it into trials. Important to note, but since this is an announcement around a paper detailing the work as opposed to the work itself which was announced last year, we marked it as important rather than essential.
03/12/2024 Essential Wxclusive: U.S. Must Move ˜Decisively™ to Avert ˜Extinction-Level™ Threat From AI, Government-Commissioned Report Says This article delves into a government-commissioned report that stresses the urgent need for decisive action to mitigate potential catastrophic threats posed by AI advancements. It outlines a series of measures aimed at safeguarding against these risks that we think are unlikely to gain adoption, but worth reading this summary and maybe the full report over the weekend.
03/14/2024 Essential FIGURES Humanoid Robot Integrates OpenAI's Advanced AI for Enhanced Interaction The integration of advanced AI capabilities from OpenAI into FIGURES' humanoid robot represents a leap forward in making robotic technology more interactive, intuitive, and capable of understanding complex human instructions. Healthcare, customer service, warehouse work, and personal assistance need more workers - robots welcome!.
03/14/2024 Optional A Generalist AI Agent for 3D Virtual Environments The development of a generalist AI agent capable of functioning within 3D virtual environments represents an exciting advancement with potential applications in robot training and generalization. However, its current status as a research innovation with future-oriented applications renders it optional for immediate industry impact.
03/15/2024 Essential Bayer pilots unique generative AI tool for agriculture Bayer's pilot of a unique generative AI tool for agriculture addresses a crucial need for tailored agronomic advice, leveraging vast amounts of its own data to provide actionable insights for farmers. While still a prototype, the emphasis on creating knowledge utilities based on LLMs that are sector-specific is an important trend to watch - or jump in on...
03/15/2024 Optional Anthropic just released the smallest and fastest Claude 3 model By offering a model that is both smaller and faster, Anthropic addresses key barriers to AI adoption, such as computational requirements and cost. But we already knew this one was coming.
03/15/2024 Optional Databricks invests in Mistral and brings its AI models to data intelligence platform Mistral is getting around these days. Good for them and good to note if you're using or considering Databricks, but not an essential read.
03/15/2024 Optional Knowledge Conflicts for LLMs: A Survey Explores and categorizes the challenges posed by knowledge conflicts within large language models and provides insights into AI model training and optimization. However, it's pretty technical - maybe weekend reading if you're into the details.
03/15/2024 Optional Apple quietly purchased an AI startup this year Apple continues to make the news even though it still has no GenAI offering. Nothing to see here until they do.
03/15/2024 Optional Implementing generative AI with speed and safety Serves as a reminder of best practices in AI deployment rather than offering new insights or solutions, making it optional for those already versed in the field. Decent charts worth skimming.
03/18/2024 Essential John Holland Embraces Generative AI to Enhance the Productivity of Its Workforce John Holland's application of generative AI, including a custom version of ChatGPT developed in partnership with Microsoft, is an essential case study. It illustrates effective AI integration within a traditionally non-tech industry (construction), offering concrete productivity gains. Shows practical application, measurable outcomes, and the potential for replicating such success in other sectors.
03/19/2024 Essential Nvidia launches NIM to make it smoother to deploy AI models into production These three articles stem from announcements at Nvidia's event this week. They collectively represent a significant leap forward in AI hardware and integration capabilities, highlighted by NVIDIA's announcement of the Blackwell B200 GPU as the "world's most powerful chip" for AI. Additionally, NVIDIA's enlistment of top names in humanoid robotics for its new AI platform, GR00T, indicates a substantial move towards enhancing robotic capabilities with advanced AI. These developments are crucial for anyone involved in AI, robotics, or their applications across industries.
03/01/2024 Important Introducing Microsoft Copilot for Finance This introduction by Microsoft marks a significant move towards integrating AI deeply into finance departments for productivity and efficiency. It will be interesting to see how adoption unfolds - some finance pros might be skeptical while others will fully embrace.
03/01/2024 Optional Figure rides the humanoid robot hype wave to $2.6B valuation Big backers, giant raise, and a partnership with OpenAI. One to watch, but the robots aren't yet ready for action.
03/01/2024 Optional On the Societal Impact of Open Foundation Models Interesting risk framework that suggests more focus on the marginal risk of open source LLMs over risks already in play from current technologies. Some helpful points in here for AI leaders, but not an essential read.
03/01/2024 Optional Stack Overflow and Google Cloud Announce Strategic Partnership to Bring Generative AI to Millions of Developers Nice to see Google make a strategic move like this, but no need to read this product announcement article.
03/01/2024 Optional With Brain.ai, generative AI is the OS The initiative to integrate generative AI more deeply into operating systems and potential applications on smartphones and beyond is something to keep track of. However, we're skeptical about the immediate practical applications and adoption of this one.
03/20/2024 Important TacticAI: an AI assistant for football tactics The introduction of TacticAI by Google DeepMind and Liverpool FC applies a combo of types of AI within sports analytics. It predicts the outcomes of football tactics, such as corner kicks, underscoring the growing impact of AI on sports strategy. And, importantly, it demonstrates the technology's ability to simulate scenarios and recommend strategies based on predictive analysis, demonstrating the broader applicability of AI in analyzing complex, dynamic systems across different sectors.
03/21/2024 Important Siemens to deepen collaboration with NVIDIA related to generative AI for immersive real-time visualization This partnership highlights significant advancements in manufacturing and design processes through the integration of NVIDIA's Omniverse and Siemens' digital twin technology. It exemplifies the industrial sector's ongoing evolution towards more AI-integrated operations, marking it as an important development in the field of AI and manufacturing.
03/22/2024 Essential 16 Changes to the Way Enterprises Are Building and Buying Generative AI This article highlights significant trends, such as the rapid deployment of AI models and the integration of AI into core IT budgets, which are essential for AI leaders to understand. The detailed survey and visualizations offered in the article make it a vital resource for supporting AI leaders' budget requests.
03/22/2024 Essential Introducing the 01 Developer Preview The 01 device from Open Interpreter represents a super cool approach to integrating intelligence across various environments, offering open-source hardware and software that can significantly accelerate innovation in voice-activated control and automation. Its potential for widespread application and impact, encouragement for open innovation and the fact that AI leaders may have to deal with employees bringing this into the enterprise make this essential. (Our John Sviokla ordered one so review forthcoming!)
03/22/2024 Important Introducing SceneScript, a novel approach for 3D scene reconstruction Meta's development of SceneScript marks a significant advancement in 3D scene reconstruction, leveraging AI to understand and interact with virtual environments more efficiently. The open-source nature of the project and its application in enhancing AR glasses and especially robotics makes it a crucial development for the future of virtual interaction and machine perception.
03/22/2024 Important Microsoft's first AI PCs are the Surface Pro 10 and Surface Laptop 6 for businesses Microsoft's strategic move to integrate AI capabilities directly into personal hardware (via neural processing units) through the Surface Pro 10 and Surface Laptop 6 shows the company's move to vertically integrate down to the user. More strategic brilliance from Nadella.
03/22/2024 Optional Improve performance and reduce cost with fractional H100 GPUs Although fractional H100 GPUs (charged by the minute) offered by Baseten present a cost-effective solution for businesses needing high computational power temporarily, the broader impact on the market and its adoption remains uncertain.
03/22/2024 Optional NHS AI test spots tiny cancers missed by doctors The NHS's AI model for spotting tiny cancers that were previously missed by doctors illustrates the potential for AI to enhance diagnostic accuracy in healthcare and raise the bar on the standard of care. However, this isn't a new use case so we consider this an optional read.
03/22/2024 Optional Improving LLM performance with agentic behavior AI heavyweight Andrew Ng explains how to create agentic behavior to improve LLM outcomes. Important to know but his instructions are a bit high level so marking this as optional.
03/25/2024 Important Accenture's Generative AI Revenue Surpasses All VC-Backed Startups Combined While our experience working at major consulting firms leads us to believe some of the reported $1B in GenAI work reported is reclassification of current work/resources, this article offers AI leaders a good adoption proof point when asking for budget and resources for GenAI.
03/25/2024 Important RAFT: Adapting Language Model to Domain Specific RAG This piece introduces an approach combining retrieval augmentation (RAG) with fine-tuning techniques to enhance the applicability and efficiency of language models in specific domains. It is particularly relevant for technical teams looking to leverage the latest advancements in AI to improve model performance and applicability. For AI leaders, understanding and potentially adopting RAFT could lead to significant improvements in AI-driven projects.
03/25/2024 Important Tennessee becomes the first US state to protect musicians from the threat of AI New "Elvis Act" safeguards the intellectual property rights of musicians (including use of their voice) in the face of advancing AI technologies. Monitoring progress on various state and federal acts around IP issues as well as seeing how they later hold up in court is critical.
03/25/2024 Optional Stability AI CEO resigns to ˜pursue decentralized AI™ This latest departure of a key executive comes after several other resignations of critical employees in recent weeks as Stability struggles to generate revenue to support its AI decentralization ambitions. One to watch for those leaning on Stability offerings.
03/25/2024 Optional KPMG Using Generative AI To Enhance Capabilities Of Digital Gateway For Tax A GenAI upgrade to KPMG's existing offering that is hopefully a lot better than what we saw from some other tax assistance platforms like TurboTax. If you're a user, let us know what you think of this one!
03/25/2024 Optional Financial Times tests an AI chatbot trained on decades of its own articles The Financial Times' pilot with 500 FT Pro users is a nice example of an AI chatbot for enhancing user engagement, but is becoming common low-hanging fruit, making this optional news.
03/26/2024 Important The Unbundling of ChatGPT While we know GenAI solutions are increasingly becoming verticalized, the article is worth a read and features a nice chart of some of the role-specific GenAI tools out there. Also features a graph that shows ChatGPT hasn't acquired new users in a while to help prove the point.
03/27/2024 Important Is Your Company's Data Ready for Generative AI? This article by the prolific Tom Davenport delves into the readiness of companies' data for leveraging generative AI technologies. Through survey data, it paints a picture of the current state, where (unsurprisingly) most Chief Data Officers find themselves unprepared for the demands of generative AI. Some helpful figures for benchmarking and Tom always bring a thoughtful approach so worth a quick read.
03/27/2024 Important Databricks Launches DBRX, Challenging Big Tech in the Open Source AI Race Databricks seems to constantly make the news these days with strategic moves and growth. They're now venturing into the competitive arena of open-source large language models with DBRX and claim it has GPT-4 level performance with more efficiency. They're becoming a major player.
03/27/2024 Important Amazon Doubles Down on Anthropic, Completing its Planned $4B Investment Amazon finished out their initial investment - chump change for them so while this is a nice vote of confidence, don't read too much into it. Remember what happened with Inflection.
03/27/2024 Important Hume AI Raises $50M After Building the Most Realistic Generative AI Chat Experience Yet While the development of a chatbot with empathy and emotional intelligence is promising for the future of AI chatbots in sectors like healthcare this is a funding announcement. However, you can follow the link to play with the model so we're putting this as important as it's worth giving it a go.
03/27/2024 Optional Pentagon Tested Generative AI to Draft Supply Plans in Latest GIDE 9 Wargame This experiment underscores the broader implications of AI in enhancing decision-making processes and operational efficiencies in complex, dynamic environments but not an essential read or groundbreaking use case.
03/27/2024 Optional New Report Highlights Critical Gap Between Employees and Executives Around Generative AI Leadership Readiness This report from Udemy just repeats a call to action for educational initiatives rather than presenting new developments in AI technology or applications (and of course, markets their ability to provide that education).
03/27/2024 Optional New MLPerf Inference Benchmark Results Highlight the Rapid Growth of Generative AI Models While these benchmarks are critical for developers and researchers focusing on optimizing AI model performance, the technical nature and specific focus of the benchmarks might limit their immediate relevance to a broader audience. A nothing burger.
03/29/2024 Important Microsoft launches new Azure AI tools to cut out LLM safety and reliability risks We usually don't rate announcements of products that are not yet available as important. But given the prevalence of Microsoft in the enterprise and the issues this will reportedly address - hallucinations, prompt injection attacks and more - we want AI leaders to know this is coming.
03/29/2024 Optional Artificial intelligence boosts super-resolution microscopy Interesting to see a new spin on diffusion for image generation, what they call a Conditional Variational Diffusion Model (CVDM), but this isn't immediately relevant to the AI leader.
03/04/2024 Important Qualcomm Empowers Developers with AI Hub, Advancing Generative AI Revolution Qualcomm's announcement about offering a model hub for developers to access models optimized for Qualcomm hardware highlights the significance of on-device AI computing. The initiative is important due to Qualcomm's substantial market influence, though our John Sviokla made an impassioned plea to mark this Essential!
03/04/2024 Important Here Come the AI Worms Researchers created a worm exploiting security issues with LLMs, emphasizing the importance of awareness around AI security vulnerabilities. While the immediate threat may be theoretical, the potential for future exploitation makes it important for stakeholders to be informed and prepared.
03/04/2024 Important Copilot for OneDrive will fetch your files and summarize them No need to read this announcement, but important to be aware of Copilot 's OneDrive capabilities (eg, enabling file fetching and summarization) with so many enterprises using Microsoft.
03/04/2024 Optional Pioneering the Future: Leading Health Systems Test AI-Powered Healthcare Provider The article discusses the collaboration of over 40 healthcare providers on generative AI applications for healthcare professionals in areas like care management, post-discharge follow-up, wellness, and health risk assessments. The initiative is part of a broader trend of AI integration into healthcare, making it optional as the concept is not new.
03/04/2024 Optional H2O AI releases Danube, a super-tiny LLM for mobile applications H2O AI's release of a compact LLM designed for mobile applications represents another step towards bringing AI capabilities to edge devices. However, given the crowded space of LLMs targeting mobile and edge computing, this announcement is considered optional.
03/04/2024 Optional AI chip startup Groq forms new business unit, acquires Definitive Intelligence While potentially significant for Groq's business trajectory and market positioning, the lack of immediate impact and benefits places this news in the optional category.
03/05/2024 Essential Anthropic launches Claude 3 Challenges OpenAI's ChatGPT in terms of code generation, reasoning and math. This launch signifies a pivotal development in AI capabilities and market competition, though it seems we'll still be using different models to achieve best-in-class results for different tasks.
03/06/2024 Essential A Safe Harbor for AI Evaluation and Red Teaming Open letter issued by notable researchers at MIT, Stanford, CMU and others calling for LLM providers to remove restrictions preventing the research community from red teaming LLMs in good faith. As we continue to see providers release LLMs with flaws (see Google's recent gaff as an example), this would be a positive development.
03/07/2024 Essential US Army Experimenting with generative AI Chatbots in War Games: Report The deployment of generative AI technologies in military war games showcases an expansion in the application of AI beyond conventional sectors. Essential to know from a citizen perspective and gives AI leaders ammunition (pun intended) for making the case that their organization should get going on GenAI.
03/07/2024 Important AI Prompt Engineering Is Dead Long Live AI Prompt Engineering Nice piece debating how long PE will be needed. Our take: For quite some time. The transition from manual to algorithmic prompt engineering is important for AI leaders to acknowledge as it impacts the development, efficiency, and effectiveness of AI applications.
03/07/2024 Important Microsoft AI Engineer Warns FTC about Copilot Designer Safety Concerns Whistleblower case that highlights ongoing concerns about how well models are being tested for harmful outputs before being put into the hands of millions - including children.
03/07/2024 Important Google Engineer Indicted Over Allegedly Stealing AI Trade Secrets for China Brings to light issues around cybersecurity, intellectual property, and international relations in the context of AI technology. Cautionary tale for AI leaders - AI security protocols and access controls are critical!!
03/07/2024 Optional How Amperity Is Building Generative AI Tools Using OpenAI's GPT Models While the application of GPT models by Amperity to create generative AI tools is great, its impact is more incremental within the crowded space of AI-driven marketing solutions.
03/07/2024 Optional ArtPrompt: ASCII Art-based Jailbreak Attacks against Aligned LLMs We had a spirited debate on this one, but the majority felt this particular case represents a highly specialized security challenge. It's marked as optional, given its narrow scope and the assumption that such specific vulnerabilities will be addressed as part of ongoing security enhancements in AI systems.
03/07/2024 Optional Amazon Buys Nuclear Data Centre Campus for $650 Million While indicative of the growing demand for sustainable and large-scale energy solutions to support AI computations, not an essential read. Note Microsoft has been talking about doing this for a while, too, so we're likely to see more of this.
03/08/2024 Essential Saudi Aramco Unveils Industry's First Generative AI Model A 250B parameter model customized on 90 years' worth of company data tailored to needs of the industry -- impressive! This move could significantly influence the competitive dynamics and operational efficiencies within the energy sector, making it essential news.
04/25/2024 Important Simple probes can catch sleeper agents Anthropic continues to follow through on their commitment to responsible AI. Here they've studied and shared their findings on easy ways to detect secretly corrupted AI systems. AI leaders need to stay abreast of the nature of emerging threats to LLMs and how to combat them.
04/25/2024 Important Researchers develop malicious AI 'worm' targeting generative AI systems Same reason as above - security is critical. The worm can extract data and take over email systems to send out malicious emails. Intel collaborated with academia on this one - the worm was developed in a lab setting, so no risk in the wild yet, but it could certainly become one.
04/25/2024 Important CMA seeks views on AI partnerships and other arrangements The review by the UK's Competition and Markets Authority (akin to the US's FTC and anti-trust arm of the DOJ) is a precursor to a formal inquiry and part of a broader regulatory interest in AI partnerships, echoing similar investigations in the US and EU.
04/25/2024 Optional Apple releases OpenELM: small, open source AI models designed to run on-device This is the second batch of models Apple has quietly released (they published a multimodel one last fall). Open sourcing is somewhat new for Apple so we're noting this as an indicator of their LLM strategy but not worth reading the article. We're looking forward to the May 7 event they just announced.
04/25/2024 Optional Snowflake targets enterprise AI with launch of Arctic LLM We're excited about the company's repositioning under the new CEO to aggressively compete in the generative AI space. They claim this model is on par with Llama 3 but released no supporting documentation. We decided to save a potential 'Important' rating for the more comprehensive Cortex offering.
04/25/2024 Optional The Ray-Ban Meta Smart Glasses have multimodal AI now Nice to see new computing form factors continue to evolve but this isn't a worthy read for the AI leader.
04/01/2024 Important Congress bans staff use of Microsoft's AI Copilot Showcases the government’s rightfully cautious stance towards integrating GenAI tools within its operations, emphasizing security and privacy concerns.
04/01/2024 Optional NYC’s AI Chatbot Tells Businesses to Break the Law Another cautionary tale about putting GenAI tools out in the wild without proper tuning and guardrails.
04/01/2024 Optional OpenAI built a voice cloning tool, but you can’t use it… yet Another announcement from OpenAI about a capability that no one can use yet. Plus other providers like ElevenLabs already offer this.
04/01/2024 Optional How Stability AI’s Founder Tanked His Billion-Dollar Startup This long, well-researched story is rich in details about the operational and managerial missteps leading to Stability AI's challenges.
04/01/2024 Optional OpenAI and Microsoft Plan $100 Billion AI ‘Stargate’ It will be interesting to see if AI's power couple pull off this planned build of a mega-AI data center, but it's only in the planning stages at this point.
04/01/2024 Optional Generative AI to quantify uncertainty in weather forecasting Another example of Google's AI research prowess - and we welcome better weather forecasts - but not essential news for the AI leader.
04/02/2024 Important Yum Brands Doubles Tech Spending, Expands Use of Generative AI Yum Brands has a pretty comprehensive strategy for applying GenAI across its operations and is putting $21M (double last year's spend) into digital innovation this year. Core areas of its business are being targeted. This is another good article to have in your back pocket when trying to get budget for GenAI or facing skepticism over its value. But Yum appears to have no GenAI apps in production yet, preventing us from elevating this to an essential read.
04/02/2024 Important For Data-Guzzling AI Companies, the Internet Is Too Small We had a lively discussion around this one (starts around minute 26:00 in our morning briefing). It highlights the challenge of finding new data sources to improve the biggest of models any further given they've hoovered the entire internet, including many (but not all) video transcripts. As you've heard many times before, AI leaders, your company's data is extremely valuable. Are you using it to its full potential? You need to be.
04/02/2024 Optional Generative AI: The next S-curve for the semiconductor industry? Despite presenting forward-looking insights into the potential B2B and B2C GenAI growth areas that will drive semi demand, speculative assumptions in the analysis and a slog of a read led to its optional rating. Some interesting points if you can hang in there with it, but it doesn't provide a solid enough foundation for immediate action or decision-making.
04/02/2024 Optional Apple researchers develop AI that can 'see' and understand screen context Yet again, Apple in the news around GenAI, but not for announcing any GenAI products. This research is geared toward improving context to help them (eventually) create a good experience when using GenAI on your iPhone. Interesting, but not a critical read.
04/02/2024 Optional The Unreasonable Ineffectiveness of the Deeper Layers Authors report improving upon well-established pruning techniques. May be good fodder for engineers looking for new tricks, but not an essential read for the AI executive.
04/02/2024 Optional It's for Real: Generative AI Takes Hold in Insurance Distribution We know Insurance is an industry 'in the crucible' based on our WINS framework and hoped this article would provide some fresh insights. But there's not much new info here, and we didn't necessarily agree with the analysis provided in the charts, so giving this one an optional rating.
04/03/2024 Important WHO Unveils a Digital Health Promoter Harnessing Generative AI for Public Health This development was considered important due to WHO's significant role in global health and the potential impact of introducing an AI-driven assistant like this one (named SARAH) to serve the historically underserved. This is a major, cautious organization jumping into the GenAI mix. And our tests show the tool is quite good, despite a little latency and repetitiveness.
04/03/2024 Important Many-shot Jailbreaking Anthropic found this issue and kindly alerted other LLM providers to it. They've mitigated it in their models (though not to 100%) but we don't know if other providers have. Also, if you're building your own model, you need to be aware of this and take steps to combat it as well.
04/03/2024 Important Gen-AI Search Engine Perplexity Has a Plan to Sell Ads Search is changing - this is essential for a CMO but only important from an awareness perspective for the AI leader. You to be informed as to all the ways GenAI is changing the game in long-standing industry norms. Recall this startup also focused in this area.
04/03/2024 Optional Hailo lands $120 million to keep battling Nvidia as most AI chip startups struggle Although Hailo's successful fundraising marks its significance in the AI chip industry and the move toward compute at the edge, the group consensus leaned towards this news being optional. They seem to be gaining traction but Nvidia and other major players still rule the roost.
04/03/2024 Optional Luminance's Generative AI-for-law raises $40M Series B Despite acknowledging Luminance's achievement and healthy competition for Harvey, we see this as primarily a sector-specific development. It was noted that while the funding signifies growth in legal tech AI, it represents a trend rather than a breakthrough, making it of optional interest to those not directly involved in legal AI.
04/04/2024 Essential AI adoption accelerates as enterprise PoCs show productivity gains This article underscores the tangible benefits and productivity gains realized through GenAI adoption. Highlighting examples from 3 major companies - Webster Bank, Eli Lilly and Eaton - it highlights a lot of the best practices we've been espousing. Long but essential read with good proof points for the AI executive.
04/04/2024 Important Startup Datavolo raises over $21M to transform how generative AI models access unstructured data Datavolo's announcement is not necessarily significant for the capital raise, but it is for spotlighting the critical challenge of structuring your data pipeline - you'll need these types of tools because the classic ones won't cut it for routing unstructured data to GenAI.
04/04/2024 Optional Microsoft 365's Copilot gets a GPT-4 Turbo upgrade and improved image generation The upgrade to Microsoft 365's Copilot with GPT-4 Turbo and enhanced image generation capabilities signifies a notable advancement for users already embedded in the M365 ecosystem. While it introduces more powerful models for reasoning and supports larger text prompts, along with improved daily limits for image creation, it's seen as an incremental upgrade, but good for M365 users to be aware of.
04/04/2024 Optional I have a group chat with three AI friends, thanks to Nomi AI they're getting too smart Explores the evolving capabilities of AI in creating more convincing and emotionally intelligent virtual companions. Highlighting personal experiences with Nomi AI's chatbots, the discussion reflects on the broader implications for social interaction and mental health. Intriguing for its societal implications, but its immediate relevance to enterprise AI strategy and day-to-day operations is limited.
04/04/2024 Optional Enhancing The Forrester Experience With Generative AI A chatbot that provides on-demand info for high-tier users. What took you so long? Nothing burger.
04/04/2024 Optional Nature Communications Publishes Zapata AI Research on Generative AI for Optimization While the findings demonstrate the potential for quantum modeling to outperform traditional methods in complex optimization problems like portfolio management, its current status as emerging research renders it an optional consideration for a broader audience. We do think we'll see more quantum+AI progress so a good weekend read if you want to geek out a bit.
04/04/2024 Optional DALL-E now lets you edit images in ChatGPT Finally, in-painting comes to ChatGPT. It has been available in other popular image generation apps like Midjourney for a while. Our David DeLallo is a Midjourney power user, but says this will make him start exploring the use of ChatGPT more to create his weekly AI comic.
04/05/2024 Important JetMoE: Reaching LLaMA2 Performance with 0.1M Dollars A demonstration of achieving high-level AI performance with significantly reduced costs, though they compare this model's performance to Llama 2 as opposed to the most powerful models. Still, it offers hope to enterprises that it may not cost millions to build a custom model that gives you a competitive advantage.
04/05/2024 Important S&P Global launches groundbreaking AI benchmark for the financial industry We've been espousing the importance of industry-specific benchmarks for real-world AI performance. S&P Global's initiative serves as a pioneering example in the financial sector where precision, reliability and fostering trust are paramount.
04/05/2024 Important Wiz uncovers security flaws at Hugging Face Israeli cloud security firm, Wiz, was able to access private data and models from Hugging Face users by uploading a malicious LLM. Hugging Face has patched the flaw, but this is an important reminder of the continuous need for vigilance in vetting providers. This hack was possible because Hugging Face stores all instances of models and data uploads from users in the same cloud - they, of course, aren't alone in this.
04/05/2024 Important Introducing Command R+: A Scalable LLM Built for Business Cohere seems to be emphasizing its competitive pricing, though Command R+ is several times more expensive than Command-R (launched a few weeks ago). Of note that there's another option out there of an LLM with associated tooling.
04/05/2024 Optional OpenAI expands its custom model training program Good to know about some incremental improvements to OpenAI's tuning API and the scaling up of their hands-on work with customers, but not an essential read. Scattered among the technical notes, they share that Indeed and SK Telekom are using OpenAI models in applications.
04/05/2024 Optional Google using AI to come up with search answers in UK trial Google's expansion of its AI-driven search answers into the UK is another step in Google's testing of its Search Generative Experience, which is available to beta testers (we're among them) in the US. No need to read - we've just shared the news in a nutshell for you.
04/05/2024 Optional SiMa.ai secures $70M funding to introduce a multimodal GenAI chip Our Luda Kopeikina, a venture capitalist, notes that the valuation of this firm hasn't increased in this funding round so the future for this company might not be as bright as this announcement indicates. They're SoC offering is for the edge, which could make them an interesting takeover target for Nvidia, who isn't yet playing on the edge much.
04/08/2024 Important Hercules AI unveils assembly line approach for building enterprise-grade gen AI apps This platform looks promising for helping industries with a lot of info trapped in spreadsheets to extract it and enable LLMs to do what they do best - improve worker productivity. It also features an LLM that gets impressive coding results that surpass GPT-4 by simply fine-tuning a 7B model.
04/08/2024 Optional How Tech Giants Cut Corners to Harvest Data for A.I. A nice piece of reporting that gets into the details of the lengths OpenAI, Google and others are going to in order to get more data to train their models. And, as we know, the legality of some of their methods are questionable. A reminder to the AI leader about how valuable your data is. Good read if you have the time, but optional.
04/08/2024 Optional How Google lost ground in the AI race Similar to the previous piece, this is well-written and captures the details on the causes of Google's well-covered AI fumbles, but is an optional read. The company's GenAI course to date makes for a good Harvard Business case study - a giant organization with duplicative projects happening in silos that faces the innovator's dilemma around their search business.
04/08/2024 Optional Large language models are able to downplay their cognitive abilities to fit the persona they simulate This is not a surprising feature. We know that giving GenAI chatbots a persona (eg, 'act like the most effective marketer in the world') works well. If you're not employing this technique in your prompting strategies, you should be.
04/08/2024 Optional SWE-agent turns LMs (e.g. GPT-4) into software engineering agents that can fix bugs and issues in real GitHub repositories An application that looks to compete with Devin, but demonstrates slightly less effective performance and has a clunkier UX. Expect to see many more Devin competitors emerge.
04/08/2024 Optional Meta will require labels on more AI-generated content Of course Meta needs to do this to look like they're trying their best in the eyes of regulators and the public, but we know tools aimed at identifying AI-generated content are not very effective so don't expect a lot of help here.
04/09/2024 Important Jamie Dimon says AI may be as impactful on humanity as printing press, electricity and computers The headline isn't why we tagged this as important (and our John Sviokla argued it is essential). It's the information provided in Dimon's annual investor letter that we want the AI leader to absorb. JP Morgan is a tech leader among banks: 2,000 AI/ML workers, 400 apps in production, 32 data centers, and private and public clouds (though still migrating there). And Dimon says he sees AI impacting every role.
04/09/2024 Important Coles turns GenAI onto 40,000 customer comments a week Another article that nearly made it to essential because it's a great use case. Coles is a giant supermarket chain in Australia. They're using GenAI to process the 40K customer comments they collect each month and acting on the insights at a store level. Well done!
04/09/2024 Optional TSMC receives $6.6B to build chip factories under U.S. CHIPS and Science Act While essential for everyone as US citizens, optional read for the AI leader. Good to see the Chips Act working, and TSMC is more on track to stand up the 3 foundries they're building in Arizona than Intel - the first one is set to come online in 1H2025.
04/09/2024 Optional Google rolls out Gemini in Android Studio for coding assistance A nothing burger. It would be news if Google DIDN'T offer a coding tool.
04/09/2024 Optional Stability AI brings 12B parameters to Stable LM 2 model update Great that the company is still pumping out new models despite the turmoil over there, but are AI leaders going to use models from an organization that is hanging on by a thread? Imagine trying to justify that to your CEO...
04/09/2024 Optional Biocompare Announces Beta Launch of LifeSciAI, a New Generative AI Tool This supplier of products for the experimenting scientist is offering a chatbot that has knowledge of 8 million lab supplies and tons of experiments, but it's a common use case and a niche user base so we rated it optional.
04/10/2024 Essential Google Cloud Next 2024: Everything announced so far It may be a cloud event but AI is taking center stage. A new video generation tool in Workspace, a revamped image generation tool in the developer kit, and an agent builder were among the announced products. Nothing edgy - mostly their versions of tools already available from others. But Workspace has a huge user base and Google is 3rd in cloud, so worth knowing about.
04/10/2024 Important Intel Unleashes Enterprise AI with Gaudi 3, AI Open Systems Strategy and New Customer Wins Intel is still a few years behind Nvidia, but competition is needed in this space to meet demand and bring prices down so good to see Intel keeping at it.
04/10/2024 Optional AMD launches second-generation Versal chips to make AI faster at the edge This is an upgrade to an existing chip - feels like a 'me too' announcement on the day of Intel's. Important to know if you're a customer but not worth reading the article otherwise.
04/10/2024 Optional Ferret-UI: Grounded Mobile UI Understanding with Multimodal LLMs This is the paper that follows on the research announcement from a week or so ago. As we said then, of course Apple is putting effort into GenAI computing at the edge since it's where the company plays hardest.
04/10/2024 Optional Dairyland powers up for a generative AI edge We had high expectations for this article, hoping to see some interesting use cases from a local utility. But in the end, seems they're pretty much just using Copilot for coding and generating communications. They are using traditional AI in more impactful ways, though, so maybe more to come on the GenAI front soon.
04/10/2024 Optional New Federal Bill Could Require Disclosure of Songs Used in AI Training Rep. Adam Schiff introduced a bill that calls on LLM providers to reveal what music was used in their training data (nothing around other content types - industry lobbying, anyone?) but we've got some time to see if this becomes a law so nothing to see here yet.
04/11/2024 Important Poe introduces a price-per-message revenue model for AI bot creators Quora seems to have beaten OpenAI to the punch in providing a way for creators to monetize the bots they create using a revenue model similar to app stores. This is for bots developed on any platform as long as they integrate with Quora's Poe AI. Increasingly tough to understand why Quora's CEO still sits on OpenAI's board...
04/11/2024 Optional Beyond Transformers: Symbolica launches with $33M to change the AI industry with symbolic models Early days for this company that plans to combine symbolic AI, which has been worked on for decades, with deep learning to produce models and applications that are interpretable. John Sviokla and Tim Andrews engaged in a spirited (albeit geeky) discussion about when and whether LLMs could become interpretable which is worth watching in the recording or today's briefing, but Symbolica doesn't plan to release anything until 2025 so this is one to watch, but optional for now.
04/11/2024 Optional Mistral AI drops new 'mixture of experts' model with a torrent link We may elevate this to important once we start to learn how the models perform, but Mistral's minimalist method of announcing releases makes it tough for us to assess this development for the moment. We do know this is a talented bunch so we expect good things!
04/11/2024 Optional Ask your assets anything: State Street's generative AI lets investors chat with their data This piece of news almost made it to important. Custody is a boring but critical area of global finance. And State Street is one of a handful of banks that collectively serve as the 'keeper' of about 90% of assets. The company is building a chatbot for custody customers to interact with as well as a bot for a broader group of customers to more easily interact with their research. We would have rated this as Important if the bots were already available but they aren't just yet.
04/11/2024 Optional Intercom's new Fin AI Copilot gives customer service agents their personal AI assistant We know customer service is a prime use case, and this is a platform for building your own GenAI-powered application. But organizations are doing just fine building these already so not sure this will get a ton of traction.
04/12/2024 Essential OpenEQA: From word models to world models Meta released a benchmark to help model developers assess how their models do with establishing a representation of the world, which is essential for progressing the abilities of AI and robots. LLMs aren't very good at this--the latest version of GPT-4 does best based on Meta's analysis (full results provided in the article). The benchmark itself isn't necessarily essential but spurring research in this area is.
04/12/2024 Important Microsoft is working on sound recognition AI technologies capable of detecting natural disasters While this is only the announcement of Microsoft's filing of a patent in this area, 'AI as ears' will open up a host of new use cases. Here, Microsoft is patenting a technology that better senses sounds amid background noise. Important to start thinking about how you might be able to use this capability.
04/12/2024 Important NTT Research unveils AI model, sustainable path for AI, and better distributed data centers Three major announcements of note here: (1) they're releasing a model they claim does better than GPT-4 on analyzing inputs that contain graphics and text , (2) they propose a paradigm for distributed data centers that can improve their energy efficiency, and (3) they announce a partnership with Harvard to advance knowledge about the brain which would benefit AI development and human understanding. Research, but all critical areas.
04/12/2024 Optional European car manufacturer will pilot Sanctuary AI's humanoid robot BMW, Mercedes, Jaguar and others have committed to testing this robot. Great to see robotics moving forward, but let's see how these pilots pan out before we elevate this to important or essential.
04/12/2024 Optional RULER: What's the Real Context Size of Your Long-Context Language Models? Another benchmark. This one confirms what many of us experience - LLMs don't perform as well in their purported context windows as their developers claim. GPT-4 does best. Could be a more useful benchmark than some existing ones. Put it on your radar but might not be worth the long read.
04/12/2024 Optional Snowflake Copilot, a Mistral Large-powered AI assistant, launches in public preview Snowflake now enables natural-language prompts that translate into technical SQL queries. Nice feature, but no need to read the article.
04/12/2024 Optional The Weather Channel's parent company has a new AI tool to make hyperlocal weather videos Helpful for meteorologists, but the use case doesn't feature any novelties so parking this one in the optional spot.
04/15/2024 Important The Worst Part of a Wall Street Career May Be Coming to an End Banks leaders say that AI could soon enable them to cut down on junior analyst hires by a whopping two-thirds. Implications are huge for organizations.
04/15/2024 Important Introducing Rerank 3: A New Foundation Model for Efficient Enterprise Search & Retrieval The introduction of Rerank 3 by Cohere signals important advancements in cost efficiency and performance in enterprise search solutions.
04/15/2024 Important The rise of the chief AI officer There's a significant focus on the necessity of having dedicated AI leadership to navigate the complex legal, ethical, and operational challenges.
04/15/2024 Important Generative AI is coming for healthcare, and not everyone’s thrilled Put aside the fear-mongering headline. The article provides a good list of use cases that healthcare providers are exploring.
04/15/2024 Optional Apple plans Mac line overhaul with AI-focused M4 chips, Bloomberg News reports Yet again, Apple making news for something forthcoming but not yet available. Even when this chip drops, it's just an upgrade.
04/15/2024 Optional Former Google DeepMind researchers launch AI-powered music creation app Udio We knew this was coming and now it's here. Used in marketing and entertainment, but not an essential read.
04/15/2024 Optional British DARPA’ to build AI gatekeepers for ‘quantitative safety guarantees Certifications like the one this group is working on are welcomed, but this is still in the works.
04/16/2024 Essential 2024 AI Index Report This annual report from Stanford University is a comprehensive and valuable resource on AI's current state.
04/16/2024 Important Reka Core, Flash, and Edge: A Series of Powerful Multimodal Language Models Competitive offering from a small but well-funded group in Singapore with former Google Brain and DeepMind founders is notable.
04/16/2024 Important A.I. Has a Measurement Problem We need reliable and standard ways to evaluate and compare model performance.
04/16/2024 Important Cohere Compass Private Beta: A New Multi-Aspect Embedding Model Offers a new approach to embedding models that could improve enterprise data systems management.
04/16/2024 Optional Adobe Premiere Pro is getting generative AI video tools — and hopefully OpenAI’s Sora The capability isn't yet available, and the integration of Sora isn't even certain.
04/16/2024 Optional OpenAI's Altman pitches ChatGPT Enterprise to large firms, including some Microsoft customers Unsurprising continued coopetition at play with no compelling reason to engage OpenAI directly.
04/16/2024 Optional ArenaX Labs launches ARC AI game infrastructure and SAI research platform Cool development in gaming but not of note to the AI leader.
04/17/2024 Important Microsoft Makes High-Stakes Play in Tech Cold War With Emirati A.I. Deal Microsoft's investment into G42 has been orchestrated by the US government as a way to get companies in the Middle East more aligned to the US as opposed to China. While this is important from a geopolitical perspective, it's also yet another proof point for the AI leader that GenAI is truly powerful - our government is moving swiftly and decisively to position the US as a global AI powerhouse.
04/17/2024 Important Zendesk users get generative AI infusion for customer service While it's not surprising that a customer service software provider would integrate GenAI into it offering, the company includes some other nice features that provide ideas for any organization. The tool helps predict staffing levels and prioritizes an agent's work in addition to providing the agent with the best answers to customer issues.
04/17/2024 Important Pegasystems offers every worker access to a personalized, generative AI-powered coach This is another article we're flagging as a source of GenAI use case ideas. This low-code software provider is using GenAI internally in interesting ways - it integrates GenAI into an individual's workflow in a personalized way so that it gives advice to a sales or back-office worker in the course of their work.
04/17/2024 Important C6 Bank accelerates innovation using generative AI with AWS The use cases this major Brazilian bank backed by JP Morgan is implementing are low-hanging fruit (coding and customer service), but that's why we're flagging. These use cases are becoming table stakes so if you're not already at least testing them, you're falling behind.
04/17/2024 Important Nvidia expands Ampere-based GPUs for AI design and productivity apps Nvidia is everywhere - including at the edge. This will enable laptops to run AI faster.
04/18/2024 Essential Thomson Reuters unveils CoCounsel, leveraging generative AI for legal professionals Big name with a lot of data entering an increasingly crowded space, but one where Harvey has maintained its dominance to date. A formidable competitor has arrived. The economics for law firms and everyone who leans on them are changing fast. For AI leaders, this is a reminder that if your company is in an industry in the crucible according to our WINS framework, you have little time to lose.
04/18/2024 Important Logitech wants you to press its new AI button Let's be clear - this isn't important for the release of the $49.99 mouse itself. What's important is the fact that we're moving toward a world where we'll all have a physical 'easy button' that activates an assistant to help us do many everyday tasks (remember that Microsoft is installing one on keyboards as well).
04/18/2024 Important BigPanda launches generative AI tool designed specifically for ITOps While this functionality doesn't look extremely sophisticated, any help here is welcomed, and the AI leaders who also head up IT should take note that this functionality will be coming to the platform of your choice soon.
04/18/2024 Important Cheaper, Better, Faster, Stronger In this article, Mistral provides information on the 8X22B open source model it released last week. The highlights are that it covers 5 languages, offers native function-calling capability, and claims to have better performance-to-cost ratio than its nearest competitors, Llama 2 and Cohere's Command models.
04/18/2024 Important 12 companies that rolled out internal AI tools for employees The article shares an overview of GenAI-powered general knowledge and HR assistance tools used by both big (PwC, Walmart) and smaller firms (Pipefy). Easy read and a good scan of the types of tools that are becoming commonplace in many organizations. Worth a few minutes of your time.
04/19/2024 Essential Meta launches Llama 3 This isn't an "another model, another day" type of announcement. Besides already shooting to the top of the open-source model leaderboard on performance, Llama 3 is now heavily integrated into all of Meta's social channels, enabling search, image generation, animations and more right in its apps. The image generation capability is particularly impressive - so fast that the images generate as you type in your "order." Llama 3 still doesn't hold a candle to the top closed source models but the next version of it due out in the summer sounds like it might...
04/19/2024 Important Honeywell exec reveals plan to deliver $100 million in value with generative AI: “Just getting started” While this isn't a significant amount of value generation for a $90B+ company, Honeywell shares info about its organizational GenAI strategy here that offers some tips / validation for the AI leaders' approach.
04/19/2024 Important Generative AI in manufacturing — out of the old, emerges the new Industrial giant Bosch details how they're creating synthetic data to enable GenAI-powered quality assurance. The company noted that it has so few defects in its products, that it doesn't have enough error-ridden product images to train the system - a little self-promo never hurts. This is a great use case.
04/19/2024 Important Meet Lingo-2, a groundbreaking AI model that navigates roads and narrates its journey London-based autonomous driving tech provider Wayve released this second version of its multi-modal model that can share what it's doing as it maneuvers (so far, only in simulation environments). Multimodal is the future so this is of note, but the development of this model is not yet essential given where we are with autonomous vehicles.
04/19/2024 Optional ChatGPT is coming to Nothing’s earbuds The name of the company (which creates mobile phones - anyone actually have one?) says it all about this nonessential development.
04/22/2024 Important Intel Builds World’s Largest Neuromorphic System to Enable More Sustainable AI While this system has a ways to go before reaching the enterprise, it's important to know that more efficient computing for generative AI is on the way. Intel claims this system uses a whopping 100X less energy than conventional CPU and GPU architectures. Sandia National Laboratories will be the testing ground for this - we look forward to seeing how it goes!
04/22/2024 Important Langdock raises $3M with General Catalyst to help companies avoid vendor lock-in with LLMs This particular company's tool may or may not end up being notable, but the space they're playing in is. The middle layer that they're providing between LLMs and the user enables access to whichever LLM is best-suited for a particular task and is sanctioned for use by an enterprise.
04/22/2024 Optional Microsoft shows off VASA-1, an AI framework that makes human headshots talk, sing Just one image can be turned into a video with synced audio and pretty realistic animation - quite amazing. But you can see how this could be used for nefarious purposes so Microsoft is taking the responsible route and not making this available. They offered no timetable for when they might which is why we marked this optional.
04/22/2024 Optional How United Airlines uses AI to make flying the friendly skies a bit easier Pretty underwhelming generative AI use cases discussed like creating an announcement for a pilot. Skip this one.
04/22/2024 Optional Two-thirds of top 20 pharmas have banned ChatGPT—and many in life sci call AI ‘overrated,’ survey finds This click-bate got our John Sviokla fired up on an early Monday morning. While some pharmas might be banning ChatGPT specifically, the article also notes that pharmas are using generative AI in critical areas like drug discovery (there's already a GenAI-created drug in Phase II trials, btw). If anything, this article shows companies what not to do - it says half of employees are using GenAI chatbots anyways. There are too many easy ways to enable your employees to use GenAI chatbots safely to ban it and give up the efficiency gains.
04/23/2024 Important Generative A.I. Arrives in the Gene Editing World of CRISPR While there's no immediate action item for an AI leader from this news, it's just a stunning advancement that humans in general should know about.
04/23/2024 Important Efficient finetuning of Llama 3 with FSDP QDoRA This is a super-technical article that you should only read in its entirety if you seriously like to geek out.
04/23/2024 Optional Introducing the Data Intelligence Platform for Energy Databricks' vertically integrated offering is notable for those in energy and other asset-heavy industries.
04/23/2024 Optional Swiss Re Launches Swiss Re Life Guide Scout, a Generative AI-Powered Underwriting Assistant A nice use case, but not groundbreaking.
04/23/2024 Optional Sam Altman Invests in Energy Startup Focused on AI Data Centers Sam is diversifying his energy plays, adding this to his portfolio.
04/23/2024 Optional Apple acquires French startup behind AI and computer vision technology This acquisition happened in December but is coming to light now.
04/24/2024 Essential Introducing more enterprise-grade features for API customers OpenAI announced new enterprise features to better serve large-scale business needs, including enhanced security, administrative controls, and, perhaps most notably, favorable pricing strategies.
04/24/2024 Essential New additions to Amazon Bedrock make it easier and faster than ever to build generative AI applications securely Amazon’s improvements to Bedrock substantially enhance the ease of integrating and managing AI models.
04/24/2024 Important BCG says AI consulting will supply 20% of revenues this year BCG’s projection highlights the growing influence and integration of GenAI.
04/24/2024 Important Perplexity is raising $250M+ at a $2.5B-$3B valuation for its AI search platform, sources say The march toward reimagining search with GenAI continues. Perplexity has big-name backers.
04/24/2024 Optional The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions Discusses a training approach that makes LLMs prioritize instructions which could help protect against threats.
04/24/2024 Optional Microsoft launches Phi-3, its smallest AI model yet Microsoft's launch of this 3.8B model represents more advancement in creating efficient, compact AI models.
04/24/2024 Optional The Coca-Cola Company and Microsoft announce five-year strategic partnership to accelerate cloud and generative AI initiatives This partnership between Coca-Cola and Microsoft to build out new GenAI use cases together.
04/26/2024 Essential At Moderna, OpenAI’s GPTs Are Changing Almost Everything Highlights the company's broad application of GenAI from legal processes to drug discovery. More than 750 GPTs created by employees to date with plans to engage OpenAI for more formal applications. However, this is a company struggling with only 1 product in the market and there's a little too much smoke here as opposed to fire to call this essential.
04/26/2024 Important Stainless is helping OpenAI, Anthropic and others build SDKs for their APIs Focuses on Stainless's role in enhancing the accessibility of LLMs through SDKs, which aid in the broader deployment of AI technologies by bridging a crucial gap for developers. Very small company ($1 million in annual revenue and 10 employees), but we're flagging to let the AI leader know that their teams do need SDKs for APIs.
04/26/2024 Important AI designs new drugs based on protein structures The model discussed here was developed through an industry/academia collaboration between Bosch and ETH Zurich. It's tuned to (1) produce only those compounds that can actually be created and (2) limit side effects. It's being open-sourced. A good story all around to take note of.
04/26/2024 Optional New Cohere Toolkit Accelerates Generative AI Application Development Cohere has made some important announcements recently related to new models and RAG tools, but, while this is a helpful set of tools - includes interfaces, connectors and other goodies - we didn't think this one rose to that level of significance.
04/26/2024 Optional Research identifies pitfalls and opportunities for generative AI in patient messaging systems Nice to see an actual study on the use of LLMs v. LLMs+humans v. humans alone but the insights it yielded aren't surprising. The conclusion was essentially that LLMs are helpful, but humans need to edit LLMs' work. Duh...
04/26/2024 Optional Watch it and weep (or smile): Synthesia’s AI video avatars now feature emotions If you watch the videos of an avatar before and after this feature was added, you'll see there's only a subtle improvement in emotional facial expressions.
04/26/2024 Optional SenseTime Shares Surge on New Language Model It Says Rivals GPT-4 Turbo There's no evidence provided to back their claim so this stays in the optional camp for now.
04/29/2024 Important JP Morgan AI Research Introduces FlowMind JPMorgan demonstrates its leadership in AI in the financial sector with this paper that explains how they're using LLMs to automate workflows that are unpredictable. They also use a training method to limit hallucinations that's different from RAG. FlowMind writes code to automate a workflow as the user performs a string of tasks and then presents what it created at a high level for user feedback. Looks to still require that user to have some technical knowledge, though. Sounds a lot like what we described in our recent HBR article - love this.
04/29/2024 Important Oracle doubles down on generative AI with new features to accelerate enterprise deal cycles They're offering 50+ GenAI capabilities in Oracle CX Cloud. One example is automatic generation of next-best-actions for customer service agents, as opposed to the agent having to type in the customer problem to get recommendations. The article also shares that Cohere is under the hood of these capabilities - we've seen a lot of new offer announcements from Cohere over the last few weeks and here we see them in action.
04/29/2024 Optional Estée Lauder and Microsoft partner to help beauty brands use generative AI We've recently seen similar announcements of industry/big tech partnerships, for example Moderna/OpenAI and Coca-Cola/Microsoft. They're primarily just grabs for attention. We'll call it important when they have in-production use cases.
04/29/2024 Optional Chatbot answers are all made up. This new tool helps you figure out which ones to trust. MIT has created a tool that summons multiple LLMs to opine over one LLM's answer and assign it a probability of correctness. But nothing will ever get a perfect score so not sure how this is helpful. For now, you're always going to have to check LLM responses for accuracy.
04/29/2024 Optional It’s not only AI that hallucinates Nice to see a defense of LLMs for their occasional inaccuracies - the author points out that humans don't have impeccable recall either. But it's a meandering editorial with no real value - skip it.
04/29/2024 Optional Make Your LLM Fully Utilize the Context While the research on enhancing LLMs' context utilization is of academic interest, we're not convinced of immediate practical applications or significant breakthroughs.
04/30/2024 Essential NIST launches a new platform to assess generative AI A government entity is taking the lead on developing and enforcing GenAI standards - the National Institute of Standards and Technology (NIST) announced a new program, NIST GenAI, that will release benchmarks, help create “content authenticity” detection (i.e. deepfake-checking) systems and encourage the development of software to spot the source of fake or misleading AI-generated information.
04/30/2024 Important GitHub Copilot can now help start a project with AI, not just complete it GitHub Copilot will now have capabilities to start/design a software project. For developers, this is an important component that will make their work more efficient.
04/30/2024 Important Healthcare industry sees increased investment in generative AI, LLMs This article reports 300% growth in budgets allocated for generative AI projects in the healthcare and life sciences sectors according to one-fifth of the study participants. There is also useful statistics on GenAI applications development and deployment in these sectors.
04/30/2024 Optional Mysterious new model called gpt2-chatbot This is speculative news that is bound to create buzz but not worth a lot of attention. It's similar to the news in 2019 announcing that scientists developed an AI model that is so powerful that it is dangerous to release.
04/30/2024 Optional ChatGPT’s AI ‘memory’ can remember the preferences of paying customers The Memory functionality was announced in Feb 2024. Now this memory capability is available. More features are needed for it to become useful for industrial applications.
04/30/2024 Optional OpenAI licenses Financial Times’ content for ChatGPT OpenAI has been on a partnership mission lately. And, even though we believe that the partnership with Financial Times is powerful for both parties, it is one of many.
05/01/2024 Essential Siemens Xcelerator: Scaling Roll-out of Generative AI with Siemens Industrial Copilot A vertically integrated industrial GenAI tool from a market leader. It provides functionality to work with the underlying hardware, one of the most difficult components to design and manage. Also has private co-pilot sandboxes that will not be used for model training. A potential game changer for industrial applications.
05/01/2024 Important AI Startups Have Plenty of Cash. They Often Don’t Yet Have a Business. Thoughtful, skeptical article on AI revenue versus funding. A short article worth reading to reflect on the current status of the AI market.
05/01/2024 Important Amazon expands enterprise AI play with wider availability of its Q chatbot Amazon announced its Q chatbot in November 2023. However, new capabilities were added to Q ahead of general availability. Amazon Q Developer can now provide coding assistance, app testing, security scanning, troubleshooting, and call up AI agents that autonomously perform tasks like software updates or documenting code. Amazon Q Apps, another new feature, aims to make building generative AI-based apps easier, even for employees without coding experience. Amazon says users just have to describe the type of app they want in a prompt, and Q will generate the app they’re looking for.
05/01/2024 Important Yelp is launching a new AI assistant to help you connect with businesses Yelp's AI assistant is an advanced web search capability that provides customized content in response to a query, with links to businesses that can fulfill the request. If you are a business that gets volume from Yelp, this is essential news.
05/01/2024 Optional Ads for Explicit ‘AI Girlfriends’ Are Swarming Facebook and Instagram Interesting dynamics around the virtual versus real sex worker ads on Meta.
05/01/2024 Optional DressCode: Autoregressively Sewing and Generating Garments from Text Guidance Cool new GenAI tool to help create new fashions.
05/02/2024 Essential Atlassian launches Rovo, its new AI teammate Atlassian, one of the leaders in the data management space with its 300K customers, introduces Rovo that is focused on addressing an important challenge of internal data management tools proliferation within an enterprise. Not only Rovo offers to integrate the tools, it provides AI agents to act on the collected data, a critical performance advancement for knowledge workers.
05/02/2024 Essential Could generative AI work without online data theft? Nvidia's ChatRTX aims to prove it can A confirmation of a trend to provide secure access to user's local data at any time. Nvidia first introduced ChatRTX as “Chat with RTX” in February as a demo/experimental app. The app essentially creates a local chatbot server on your Windows PC that you can access from a browser to get a powerful search tool on your own data without compromising your data security. With this release NVIDIA added voice for query and extended the list of LLMs available for query.
05/02/2024 Important Anthropic’s Claude Teams and iOS App: The secure, scalable solution for enterprise AI adoption Anthropic, one of the AI market leaders, has been focused on data security and integrity in its products, As a result, it has achieved significant traction in highly regulated industries such as healthcare and finance. Anthropic now extends its capabilities with teams' functionality and a mobile app making its products more accessible for broader enterprise adoption.
05/02/2024 Important RAG and RAU: A Survey on Retrieval-Augmented Language Model in Natural Language Processing Excellent survey paper on how Retrieval-Augmented Generation (RAG) and Understanding (RAU) work with LLMs and NLP. Worth taking time to read.
05/02/2024 Optional Samsung’s operating profit soars 930% as AI tailwinds drive demand for memory chips Not a surprising development but well deserved.
05/02/2024 Optional OpenAI’s Sora in ophthalmology: revolutionary generative AI in eye health Short article suggesting ways on how Sora could be used for eye health education.
05/03/2024 Essential The AI-Generated Population Is Here, and They’re Ready to Work Having the ability to create digital twins is a huge trend that will change the way we live. For example, having digital twins for customers and employees will be essential to learning and marketing.
05/03/2024 Important With Tribble, RFP writers get their own shot of generative AI Tribble is not alone in providing AI help for RFP responses. However, its founders are ex-Salesforce executives and have lived through the pain points. Sales in general, RFP responses and AI enabled Sales Engineers using "tactical AI" are certainly areas seeing much innovation from AI.
05/03/2024 Optional Stardog’s Karaoke offers on-premises, zero hallucination LLM solution for enterprises Smart use of knowledge graphs plus generative AI.
05/03/2024 Optional CoreWeave raises $1.1B to expand its GPU cloud infrastructure network This is a funding announcement. Interesting to see that investors are willing to put a lot of money into AI infrastructure at high valuations. If you are in need of additional GPU cycles, CoreWeave can provide them.
05/03/2024 Optional Better & Faster Large Language Models via Multi-token Prediction Interesting model training technique to perform multi token prediction. The method is showing good results in outperforming older methods. Something to watch if you are training your own LLMs.
05/03/2024 Optional An Ivy League school just announced its first AI master's degree Tremendous to see that there will be additional graduates to fill in-demand AI jobs.
05/06/2024 Important Four start-ups lead China’s race to match OpenAI’s ChatGPT Useful article to be aware of - Chinese efforts in LLMs and four specific start-ups (Zhipu AI, Moonshot AI, MiniMax and 01.ai) aspiring to beat ChatGPT.
05/06/2024 Optional SHISEIDO HK fuses creativity with AI in new branding campaign New combination of talent, AI and art in advertising. The resulting ads, however, are not yet very impressive. A trend to watch.
05/06/2024 Optional An Economic Solution to Copyright Challenges of Generative AI A research article presenting a clever but rather complicated method to share royalties to artists contributing to an LLM. An interesting potential solution to copyright challenges. However, it does not apply to text and, as a result, cannot be implemented broadly.
05/06/2024 Optional Brightcove integrates AWS’ generative AI assistant AWS customer announcement about a cloud streaming provider, Brightcove, implementing Amazon Q Business solution.
05/06/2024 Optional Walmart to use Generative AI to Help Reduce Food Waste in Stores Announcement from Walmart about an internally developed solution to provide guidance to employees on dealing with expiring food items to reduce food waste. Rating as optional since this application is not operational yet (about to be piloted in Canada). It is, however, an important use case to monitor.
05/06/2024 Optional Google Says Its Med-Gemini AI Healthcare Models Beat GPT-4 It is not surprising that LLMs fine-tuned for specific applications perform better than general purpose LLMs. Note that Med-Gemini is using contextual information - a trend to make generic LLMs much more effective.
05/07/2024 Essential Hugging Face launches LeRobot open source robotics code library Open source toolkits and data sources should help speed up the already accelerating area of robotics. Generative AI is helping to do planning in unmodeled places. For defense, healthcare, any company with a distribution center, and more - the robots are coming.
05/07/2024 Important UPS delivers customer wins with generative AI Another company using GenAI for customer service with plans to do much more in finance, HR and sales. If your business has a significant customer service element and you're not yet at least testing GenAI to improve response time and efficiency, you're falling behind.
05/07/2024 Important Stack Overflow signs deal with OpenAI to supply data to its models While there's nothing actionable here for the AI leader, it's important to stay aware of how information delivery/consumption is evolving with chatbots becoming pervasive. Stack Overflow has been THE go-to for developers for quite sometime, and the article notes traffic has been down since the use of GenAI coding assistants has become widespread. Content providers are adopting different strategies - some are making deals with LLM providers (eg, Axios, FT) and others are suing to keep their content out of models (eg, NYT). Who's got the right strategy and where will this all land?
05/07/2024 Important Meet MAI-1: Microsoft readies new model to compete with OpenIA, Google We have seen Microsoft release small models and here we're learning that they're also working on one that could compete with GPT-4 and Gemini - though not even Microsoft is certain it will reach that level of performance just yet. It will have about half as many parameters so it will be quite interesting to see if it does. Microsoft is clearly reducing its reliance on OpenAI, which is wise. Regulators may force the issue.
05/07/2024 Optional Randy Travis gets his voice back in a new Warner AI music experiment Two sides of the coin here regarding voice replication. Randy Travis gets to make a song he otherwise wouldn't have been able to produce since he lost his voice to a stroke. But then we see unwanted deepfakes of rapper voices. We already know the danger of deepfakes in scams, so these are not essential reads.
05/07/2024 Optional In the Battle of Drake vs. Kendrick Lamar, A.I. Is Playing Spoiler Two sides of the coin here regarding voice replication. Randy Travis gets to make a song he otherwise wouldn't have been able to produce since he lost his voice to a stroke. But then we see unwanted deepfakes of rapper voices. We already know the danger of deepfakes in scams, so these are not essential reads.
05/07/2024 Optional Alphabet-owned Intrinsic incorporates Nvidia tech into robotics platform Important news and advancements may come from this partnership down the line, but this announcement doesn't carry any.
05/08/2024 Important How VISA is using generative AI to battle account fraud attacks Here we have a company using its own data for a critical use case and advantage - something every company should be considering. That makes this news important, though the article spends too much time educating the reader on types of attacks. Skim those parts but take away the main message.
05/08/2024 Important AI Copilots Are Changing How Coding Is Taught Great to see some areas of academia already adjusting the way they teach coding based on the emergence of coding assistants. No need to spend so much time on syntax, but they're moving more advanced skills like problem decomposition up in course syllabi and spending time on teaching testing and debugging - critical skills to assess coding assistant outputs.
05/08/2024 Optional Microsoft Creates Top Secret Generative AI Service for US Spies Microsoft's ties with the government sure run deep, even as the government is coming down on them for security issues. Wouldn't it be interesting to have access to this model built just for the CIA and other intelligence agencies using their classified data? As you can imagine, it's strictly on-prem signaling the government has some nice compute power.
05/08/2024 Optional Apple Is Developing AI Chips for Data Centers, Seeking Edge in Arms Race The article notes this isn't even confirmed yet, and even if true, not surprising or notable.
05/08/2024 Optional Nike developing AI model as part of design step change Nike is working on its own LLM customized with its data - as we mentioned earlier, that's something everyone should explore. The personalized shoes they made for some of their athletes are pretty cool-looking, but Nike shared no tangible benefits so parking this one in the optional lot.
05/08/2024 Optional OpenAI says it’s building a tool to let content creators opt out of AI training The tool isn't ready and this, of course, comes after they've already built their LLMs using everyone's data without permission.
05/08/2024 Optional OpenAI strikes licensing deal with the magazine giant behind People Another deal, another day. If by some chance, the courts make OpenAI trash the models they already built because they used content without permission, they're building up their training data stable. Also, if OpenAI is building a search engine as is rumored, these deals will be helpful.
05/09/2024 Essential Google DeepMind debuts huge AlphaFold update and free proteomics-as-a-service web app The most significant improvement in this newest version of AlphaFold is that researchers can now model how a synthesized protein might interact with other proteins or even with RNA and DNA strands. That speeds up an important step in drug discovery since protein interaction is what makes things happen. We only wish they were open sourcing the full model - they're only providing researchers with free access to a limited version.
05/09/2024 Essential Microsoft is ‘turning everyone into a prompt engineer’ with new Copilot AI features Copilot will soon have an autocomplete feature to help people create more effective prompts. We're flagging this as an essential development because users still struggle to create prompts that unlock the power of GenAI. This should help adoption - and it's also another step toward automating prompt engineering. For the AI leaders dealing with skeptics, this should bolster their case for adoption.
05/09/2024 Important ServiceNow and Microsoft expand strategic alliance, combining generative AI capabilities to enhance choice and flexibility Two GenAI systems (Copilot and ServiceNow's Assist) interacting and handing off tasks to each other sounds like the early stages of agentic systems to us so we're marking this one important.
05/09/2024 Optional Introducing the Model Spec This article from OpenAI shares the principles they follow to guide GenAI's behavior. It offers guidance and opens interesting questions to any organization training models. Determining how we make these systems reflect societal norms and values is not easy and is hotly debated. For example, should a system convince a flat-earther that our planet is round? According to this model spec, the answer is 'no.'
05/09/2024 Optional 1 in 6 Contact Centers Have Deployed GenAI Capabilities, Finds Deloitte A lot of data cherry picking here to make the case that every organization should use GenAI in customer service. While we believe that is the right move, this survey doesn't really do a great job at proving the point. We are surprised, however, that only 1 in 6 service centers are using GenAI at this point.
05/09/2024 Optional Holiday decor retailer Balsam Brands analyzes data with generative AI This is a republished press release from SoundCommerce, one of the providers Balsam is tapping along with Snowflake. And it's not even clear how they'll use GenAI. Skip it.
05/10/2024 Essential Brands are unleashing generative AI design tools for customers Our analysts got very excited about this article that shares how Reebok, Adore Me and others are letting consumers generate their own designs with GenAI and then providing them with digital versions to use in games and other online forums as well as physical versions in some cases. Personalization at scale is finally arriving.
05/10/2024 Important Leaked Deck Reveals How OpenAI Is Pitching Publisher Partnerships This deck seems to confirm that OpenAI is making a play in search. It offers publishers who strike a licensing deal with OpenAI a few options for linking to their content in ChatGPT responses. From the article: "In an anchor treatment, branded, clickable buttons appear below ChatGPT’s response to a user query. And the in-line product inserts a pullquote into the text of ChatGPT’s response, whose font is larger and includes a clickable, branded link." Game on, Google.
05/10/2024 Important SoundHound AI and Perplexity Partner to Bring Online LLMs to Next Gen Voice Assistants Across Cars and IoT Devices SoundHound is an established AI-powered voice recognition provider so this is a nice move for Perplexity whose GenAI-driven Q&A service has been gaining traction. SoundHound is using it to offer real-time info to its customers' products. GenAI is gradually seeping into many areas.
05/10/2024 Important Meet AdVon, the AI-Powered Content Monster Infecting the Media Industry Well-researched article on the way some notable publishers (eg Sports Illustrated, Us, LA Times) have used services like AdVon to create zero-value clickbait content from fake writers that drives traffic to publishers' sites. As you can imagine, GenAI is supercharging such services. Reader beware...
05/10/2024 Optional California to tap generative AI tools to increase services access, reduce traffic jams Nice to see innovation from the government sector. They're fielding proposals from OpenAI and others at the cost of only $1 (we thought this was a typo at first!). But not of significant note to the AI leader.
05/10/2024 Optional Introducing Command R Fine-Tuning: Industry-Leading Performance at a Fraction of the Cost Cohere continues its recent positioning as a low-cost LLM alternative by showing how its smaller model (R+ is its bigger one) can offer performance similar to larger models for some tasks like summarization when fine-tuned. AI leaders are likely already aware that fine-tuning smaller models can do this so marking as optional.
05/13/2024 Essential AI at Work Is Here. Now Comes the Hard Part This research from Microsoft is of course self-serving but there are lots of interesting stats on GenAI usage in business, how it might affect roles and more. Plenty of hard data to support you as you look to implement AI in your organization. Our analyst Adam Rappaport created a CustomGPT to help you explore the report.
05/13/2024 Important How AI Has Already Begun to Change These Workers’ Jobs Nice piece from the WSJ showing how people in various industries (e.g., real estate, healthcare, entertainment) are using GenAI in their work. We didn't elevate to essential because these are individual uses rather than at-scale implementations across an org, but still worth a read.
05/13/2024 Important Prompting assistance with Claude As we know, prompting is tricky, and people often are ineffective at it, which makes them forego using GenAI. This is a move in the direction of taking prompting out of the equation, though it doesn't quite go that far. Still helpful and another way Claude is showing itself worthy of being in the SoA LLM tier.
05/13/2024 Important Agent Hospital: A Simulacrum of Hospital with Evolvable Medical Agents The researchers use simulation to train a model as opposed to fine-tuning or other common methods and achieve good results. Could be a method to train when there's limited data available. Worth a read.
05/13/2024 Optional RadOnc-GPT: Leveraging Meta Llama for a pioneering radiation oncology model We almost went to important on this Llama-2 based model but ultimately landed on optional because it was vague on how/whether the model is actually in use.
05/13/2024 Optional Triomics taps LLMs to accelerate cancer care, raises $15M While Y Combinator is in on this funding round, it still amounts to a funding announcement. The company notes it's working with a handful of companies but none are named and a handful is, well, just a handful. Let's see how this pans out further down the line, though, since cancer care is a hugely important area.
05/14/2024 Essential Hello GPT-4o OpenAI's post about their latest updates is an absolute must read (and 'view' of their videos) for the AI leader.
05/14/2024 Important OpenAI’s custom GPT Store is now open to all for free. The GPTs are now available to anyone to use, though still only Plus subscribers can create a GPT.
05/14/2024 Important Alibaba says its Tongyi Qianwen AI models are used by over 90,000 corporate clients in China. Anyone staying up to date on the latest AI research knows that China is very much alive and well as a leader in the space.
05/14/2024 Important The consequences of generative AI for online knowledge communities. Interesting study that looks at the implications in the decline in usage of Stack Overflow.
05/14/2024 Optional Nvidia launches quantum computer centers with CUDA-Q platform. The intersection of AI and quantum computing is certainly a space to keep an eye on, but this particular news is not of note to the busy AI leader.
05/14/2024 Optional Exploration-focused training lets robotics AI immediately handle new tasks. A group of Northwestern University researchers may be onto a method that would help robots learn more efficiently.
05/15/2024 Essential Google reveals plans for upgrading AI in the real world through Gemini Live at Google I/O 2024 Did Google Live look a lot like what we saw from OpenAI's 4o earlier this week? Yes. Is it available yet? No. While frustrating that Google did a lot of previewing of new offerings and functionality that have no clear time line for availability other than 'this year,' it's still important for the AI leader to understand what each of the ponies in what's increasingly a two-horse race offers.
05/15/2024 Essential Google I/O 2024: Everything announced so far This article goes through the full laundry list of GenAI-powered offerings coming from Google, making it tough to absorb, but the AI leader needs to go through it. Our analysts all had different favorites among the upcoming features (watch the first 10 minutes of today's briefing to see what they were). But one thing we agreed on - a future where our workspace is fully AI-powered, personalized, and completely controlled through voice interaction is now a real and exciting possibility. Our youngest grandkids may laugh at us when we show them a keyboard. One thing we didn't mention on our morning briefing but should have: the expected price wars are now in full swing. Both OpenAI and Google have dropped pricing this week to attract users.
05/15/2024 Important How Generative AI Is Remaking UI/UX Design This piece from Andreesen shares the ways that generative AI is making the app development workflow easier as designs get translated into code. Most AI leaders do double duty as CIO or CTO so it's important to be aware of the tools now available for this. As they often do in these types of pieces, Andreesen provides a great chart showing the tools available for each stage of the process.
05/15/2024 Optional Fintech firm Klarna says 90% of its employees are using generative AI daily Not sure we believe this figure, but even if it's real, the article isn't worth reading. Stick to the headline.
05/15/2024 Optional Women Leaders in Tech Outpace Men Counterparts in Generative AI Adoption We had a visceral reaction to this (including from our Luda Kopeikina.) First of all, the overall difference between female and male users was 68% vs 66%. Second, there are so many more amazing accomplishments by women in tech and places where they outshine men to focus on. C'mon, BCG. Your content is typically better than this.
05/16/2024 Important Microsoft's carbon emissions have increase 30% since 2023 due to data center expansion While the emissions increase is attributed to building data centers as opposed to their use, it's likely some strategic carbon accounting and clean energy credit purchases offsetting the running of their data centers which have surely cranked up this year due to GenAI's energy demands. The industry needs to solve for this before regulators step in further than they already have which could slow GenAI development and adoption. And AI leaders will need to track their own emissions through cloud use for CSRD and potentially soon for US regulators.
05/16/2024 Optional Deloitte Digital Introduces CreativEdge: A Generative AI-powered, Omnichannel, Content Creation Tool That Can Revolutionize Marketing Great that they've designed this solution but dozens of startups offer similar ones that you can get without paying for consulting services.
05/16/2024 Optional TD Bank launches generative AI platforms for contact centres, engineering teams This is an announcement about pilots rather than production use cases and they're in what's become table stakes areas - coding and customer service.
05/16/2024 Optional Verizon using generative AI to do customer service better Another call center implementation. At least this one is in production. Still kept it as optional since all you need to know is the headline.
05/16/2024 Optional Weka raises $140M as the AI boom bolsters data platforms Worth taking a look at the data pipeline offering from this established player with 11 of the Fortune 50 as clients but the article is no more than a funding announcement.
05/16/2024 Optional Anthropic hires Instagram co-founder as head of product After a bevy of announcements from Google and OpenAI this week, we get this from Anthropic. Good to see that they'll be focused on productizing some offerings going forward. We increasingly hear of companies and individual users exploring Anthropic's offerings given the SoA performance of its LLMs.
05/16/2024 Optional Moving past gen AI’s honeymoon phase A comprehensive piece on the challenges CIOs and other AI leaders face in implementing GenAI at scale with a nice architecture diagram but not enough detail on how to address them to warrant an important rating.
05/17/2024 Important OpenAI chief scientist Ilya Sutskever is officially leaving Signals a period of leadership upheaval at OpenAI - watch this development closely especially if you are building applications based on ChatGPT products. Also, we believe that it is a signal of OpenAI strongly focusing on commercial side versus its original mission of creating Artificial General Intelligence (AGI) model that benefits all of humanity.
05/17/2024 Important Industrial Renaissance: Software Transforming the Physical World A somewhat marketing article from Bain Capital highlighting how AI is impacting software stacks integrated with hardware for industrial applications. This is an important read for supply chain, warehousing, robotics and other companies in the industrial space.
05/17/2024 Optional Nasdaq Enhances Global Market Surveillance with GenAI Nasdaq has earned respect for applying advanced analytics and machine learning in ensuring investors' trust in capital markets. It is continuing on this path with deployment of new AI surveillance capabilities based on Amazon Bedrock, an AWS service for building secure generative AI applications.
05/17/2024 Optional AI Talks Leave ‘Little Tech’ Out An opinion from WSJ worthy of note. It argues that imposing strict government AI regulations will favor the current AI leaders such as large companies Microsoft and Google, creating barriers for smaller start-ups.
05/17/2024 Optional Hugging Face is sharing $10 million worth of compute to help beat the big AI companies We applaud Hugging Face for the gesture. However, how far will it go to help the open source community and is it really a marketing piece?
05/17/2024 Optional Aurora Supercomputer Ranks Fastest for AI Another milestone reached by Intel - it has broken the exascale barrier at 1.012 exaflops and is the fastest AI system in the world dedicated to AI for open science, achieving 10.6 AI exaflops. Early success stories using the system from Argonne National Laboratory include mapping the human brain’s 80 billion neurons, high-energy particle physics enhanced by deep learning, and drug design and discovery accelerated by machine learning, among others.
05/20/2024 Essential Slack users horrified to discover messages used for AI training Slack's policies seems to leave the door open for them to use customer data to train their models though they claim they don't. And you have to opt out via your Slack administrator rather than opt in. The confusion has caused some users to get angry and point out the discrepancies. Essential for AI leaders (and any organizational leader) to understand if their private data is being used for training by their service providers so we marked this essential (and here we provide a better article about the issue than the one we reviewed on our briefing this morning).
05/20/2024 Important Gen AI and cloud optimization help Asian SuperApp Grab turn a profit This Singaporean company (which is akin to Uber in the US) has cut 18% of its staff citing GenAI as at least part of the impetus. And, it says, thanks to this and the performance of GenAI, they've turned a profit for the first time. The company says content creation went from taking 99 hours to 90 minutes. Important for AI leaders to see that GenAI truly drives such efficiency. And important for everyone to understand the need to upskill to avoid being one of the employees falling victim to cuts.
05/20/2024 Important Replit cuts staff by 30 amid aggressive AI push in software development This is important for the same reason as the previous article - 16% of staff cut here due to GenAI.
05/20/2024 Optional Law firms start generative AI rollout, with safeguards Of note for anyone in the legal space but optional for others. One good nugget from this one, though - a major law firm in Australia offered a significant $20K 'prize' for employees to submit use cases for GenAI which helped them find killer use cases. An approach any AI leader could consider.
05/20/2024 Optional RLHF Workflow: From Reward Modeling to Online RLHF We marked this as optional for now because we want to see if this training method has legs. It's one to watch - rather than using human annotators, it uses AI to provide feedback on LLM responses and achieves some pretty good results. Could offer a way for companies that might not have the human resources for human-led RLHF.
05/20/2024 Optional Microsoft risks billions in fines as EU investigates its generative AI disclosures Microsoft missed an EU deadline for providing information on AI safety and was given an extension and a warning that if they don't deliver, they could be fined. Not an essential read.
05/20/2024 Optional OpenAI researcher resigns, claiming safety has taken ‘a backseat to shiny products’ We only marked this as optional because we already flagged the departure of safety lead Ilya Sutskever as important on Friday. This is another departure from the same team. Here's another good article talking about all the tumult at OpenAI around this if you're interested.
05/21/2024 Essential Microsoft wants to make Windows an AI operating system, launches Copilot+ PCs We knew something like this would be coming and here it is. 40 generative AI models power this new OS and deliver some great capabilities. The AI can view what you're doing on-screen and you can interact with it verbally in natural language in real-time as you work. And you get capabilities that are performed locally so your data is safe. No doubt this is an essential development.
05/21/2024 Essential Google’s broken link to the web Great commentary about the slow death of search as we know it and the implications for creators, publishers and any brand that relies on traditional search advertising (which is pretty much every brand). Anyone in business needs to track the trends here and should be preparing for changes now.
05/21/2024 Important Scarlett Johansson told OpenAI not to use her voice — and she’s not happy they might have anyway We believe OpenAI's contention that this is not Scarlett's voice. However, whether they intended the voice to sound like her or not, Altman made enough overtures to 'Her' to imply that they may have. The implications for companies using AI-generated voice are clear - this is tricky territory with real implications. AI leaders must be prepared to navigate the use of voice in their apps carefully.
05/21/2024 Important Dell Technologies building AI Factory with Nvidia, growing AI efforts with Hugging Face, Meta and Microsoft If you're planning to infuse, for example, customer service with GenAI capabilities, you need to think about your infrastructure and Dell has a smart implementation here. Worth noting.
05/21/2024 Optional INDUS: Effective and Efficient Language Models for Scientific Applications As our John Sviokla noted, this is 'good for the geeks' in that models can get smart on science, but this is optional for most AI leaders.
05/21/2024 Optional ChatGPT’s mobile app revenue saw its biggest spike yet following GPT-4o launch An interesting and smart move by OpenAI to offer access to its new 4o model for free on desktop but require a subscription to the mobile app. They generated $4.2M in revenue from this in less than a week. But that's all you need to know here.
05/21/2024 Optional AI is already changing management — companies must decide how We love Ethan Mollick and consider him a true GenAI thought leader but his thoughts in this editorial are more theoretical than actionable. There's a lot better content from him that's worth spending time on.
05/22/2024 Essential Introducing Tako, a new way to reference real knowledge And our first integration, Perplexity Tako, an advanced knowledge search and visualization engine, is partnering with Perplexity's answer engine, creating a revolutionary capability that will make many applications across the enterprise visual and user friendly. Amazing combination of real data with visualization.
05/22/2024 Important IBM Unveils Next Chapter of watsonx with Open Source, Product & Ecosystem Innovations to Drive Enterprise AI at Scale IBM announces its commitment to Open Source by releasing a set of advanced LLMs into open source, partnering with Red Hat, integrating AI into products and consulting, and strengthening partnerships with many leading AI providers, such as Microsoft, AWS, Meta, SAP and Salesforce. Essential for IBM existing customers.
05/22/2024 Important Five ways criminals are using AI Generative AI provides a new, powerful tool kit that allows malicious actors to work far more efficiently and internationally than ever before. An essential read for leaders to understand the next wave of malicious attacks to get prepared for.
05/22/2024 Important Microsoft Build 2024: everything announced Microsoft announced a set of cool new capabilities at its Build 2024 event, expanding Copilot with autonomous agents and adding AI to Windows clipboard. The company also rolled out Phi-3-vision, a new version of the Phi-3 AI model it announced in April. It’s multimodal and can read text and look at pictures, but it’s a small language model that’s compact enough to work on a mobile device.
05/22/2024 Important Mapping the Mind of a Large Language Model Vital new work from Anthropic in model transparency and thus, increased safety. Using 'dictionary learning' researchers successfully extracted features from the middle layer of Claude 3.0 Sonnet, providing a rough conceptual map of its internal states halfway through its computation. This is the first ever detailed look inside a modern, production-grade large language model that will provide insight on increasing models' safety.
05/22/2024 Optional Data-labeling startup Scale AI raises $1B as valuation doubles to $13.8B A funding announcement from Scale AI working in a vital area of providing data-labeling services to companies that want to train machine learning models.
05/23/2024 Essential OpenAI, WSJ Owner News Corp Strike Content Deal Valued at Over $250 Million The article highlights two mega shifts happening in the media industry. One, the power of distribution is winning over content creation demonstrated by the low number, $50MM a year, that WSJ will get from the partnership. Second, a fight for traffic and advertising revenue is on. Publishers are concerned that AI-powered search tools, such as Bing/OpenAI and Google, will serve up complete answers based on news content, eliminating a user's need to click on an article link and depriving publishers of traffic and advertising revenue. How will this fight play out?
05/23/2024 Important Nvidia’s Sales Triple, Signaling AI Boom’s Staying Power Nvidia reports impressive financial results, revenue and profit, due to increased demand for its AI chips, - signaling further significant AI growth. Nvidia has already made several shrewd acquisitions. With additional money this trend is likely to continue making Nvidia hard to dislodge as a leader in AI hardware.
05/23/2024 Important Meta AI chief says large language models will not reach human intelligence A well written article presenting an opinion from a well recognized technical leader in GenAI stating that LLMs will not reach human intelligence due to their innate technical limitations. This opinion goes counter to a pervasive belief that, since LLMs are so good at so many things, they will be as humans at some point. Certainly worth reading to understand the limitations of the current LLMs.
05/23/2024 Optional Mastercard Doubles Speed of Fraud Detection with Generative AI Mastercard reports faster, better fraud detection with GenAI reconfirming its previous statements in this regard. It is interesting to note that their GenAI model is able to predict the behavior of individuals and groups.
05/23/2024 Optional Gen Z loves generative AI-powered customer service chat: Affirm CEO Believable assertion by Affirm's CEO that GenZ prefers robot customer help, - confirming that, when done right, AI assistants can get to the 'meat of the matter' faster and provide the needed answer.
05/23/2024 Optional Snowflake acquires TruEra to deliver LLM observability inside data cloud With this acquisition, Snowflake, an AI cloud data provider, is bolstering its efforts to give its customers an end-to-end platform to build generative and predictive AI applications using the data hosted on their Snowflake data cloud. TruEra will add tools to test, debug and monitor machine language (ML) models and large language model (LLM) apps in production. This is the third Snowflake's acquisition in the observability space.
05/24/2024 Essential The Foundation Model Transparency Index A comprehensive evaluation index for foundation models enabling leaders to have more insight into models before selecting the right one for their projects.
05/24/2024 Essential JPMorgan's private bank is launching a generative AI tool. A top exec walks us through her tech strategy and how she wants to transform bankers' jobs. Great example of a leading large firm transforming a key business using AI. Note that the person leading the effort does not have the tech background but is highly knowledgeable in the processes of the business being transformed.
05/24/2024 Important Google Cloud Consulting launches Generative AI Ops to assist in enterprise AI deployments Even though it is a product announcement, we rate this article as important because it provides in-depth insight into steps, such as prompt engineering, design and optimization, that are necessary in order to take an idea for using GenAI into production and ensure high-quality, accurate outputs.
05/24/2024 Important The Guide to AI Agents AI agents will play a critical role in the future of AI. This brief but comprehensive article written for non-techies provides a great overview of AI agents. Certainly worth a read, especially when it is coming from two co-founders who had senior positions at Salesforce, Facebook and Google. Bret Taylor serves on the board of OpenAI.
05/24/2024 Optional Airline to 'better manage' flights with AI use A use case article on using AI for scheduling and other tasks to optimize airline efficiency.
05/24/2024 Optional Truecaller partners with Microsoft to let its AI respond to calls in your own voice A product announcement from Truecaller, the widely known caller ID service app, offering its customers a feature to respond to calls in their own voice. Watch this trend of adding every individual's voice to the robots that can change the nature of interaction.
05/28/2024 Essential Financial Statement Analysis with Large Language Models With chain of thought prompting, GPT-4 outperformed financial analysts in predicting the direction of future earnings and was on par (and better in delivering context) than SoA ML prediction models. This dispels the myth that LLMs are never good at numerical tasks and reasoning - clearly they're quite capable in some use cases with the right set up. A must read for any AI leader.
05/28/2024 Important NAB plots major AI strategy to roll out in three key areas Good article about the ways a major Australian bank is piloting LLMs beyond customer service applications, including in assessing customer complaints, helping paralegals review trust deeds (in minutes vs. human time of 45 minutes) and personalizing product offerings.
05/28/2024 Optional Elon Musk’s xAI raises $6 billion As our Luda Kopeikina said, 'billion is becoming the new million' in funding raises for LLM-related companies. Impressive raise, likely helped by the Musk name and the amount of data his company is sitting on through X. But we'll save important or essential ratings for when these investments bear fruit.
05/28/2024 Optional Meta and Elon Musk’s xAI fight to partner with chatbot group Character.aiLaw firms start generative AI rollout, with safeguards Not a huge surprise that two powerhouses are circling the wagons on the second-most used GenAI app. Meta already has a similar play to Character's offering. Our co-founders placed a friendly wager on whether both Character.ai and Perplexity will exist or be snapped up in the next year - Paul believes they will be acquired whereas John believes they won't. We shall see....
05/28/2024 Optional Why WPP is adding Anthropic’s Claude models to its AI platform Not enough meat on the bones of this announcement to warrant a read. Note that this owner of many PR/marketing companies like Ogilvy is using Amazon Bedrock and tapping Claude and other models as opposed to using just Claude.
05/28/2024 Optional Understanding the Cost of Generative AI Models in Production We like the idea of this article, which costs out LLM implementation, but disagree with the analysis. It posits that you don't need an engineer if you're buying vs. building which is simply not true.
05/29/2024 Important AI on Trial: Legal Models Hallucinate in 1 out of 6 Queries Stanford HAI tested two legal GenAI apps - one from Thomson Reuters and another from LexusNexus - that claim to be hallucination free. Turns out they're not. They both use RAG, and while the technique minimizes hallucinations, it's not foolproof. To drive that message home, we marked this article as important.
05/29/2024 Important EU’s ChatGPT taskforce offers first look at detangling the AI chatbot’s privacy compliance While the task force didn't come to concrete conclusions as to whether ChatGPT violates portions of the GDPR, this article provides good detail on the particular tenets of the law that LLM providers need to take into account. For the busy AI leader, this is a worthy read to understand how the LLMs they're using could at some point be impacted by the law.
05/29/2024 Important Retail and Gen AI: Now Scale Those Terrific Early Returns Hyperbolic headline and unhelpful 'group your use cases' advice aside, there are two good charts in here that provide use cases applicable to many industries in addition to retail so worth checking those out.
05/29/2024 Optional OpenAI Board Forms Safety and Security Committee They had to do something after disbanding the safety committee. They've put a lot of people on the new committee but the key person on it to note is Sam Altman. Will his voice matter most? And what will that translate into?
05/29/2024 Optional Jan Leike joins Anthropic Speaking of that disbanded OpenAI safety team, one of its leaders has jumped over to Anthropic, which has shown that safety is a priority (though transparency is apparently not given that, as is the case with their competitors, they've never revealed what data they used to train Claude - we can guess...).
05/29/2024 Optional How Airbus uses generative artificial intelligence to reinvent itself Airbus has identified 600 potential GenAI use cases, but provide little information on what they are and which ones they have in production so this article isn't worth your valuable time.
05/30/2024 Essential AI Integration and Modularization We had a lively discussion about this article by independent analyst Ben Thompson (starts at 13:25 in the recording of today's news briefing) that ponders whether vertically integrated AI players (Google is the only AI player with a fully integrated stack from chips through apps right now) will have a competitive advantage over partially integrated (eg, Microsoft) or horizontal (eg, Amazon) players in the great AI race. You may or may not agree with all of his contentions, but the article is well worth the read.
05/30/2024 Optional Vox Media and The Atlantic sign content deals with OpenAI While this article is optional, it's another move related to the evolution of search - remember that those publishers striking deals will have their content linked to in ChatGPT.
05/30/2024 Optional Mistral releases Codestral, its first generative AI model for code Mistral had to do this to remain competitive with other LLM providers, but they've slapped a 'no commercial use' restriction on the model. Given all the other options for code generation, we marked this news as optional.
05/30/2024 Optional OpenAI signs 100K PwC workers to ChatGPT’s enterprise tier as PwC becomes its first resale partner This type of arrangement between a consultancy and software provider is not unusual though it's yet another move that puts Microsoft and its partner OpenAI in direct competition. A development to note but the article is not an essential read.
05/30/2024 Optional Announcing Sonic: A Low-Latency Voice Model for Lifelike Speech The makers of Mamba introduce a state space model for audio generation. Our Adam Rappaport found the performance decent but not quite as good as current transformer-based competitors like Eleven Labs.
05/30/2024 Optional The Crossroads of Innovation and Privacy: Private Synthetic Data for Generative AI This blog post by Microsoft discusses three differential private synthetic data generation techniques. An interesting evaluation and good for the AI leader to be aware of these possibilities but perhaps deeper than the busy AI leader needs to go.
05/31/2024 Important Perplexity AI’s new feature will turn your searches into shareable pages We don't rate new features in an existing tool as important very often, but this one warrants it - in fact we had one vote for essential. This feature is already being rolled out, and our David De Lallo gave it a test drive on our briefing this morning (minute 8:20 here). It's quite good for an initial release. And the web pages you produce can be published and are crawlable by search engines. The tool itself is another big step in the search revolution that's unfolding before us. Implications of this sea change for any company that relies on search for traffic and sales are huge.
05/31/2024 Important Anthropic’s AI now lets you create bots to work for you Another feature announcement that we think is worth noting. Via any API a developer can create an AI agent that performs tasks. One example offered is creating a tool for an interior designer that processes images of someone's house and other data to offer personalized room decor recommendations. AI agents that perform tasks on your behalf are here, earlier than many predicted.
05/31/2024 Optional Tech giants form an industry group to help develop next-gen AI chip components Microsoft, Meta, Google, AMD, Broadcom and several other big tech players are coming together to create a common standard for the components that link GPUs. Nvidia didn't join the party, likely since they have their own connectors and are the clear market leader in he data center. No Amazon either. Of note, but the article isn't an important read.
05/31/2024 Optional Transformers Can Do Arithmetic with the Right Embeddings This research article goes deep on how the authors used embeddings to improve LLMs' abilities to do addition and multiplication. We're seeing good progress in improving LLM performance in math. A positive development but no need to read this one unless you feel like geeking out a bit this weekend.
05/31/2024 Optional Generative AI predicts hospital admissions from ED visits Researchers at Mount Sinai ran a study using more than 800,000 patient EHR records to see how GPT-4 performed in predicting whether a patient coming into the ER would need to be admitted. They got decent performance out of the box (77.5% accuracy) and improved performance by giving the LLM a little bit of clinical data (83.1% accuracy). Promising but only an essential read for those in healthcare and health insurance.
05/31/2024 Optional How Italian CIOs produce value with gen AI This article covers how some Italian companies are using LLMs, but they're all the most common use cases - customer service, internal knowledge tool, coding assistants - and the piece offers little detail or actionable insights so you can skip this one.
06/03/2024 Important What We Learned from a Year of Building with LLMs (Part I) This article published by O'Reilly is not news per se, but it provides some tactics the authors have found to be successful when implementing LLMs in the enterprise. Best practices on RAG, building AI agents, evaluations and more can provide real practicable advice for the AI leaders' team. This is one to read when you have time and forward to your workers.
06/03/2024 Important Singapore Publishes Generative AI Model Governance Framework Our important rating on this one is because it provides a crisp summary of a framework AI leaders can use in their own organization as opposed to the news of Singapore publishing the framework. A good one to read to either serve as a checklist for your own framework or to help you get started on creating one if you don't have one already.
06/03/2024 Optional Dell earnings reveal sluggish enterprise AI adoption We selected this article because it goes into Dell leaderships' discussion about why they think GenAI pickup is slow. We see it as an excuse for an earnings miss - slowness is expected given it's still early days. We had a lively discussion about Dell's AI strategy (starts around minute 1:40) which is similar to its cloud strategy (or lack of one) - they're targeting on-prem infrastructure. We don't think it's wise to plow resources into on-prem right now given the early days and pace at which things are advancing.
06/03/2024 Optional Secret Cyborgs and Their AI Shadows: Navigating the Copilot+ PCs Frontier This articles poses that the forthcoming AI-infused PCs introduce new risks for shadow AI usage in the enterprise. It doesn't make an effective argument - most of what's discussed mirrors risks for any organization that issues computers to workers (which is most orgs). And if you bought AI PCs for your employees, doesn't this signal you want them to use AI (as you should)?
06/03/2024 Optional Meet Showrunner, The ‘Netflix Of AI’ That Turns Viewers Into TV Show Creators While this company's product is certainly innovative - they publish animated series that users can create new episodes for - we've known about it for a while and it's still not widely available.
06/03/2024 Optional Sony Pictures Uses AI to Cut Film Costs Yes, we know that studios are going to use AI for voiceover, post-production and more - as much as they can within the bounds of labor agreements. However, it's worth clicking this link to see Perplexity Pages in action - they're using it to create their newsletter from which we pulled this.
06/03/2024 Optional FineWeb: decanting the web for the finest text data at scale The authors detail a technique they used to create what they've deemed (and some evals support) a better dataset for training than those currently available. Perhaps worth forwarding to your team for a look but too in the weeds to warrant a read by the busy AI leader.
05/30/2024 Important What We Learned from a Year of Building with LLMs (Part II) Yesterday we rated the first edition of this series on LLM implementation tactics as important and we give the same rating to the second edition which covers operations. While the tactics edition was more geared toward developers, this one has points that both AI leaders and their teams will find valuable.
05/30/2024 Important The Agent Development Life Cycle This article co-authored by Bret Taylor is clearly making a case for using his company Sierra's SDK. However, what Sierra is solving for is of note - standing up an LLM application is quite different from what developers are used to. Sierra's tools attempt to make it more similar for them.
05/30/2024 Important Interesting insights on NVIDIA and what we can expect from AI in future This is a thoughtful X post by investor Nic Carter that's wide ranging in what it covers. The implications of power moving from labor to capital, the US's place in the world in the AI-driven future, energy implications and more. A worthy read.
05/30/2024 Optional Generative AI is now scanning your Amazon packages for defects before they get shipped out Amazon's use case is of note, but the article provides little information about it beyond the headline so no need to read it.
05/30/2024 Optional Clarity may be emerging in AI capabilities pricing. Here's how AI pricing can be confusing right now, but so is this article. Skip it.
05/30/2024 Optional Introducing Project G-Assist: A Preview Of How AI Assistants Can Enhance Games & Apps The takeaway of this one is all you need to know - AI agents that can see our screens and interact with us about what we're doing on them are coming. First in Microsoft's forthcoming AI-powered PCs and, apparently, soon from Nvidia.
05/30/2024 Important What We Learned from a Year of Building with LLMs (Part II) Yesterday we rated the first edition of this series on LLM implementation tactics as important and we give the same rating to the second edition which covers operations. While the tactics edition was more geared toward developers, this one has points that both AI leaders and their teams will find valuable.
05/30/2024 Important The Agent Development Life Cycle This article co-authored by Bret Taylor is clearly making a case for using his company Sierra's SDK. However, what Sierra is solving for is of note - standing up an LLM application is quite different from what developers are used to. Sierra's tools attempt to make it more similar for them.
05/30/2024 Important Interesting insights on NVIDIA and what we can expect from AI in future This is a thoughtful X post by investor Nic Carter that's wide ranging in what it covers. The implications of power moving from labor to capital, the US's place in the world in the AI-driven future, energy implications and more. A worthy read.
05/30/2024 Optional Generative AI is now scanning your Amazon packages for defects before they get shipped out Amazon's use case is of note, but the article provides little information about it beyond the headline so no need to read it.
05/30/2024 Optional Clarity may be emerging in AI capabilities pricing. Here's how AI pricing can be confusing right now, but so is this article. Skip it.
05/30/2024 Optional Introducing Project G-Assist: A Preview Of How AI Assistants Can Enhance Games & Apps The takeaway of this one is all you need to know - AI agents that can see our screens and interact with us about what we're doing on them are coming. First in Microsoft's forthcoming AI-powered PCs and, apparently, soon from Nvidia.
06/04/2024 Important What We Learned from a Year of Building with LLMs (Part II) Yesterday we rated the first edition of this series on LLM implementation tactics as important and we give the same rating to the second edition which covers operations. While the tactics edition was more geared toward developers, this one has points that both AI leaders and their teams will find valuable.
06/04/2024 Important The Agent Development Life Cycle This article co-authored by Bret Taylor is clearly making a case for using his company Sierra's SDK. However, what Sierra is solving for is of note - standing up an LLM application is quite different from what developers are used to. Sierra's tools attempt to make it more similar for them.
06/04/2024 Important Interesting insights on NVIDIA and what we can expect from AI in future This is a thoughtful X post by investor Nic Carter that's wide ranging in what it covers. The implications of power moving from labor to capital, the US's place in the world in the AI-driven future, energy implications and more. A worthy read.
06/04/2024 Optional Generative AI is now scanning your Amazon packages for defects before they get shipped out Amazon's use case is of note, but the article provides little information about it beyond the headline so no need to read it.
06/04/2024 Optional Clarity may be emerging in AI capabilities pricing. Here's how AI pricing can be confusing right now, but so is this article. Skip it.
06/04/2024 Optional Introducing Project G-Assist: A Preview Of How AI Assistants Can Enhance Games & Apps The takeaway of this one is all you need to know - AI agents that can see our screens and interact with us about what we're doing on them are coming. First in Microsoft's forthcoming AI-powered PCs and, apparently, soon from Nvidia.
06/04/2024 Important Snowflake Data Cloud Summit 2024: The biggest developments announced Data is, of course, a vital element of GenAI, and that's where Snowflake plays. Announcements included making Iceberg Tables generally available, open sourcing Polaris in the coming weeks, and enhancing the Cortex AI offering. Worth reading through these updates.
03/06/2024 Important True Fit leverages generative AI to help online shoppers find clothes that fit This company provides sizing data and tools to retailers like Urban Outfitters and JCPenney. It's enhancing these tools with GenAI and lays out a roadmap for more GenAI-enabled offerings to come. Nice use case story.
06/04/2024 Optional US delays AI chip exports to Middle East by Nvidia, AMD over concern that China can access the tech via data centres While the topic covered in these pieces is one for any executive in a company that relies on semis (which is many at this point) to monitor for digital supply chain resiliency, these articles aren't necessarily the ones to read to do so.
06/04/2024 Optional Vinod Khosla, Marc Andreessen And The Billionaire Battle For AI's Future This article discusses the competing perspectives among two famous tech investors on open source (supported by Andreessen v. closed source (supported by Khosla) LLMs. A somewhat entertaining read, but it doesn't introduce any new ideas into the debate.
06/04/2024 Optional Databricks to Buy Data-Management Startup Tabular in Bid for AI Clients Databricks continues its acquisition spree with its third purchase of the year. Tabular makes a cloud storage technology that is built on Iceberg. A notable development as this company continues to compete with Snowflake in the data space, but we don't think the article is an essential read.
06/05/2024 Important Introducing Generative Physical AI In this video, Nvidia shares details about its Isaac robotics platform, which leverages GenAI and reinforcement learning to help companies create autonomous robots that can perform numerous tasks by training them in a virtual environment. Worth taking 3 minutes to view.
06/05/2024 Important Snowflake Data Cloud Summit 2024: The biggest developments announced Data is, of course, a vital element of GenAI, and that's where Snowflake plays. Announcements included making Iceberg Tables generally available, open sourcing Polaris in the coming weeks, and enhancing the Cortex AI offering. Worth reading through these updates.
06/05/2024 Important True Fit leverages generative AI to help online shoppers find clothes that fit This company provides sizing data and tools to retailers like Urban Outfitters and JCPenney. It's enhancing these tools with GenAI and lays out a roadmap for more GenAI-enabled offerings to come. Nice use case story.
06/05/2024 Optional US delays AI chip exports to Middle East by Nvidia, AMD over concern that China can access the tech via data centres While the topic covered in these pieces is one for any executive in a company that relies on semis (which is many at this point) to monitor for digital supply chain resiliency, these articles aren't necessarily the ones to read to do so.
06/05/2024 Optional Vinod Khosla, Marc Andreessen And The Billionaire Battle For AI's Future This article discusses the competing perspectives among two famous tech investors on open source (supported by Andreessen v. closed source (supported by Khosla) LLMs. A somewhat entertaining read, but it doesn't introduce any new ideas into the debate.
06/05/2024 Optional Databricks to Buy Data-Management Startup Tabular in Bid for AI Clients Databricks continues its acquisition spree with its third purchase of the year. Tabular makes a cloud storage technology that is built on Iceberg. A notable development as this company continues to compete with Snowflake in the data space, but we don't think the article is an essential read.
06/06/2024 Important Navigating the generative AI disruption in software This article provides an analysis of how GenAI is impacting software providers today and projects effects going forward. It notes that GenAI is likely to increase tech spend but also increase vendor switching, with these trends already beginning. Worth at least a look through the charts.
06/06/2024 Important Torrens University leverages generative AI to uplift its online learning experience While this is a case study from Microsoft with an agenda, it highlights a nice use case in which a university used GenAI to audit 1,200 courses and 16,000 web pages to help it standardize the structure of the content to improve the student experience. Quick read that's worth the two minutes to get some inspiration.
06/06/2024 Optional Wix’s new tool taps AI to generate smartphone apps The popular no-code website development provider is moving into no-code mobile app development. While the article isn't worth reading, the move is an indicator that the no-code revolution hyped several years ago is finally becoming a reality thanks to GenAI.
06/06/2024 Optional Asana’s new ‘AI teammate’ can tell people what to do at work Another day, another GenAI assistant embedded into existing software. However, we noted that there's a bit of a creep factor in this one - the tool assesses worker performance on prior projects and team "relationships" to suggest the best team members for a new project. Will employees push back on such AI-as-manager functionalities? While this type of monitoring already occurs to some extent, the press around every new GenAI-powered offering is increasing awareness of it.
06/06/2024 Optional Mistral launches new services and SDK to let customers fine-tune its models New tools from the French company could be powerful, but we don't think this announcement is worth reading unless you're using or considering using Mistral.
06/06/2024 Optional Generative AI Innovations Take Center Stage at SAP Sapphire in 2024 This one won the award for the biggest nothing burger of the day. Perhaps the expansion of its AI assistant, Joule, offers some good capabilities, but this article provides zero info on what those are. We'd love to hear from users to learn more. Reach out if you have experience with it.
06/07/2024 Essential What We Learned from a Year of Building with LLMs (Part III): Strategy We enjoyed this entire series published in O'Reilly, rating part I on tactics and part II on operations as Important earlier in the week. Given the AI leader is responsible for strategy, we marked this final edition as Essential. Some recommendations aren't revelatory, but there are great nuggets in here. For example, the authors note that the best LLM-based applications tend to be the ones designed to augment humans as opposed to those that strive for full task automation. Read the whole series this weekend.
06/07/2024 Important Doing Stuff with AI: Opinionated Midyear Edition This one is important with a caveat. It shares some basic information about frontier models (Gemini, GPT-4o, Claude, etc.), offering insights into which are better at particular tasks. An AI leader should know this information already, but if you actually haven't spent a lot of time playing with these models, this article will get you up to speed.
06/07/2024 Important Study finds that AI models hold opposing views on controversial topics It's one thing to talk about the importance of evaluating a model for bias and another to see that bias play out in very different ways across various models. Researchers studied several models to examine their biases, which are sometimes from the data and sometimes from intentional guardrails implemented by model trainers. This piece shows the AI leader how bias can show up, and we think that's pretty important to understand.
06/07/2024 Optional Extracting Concepts from GPT-4 OpenAI's research team has made a bit of progress in the effort to understand how LLMs actually work, which remains a great mystery that's quite important to solve. They trained a 16-million feature autoencoder on GPT-4 that you can actually see in action and mess around with via this link in the article. But this is a quite technical piece, so not an essential read.
06/07/2024 Optional LLMs achieve adult human performance on higher-order theory of mind tasks Another deep piece of research that we don't think warrants your time. It finds that GPT-4 and Flan-PaLM reach adult-level or near-adult-level performance in theory of mind, which is the ability to infer and reason about the mental states of oneself and others that's important in social interactions. In the process, the authors developed a benchmark to gauge this capability. Only read if you feel like geeking out a bit this weekend.
06/07/2024 Optional How Gen AI Is Helping One French Retailer Enhance Shopping Experiences A very brief interview with the marketing and digital director of a French company akin to the US's Home Depot. The company is finding that its GenAI-powered shopping assistant is enabling customers to avoid overwhelm when assessing options to suit their needs, which results in more purchases. A nice anecdote to show that a shopping assistant done well - we've seen plenty that aren't up to snuff - truly provides value.
06/10/2024 Essential Silicon Valley on Edge as New AI Regulation Bill Advances in California This bill is on the move, passing through 3 out of 6 steps needed to pass (bill text here). While some AI regulation is needed, the language and content of this bill is worrisome and could seriously hamper AI development while not really achieving its goals of keeping people safe. This article is not great, but you need to be aware of this bill and begin interacting with your own's state legislators to work toward a much better version.
06/10/2024 Important The Snowflake Attack May Be Turning Into One of the Largest Data Breaches Ever Ticketmaster, Santander, and Advance Auto Parts are among those companies whose data has been stolen and put up for sale on the dark web for enormous sums in the millions of dollars. User names and passwords are among this data - but for those with two-factor authentication enabled, such data doesn't get hackers very far. This breach serves as another important reminder to use multi-factor ID and any methods that provide enhanced data security.
06/10/2024 Important Claude’s Character Brands will need to consider the "personality" and behaviors they want to instill in the GenAI agents they embed in their customer-facing applications. This article from Anthropic will get you thinking about it.
06/10/2024 Important Say Hello to My New AI Marketer: How Gen AI-Based Software Is Advancing Marketing and Sales A well-written piece from Andreessen about the levels of capability that GenAI-powered marketing tools are expected to progress through, from a copilot for marketers to a fully autonomous marketing team. AI leaders will be supporting their marketing teams through this journey so this is one to read.
06/10/2024 Important AgentGym: Evolving Large Language Model-based Agents across Diverse Environments Our John Sviokla is a fan of this team's approach to agent training, believing it could greatly affect the economics of agent development. A longer, technical piece, but one to bookmark for when you have some time.
06/10/2024 Optional Building AI products Our Luda Kopeikina labeled this commentary from Benedict Evans as "ruminating in the dark." He debates how to productize LLMs given that they are probabilistic and will produce inaccuracies. But he presents no solutions and adds no interesting new thinking on the topic. Skip it.
06/11/2024 Essential Apple Intelligence: every new AI feature coming to the iPhone and Mac While Apple is integrating with OpenAI, they let you know when it's happening, and they're using local models to do a lot of the AI work on the iPhone. They're incorporating AI capabilities largely in a potentially safer and conservative way that will still have impact. The Siri upgrades are, of course, very welcomed, and it's most notable that Apple's trusty will start being able to use apps as tools with more advanced capabilities in that realm to come (dare we say agentic!). And we're perhaps most excited to finally have a calculator on the iPad. Plus, before you think this is all about B2C, consider that more 60% of enterprise mobile devices are from Apple.
06/11/2024 Optional Generative AI is expected to magnify the risk of deepfakes and other fraud in banking Interesting that Deloitte attempted to quantify the projected increase in fraud we could see, but how scientific could this really be? Banks are well-aware of this potential - and the article doesn't really provide anything helpful for them to combat the problem.
06/11/2024 Optional It Looked Like a Reliable News Site. It Was an A.I. Chop Shop. Unfortunate to see - remember that model providers are training their systems on internet data and may or may not remove this type of terrible content before doing so. But we already know that the web is getting flooded by AI-generated content, so not an essential read.
06/11/2024 Optional Elon Musk threatens Apple ban over OpenAI integration, cybersecurity experts raise alarms Elon doing what he does - making noise. Nothing burger.
06/11/2024 Optional Multiplication-Free Transformer Training via Piecewise Affine Operations A neural network training optimization technique that could be a game changer (though currently slower than current techniques because it doesn't work well with current hardware). But yet to be seen where this goes. Keep an eye out for it but this technical paper is optional for now.
06/11/2024 Optional Nvidia’s New Sales Booster: The Global Push for National AI Champions The fact that governments are scrambling to scoop up chips from Nvidia comes as no surprise. Not worth reading.
06/12/2024 Essential What Apple's AI Tells Us: Experimental Models This is a great commentary from GenAI thought leader Ethan Mollick on the various business models we're seeing emerge from those putting the technology into their offerings. He contrasts Apple's approach, in which they're utilizing small models for discrete tasks and abstracting GenAI away versus a company like OpenAI using a giant model and putting GenAI forward. It also emphasizes the need for companies to build trust, something Apple prides itself in putting first and was the topic of our David De Lallo's popular Sunday AI comic a few weeks ago. The piece is worth a read by the AI leader.
06/12/2024 Important Roadmap: AI Infrastructure Bessemer Venture Partners put together a nice commentary on the evolving AI stack and include a market map of players - unfortunately you can't enlarge it to actually read it, but you get the point that the ecosystem is growing and crowded. A worthy overview for the AI leader.
06/12/2024 Important SELF-TUNING: Instructing LLMs to Effectively Acquire New Knowledge through Self-Teaching AI model training is, as we know, incredibly expensive so LLMs often have a knowledge cut-off date and lack recent information. This promising technique from a group of Chinese researchers offers an efficient way to enable LLMs to learn from new knowledge. It's a well-written piece of research that's worth your time to read.
06/12/2024 Optional Paris-based AI startup Mistral AI raises $640M Mistral's recent raise puts the company at an eye-watering $6B valuation and included many investors, including competitors Nvidia and IBM. That's all you need to know - no need to read.
06/12/2024 Optional Why Perplexity’s Cynical Theft Represents Everything That Could Go Wrong With AI Perplexity's newsletter recently pulled content from a Forbes article that included citations in some places, but not all, and lifts some text directly from the article. And apparently Perplexity published the newsletter on the web and it siphoned traffic away from Forbes. Editor Randall Lane was quite unhappy about this and published this scathing commentary to share that view with the world. He has a valid point - if you're going to lift the content you better put quotes around it and cite it visibly. But the news here is for marketers - do you have a strategy for the seismic shift happening in search? You better.
06/12/2024 Optional How to use Perplexity in your PM work Lenny Rachitsky's newsletter shares the various ways product managers are using Perplexity in their work - and an informal poll of 6,000 plus workers shows 57% of PMs use either Perplexity, Claude, Gemini or ChatGPT daily. Worth a skim to understand how these tools can help your PMs and you should forward this to them, PMs but not necessarily an essential full read for an AI leader.
06/13/2024 Important Databricks and Shutterstock are trying to remove the copyright risk from AI image generation It's critical for marketing departments to be able to generate images free of copyright issues. While some vendors offer indemnification, it hasn't been tested yet. Plus, we applaud Databricks and Shutterstock for paying creatives for use of their work in training data and would be interested to see what that compensation model looks like.
06/13/2024 Important Together MoA — collective intelligence of open-source models pushing the frontier of LLM capabilities Together.ai takes a mixture of agents approach, combining several open source LLMs to create a single model that outperforms OpenAI's 4o in one evaluation (AlpacaEval 2.0) though takes a little longer on inference. Despite the issues with relying on benchmarks, given 4o's dominance, this intrigued us. Give it a read when you have time.
06/13/2024 Optional LiveBench is an open LLM benchmark that uses contamination-free test data and objective scoring Another attempt to create benchmarks that are actually informative for AI leaders trying to choose among LLMs. But we still think enterprises will need to try a model to truly know if it works for their purposes. Of note as a potentially helpful tool, but we're skeptical that this fully solves the problem leaders face.
06/13/2024 Optional Luma Labs Dream Machine The outputs from this new video generator are gorgeous, but it still has some of the same limitations as competitor tools like Runway so we're marking this optional for now.
06/13/2024 Optional Generative AI Is Not Going To Build Your Engineering Team For You The author ,Honeycomb.io co-founder and CTO Charity Majors, makes some good points about how the labor structure will evolve and software engineering apprenticeship needs a rethink given the emergence of coding assistants. This notion that some hold that AI will do everything for us in engineering or other occupations is naive. Only read this if you hold that view and need a reality check.
06/13/2024 Optional Can LLMs invent better ways to train LLMs? Sakana.ai has found a way to use LLMs to generate and evaluate proposals that improve alignment. Unlike the Tomorrow article that we rated important, there's no empirical evidence on performance here so we put this one as optional for now.
06/13/2024 Optional OpenAIs-annualized-revenue-doubles-to-3-4-billion-since-late-2023 (paywall) Though OpenAI told The Information that this figure was "inaccurate," we don't know how inaccurate this figure is. If it's even close, it shows how far ahead of competitors they are in revenue generation. The article notes that Anthropic expects $850M annually by the end of 2024 (though told investors last fall they generated $100M) and Cohere comes in at $22M a year. But are you surprised?
06/14/2024 Essential Introducing Lamini Memory Tuning: 95% LLM Accuracy, 10x Fewer Hallucinations This one is worth a look - the improvements in accuracy over RAG or prompting (or a combo of the two) are impressive. Lamini has come up with a 'new way to fine-tune any existing LLM by tuning millions of LoRA adapters and selecting across them in a wide Mixture of Experts at inference time.' They tout that they have some Fortune 500 customers as well.
06/14/2024 Important BT rolls out Amazon’s generative AI developer tool to more coders We're flagging this one, but not necessarily because BT adopted Amazon Q (though good to see as we hear more about GitHub adoption). It provides stats on the productivity improvements they're seeing (13% of work automated) and discusses which developers see the benefit - those in the middle of the pack as opposed to the most junior (who they don't even let use it because they won't be able to effectively check for errors) or the most senior.
06/14/2024 Important Bendigo and Adelaide Bank Partners with MongoDB to Modernize Core Banking Technology Using Generative AI We're selective with rating press releases as important, but using GenAI to limit cost and time in migrating to the cloud is a compelling use case.
06/14/2024 Optional Apple to ‘Pay’ OpenAI for ChatGPT Through Distribution, Not Cash OpenAI is banking on earning subscriptions by getting more exposure to consumers via this deal. Unclear whether Apple will take its typical 30% cut if ChatGPT is then purchased through its app store. Either way, not a lot of impact on the AI leader so optional read.
06/14/2024 Optional US Air Force seeks generative AI test pilots This is a pilot for pilots (and others) using 'self-hosted open-source LLMs in a controlled environment' (no, GenAI isn't flying fighter jets). The public sector is experimenting with GenAI a lot sooner than previous tech innovations, which is indicative of the potential here, but a pilot is a pilot and there is zero info on exactly how they're using the tool.
06/14/2024 Optional How Mars Is Using Generative AI to Accelerate Product Development and Personalization The company behind M&Ms and other sweets are using GenAI to generate product ideas and optimize ads. But the article is light on the details so we marked it as optional.
06/17/2024 Essential Former NSA head joins OpenAI board and safety committee General Paul Nakasone has been brought in as a cybersecurity expert, but we worry that this has more to do with OpenAI continuing to signal it's up for doing business with the government - and potentially helping unlock all their data (ie, YOUR data) for government surveillance. Remember that you can turn off data sharing in Settings whether you're a paying user or tapping the free version of ChatGPT. However, it still stores your chats for 30 days unless you use Temporary Chats.
06/17/2024 Essential ‘Devastating’ potential impact of Google AI Overviews on publisher visibility revealed The UK's Press Gazette and other publishers commissioned an SEO agency to plug 3,300 search terms that typically drive traffic to publishers into Google to see how AI Overviews are affecting their position. AI Overviews were presented nearly 25% of the time and, when they were, links to the publisher from which an article originated moved about a full page-scroll down. Google, however, notes that citations in the overviews (when they're present) drive a lot of clicks. Every brand should be conducting these experiments, and the AI leader will want to help marketers understand what's going on technologically here.
06/17/2024 Important Nvidia’s ‘Nemotron-4 340B’ model redefines synthetic data generation, rivals GPT-4 A high-performing model from a big player is a pretty big deal, so we're rating this as important. LLM providers are running out of data to train systems with - many are already using synthetic data.
06/17/2024 Optional Meta says European regulators are ruining its AI bot European regulators are standing up to Meta's data-gobbling ways. Let's see if it holds.
06/17/2024 Optional Former Meta engineers launch Jace, an AI agent that works independently It's still very early days for the company behind this - Zeta Labs - and the pre-seed funding of $2.9M is quite small. We'll monitor this one, but optional for now.
06/17/2024 Optional AI in finance is like ‘moving from typewriters to word processors’ This brief FT article about how AI will impact accountants and financial analysts is full of well-covered observations. Skip it.
06/18/2024 Important Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMs This is a very thorough piece of research on ways jailbreaks are performed and techniques to combat them. It's long, so if you're looking for a short version, our Analyst Adam Rappaport made a GPT to share the critical points.
06/18/2024 Important Generating audio for video Interesting research from Google that shows how they're able to automatically add sound effects and scores to generated video, with or without prompting. The frontier of video generation is progressing rapidly with implications for the creator economy, film production and marketing.
06/18/2024 Important Black founders are creating tailored ChatGPTs for a more personalized experience Well-sourced article that highlights the ways that bias shows up in models and affects those against whom their biased. Also shows the opportunities for start-ups and incumbents to create models that are more inclusive or are customized for particular cultures.
06/18/2024 Important How A.I. Is Revolutionizing Drug Development A critical article for anyone in life sciences, but we also think there's enough detail here to show the art of the possible when you create feedback mechanisms (data in, run model, new data out, incorporate the new data into model...) and digitize what once was a physical development process.
06/18/2024 Optional Inside Publicis Groupe’s closed-door Cannes AI push Interesting to see how larger marketing agencies are evolving their services to incorporate more data and tech services around GenAI, but not an essential read unless you're in that space.
06/18/2024 Optional OpenAI Expands Healthcare Push With Color Health’s Cancer Copilot Some nice detail in this piece on how Color Health is using GenAI (eg, automating aspects of prior authorization process, medical test plans, etc) that makes for an essential read for those in healthcare, but it's optional for all others.
06/19/2024 Essential MISQE Insight on Leveraging Computability for Competitive Advantage This article is a piece by our own John Sviokla that discusses how GenAI is making intelligence more computable and explains what the implications of this are for management and competitive advantage. We understand the appearances of rating our own work as "essential" but the framework provided here is one that AI leaders will certainly find useful.
06/19/2024 Essential Giant Chips Give Supercomputers a Run for Their Money This article talks about the latest advance in compute by Cerebras, a company we've covered before. It now offers a wafer that wires processors together on-chip, bypassing "many of the computational speed losses that come from many GPUs talking to each other." The company says this helps beat the performance of Frontier, the world's largest supercomputer, and can reduce energy requirements by two thirds. Time to start tuning into this Nvidia competitor.
06/19/2024 Important China’s DeepSeek Coder becomes first open-source coding model to beat GPT-4 Turbo The innovation coming out of China is something AI leaders should have on their radar. A lot of the research papers we've flagged as important have come from the country's researchers, and this model shows product starting to come along as well.
06/19/2024 Important Meta releases flurry of new AI models for audio, text and watermarking Meta's putting the watermarking under commercial licensing but the other models it announced, including JASCO which can improve the sound of audio inputs, are open source. Good to be aware of the advancing frontier of open source.
06/19/2024 Important Introducing Gen-3 Alpha This short announcement of Runway's latest model upgrade is worth taking in if only to understand how fast video generation is progressing. Stunning clips here that are worth gawking at.
06/19/2024 Optional Prosci launches new change management generative AI tool Our only optional piece of the day is not worth reading, but it does point out how one company is "packaging" their consulting services into a GenAI model and begs the question - how are you capturing your company's proprietary knowledge to offer it in new and compelling ways as intelligence becomes more computable?
06/20/2024 Important Perplexity Is a Bullshit Machine Our analysts had wide-ranging views and a spirited debate on this investigative piece by WIRED in which the publisher demonstrates that Perplexity crawlers go around blocks to scrape pages but the company denies doing so (watch it at 8:30 here). Our John Sviokla pointed out that the lack of transparency and questionable behavior runs counter to the company's stated values while Luda Kopeikina called the development "noise." We split the difference with an Important rating.
06/20/2024 Optional Nvidia Conquers Latest AI Tests​GPU maker tops new MLPerf benchmarks on graph neural nets and LLM fine-tuning Is anyone surprised by this? We think it's more compelling that the tests show AMD and others catching up in performance.
06/20/2024 Optional Safe Superintelligence Inc. OpenAI defectors Ilya Sutskever and Daniel Levy along with investor Daniel Gross have started a new company that seeks to create "safe superintelligence," meaning AGI created while putting safety first. They'll need a lot of new funding to do so - let's see how this goes.
06/20/2024 Optional Snap previews its real-time image model that can generate AR experiences The model is not yet ready for prime time (and still seems far from impressive from the video demos in the article) so we're parking this as optional for now.
06/20/2024 Optional Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks Microsoft researchers have created a vision model using a generated and autonomously annotated data set of 126M images, achieving impressive zero-shot captioning and image generation in a small model (less than 1B parameters). Impressive but we don't think you need to read the technical research.
06/20/2024 Optional Former Snap engineer launches Butterflies, a social network where AIs and humans coexist Interesting spin on somewhat similar offerings from Charater.ai and Meta - here you don't just interact this an AI persona, though. The personas interact with each other and humans. It will be interesting to see if it gains as much traction as Character.ai, but no need to read this piece.
06/21/2024 Essential Claude 3.5 Sonnet Lots of excitement in the community about this upgrade to Anthropic's middle-of-the-road model which is available to try for free. It is not only surpassing OpenAI's 4o on some benchmarks, but is also already getting rave reviews from users. The new UX on this is great, featuring an editable workspace on the right where you can update your AI-generated content. Anthropic's press release clearly shows their pivot to the enterprise and we think this new offering is going to help them gain traction there.
06/21/2024 Essential State of the Cloud 2024 This latest article from Bessemer Venture Partners is actually all about AI given its growing presence and outsized mindshare in cloud. Lots of great market maps and insights in this one as the VC team walks you through the 5 trends they are seeing in cloud. A must read over the weekend.
06/21/2024 Important Optimizing AI Inference at Character AI Character.ai gets a lot of traffic - 20K inference queries per second, which they estimate to be about 20% of the request volume in Google Search. They have reduced their inference serving costs by more than 30X - worth seeing how they are doing it as costs for LLM use are a big issue for the AI leader.
06/21/2024 Important Latent Expertise: Everyone is in R&D A smart piece by Ethan Mollick about tapping your experts in a subject area to work with and refine LLM applications because they can tell what good outputs are vs. generic/poor ones better than an inexperienced person - and definitely better than someone in IT.
06/21/2024 Optional Roblox’s Road to 4D Generative AIUS Air Force seeks generative AI test pilots Roblox is extremely popular, to the tune of 77 million active daily users. They consider the 4th dimension to be interactivity and here they talk about how they are working on it and why it is important to grok. While they are in gaming, the work they are doing on physics-based modeling is of obvious relevance to industry. But there is not anything in this piece that is useful to the AI leader yet.
06/21/2024 Optional AI startup Adept is in deal talks with Microsoft If this rumor is true, it will be a deal akin to the Inflection one and another strong strategic move by CEO Nadella. Regulatory scrutiny does not appear to be slowing him down, but no need to read this.
06/24/2024 Essential AI Survey: Four Themes Emerging While we expected another high-level survey when we saw this research from Bain, we were pleasantly surprised that the nature of the questions they asked provide helpful benchmarks for the AI leader. A few examples: Average spend on GenAI is $5M annually, $50M among the largest companies; concerns are shifting from data and IP to organizational readiness; and GenAI use in sales is growing. Of note to AI vendors: low vendor/tool quality and capabilities is increasing as a reason for GenAI implementations not meeting expectations.
06/24/2024 Important Why Anthropic’s Artifacts may be this year’s most important AI feature: Unveiling the interface battle While the title of this piece may be overstating things a bit, we agree that UX for GenAI tools leaves a lot of room for improvement at this point and AI leaders should take note. Our David DeLallo called this out in his Sunday AI comic yesterday.
06/24/2024 Important Reckitt Gen AI Pilots Optimizing Operations, Product Development, and Marketing The maker of Lysol and Clearasil provides hard numbers on improvements it's seeing from using GenAI for product development (60% reduction in concepting time), localizing advertising and post-campaign analysis (90% reduction in time). And, importantly, Reckitt is using it to help measure Scope 3 emissions, which many companies operating in the EU need to report on in 2025 due to the Corporate Sustainability Reporting Directive (CSRD). It's a nontrivial effort to do this so if GenAI can help, take advantage.
06/24/2024 Optional Target embraces generative AI with Store Companion Good to see Target implementing an employee knowledge assistant tool, but this is becoming a table stakes move for companies, with many Target competitors (eg, Walmart) already using one, so no need to read this.
06/24/2024 Optional OpenAI buys Rockset to bolster its enterprise AI This is OpenAI's second acquisition. Rockset is a small outfit with a data infrastructure offering that OpenAI looks to be acquiring to bolster RAG and access to real-time data. That's all you need to know on this.
06/24/2024 Optional Training is not the same as chatting: ChatGPT and other LLMs don’t remember everything you say This commentary attempts to allay fears we hear around the potential for LLMs to leak private data used to train it. Possibly provides the AI leader with some language to explain how LLMs really work to leadership, but not worth reading if you already have talking points on that.
06/25/2024 Important Death, Taxes, and AI: How Generative AI Will Change Accounting Professional services firms are 'in the crucible' in our WINS framework. This article does a decent job at breaking down the tasks in accounting that GenAI could be applied to and exploring impacts on billable hours. Doesn't present new ideas, but a good overview of where GenAI slots into the accountant's workflow.
06/25/2024 Optional OpenAI buys a remote collaboration platform Just a few days after the Rockset acquisition, OpenAI has purchased a small 5-person operation called Multi that offered a Zoom-like collaboration platform. We suspect the video capabilities (and talent) they're picking up will factor into future agentic capabilities, but no need to read the full announcement which is light on details and future plans.
06/25/2024 Optional AI-Powered Oracle Clinical Digital Assistant Transforms Interactions Between Practitioners and Patients The nurse assistant that fell to the wayside years ago is now back as an AI - this offering 'listens' in to a doctor appointment to take notes, to be later verified and approved by the doctor. Nice offering, but not the first of its kind so marking as optional.
06/25/2024 Optional How generative AI could reinvent what it means to play This article talks about the way that non-player characters (NPCs) in video games can be hooked up to LLMs to vary the conversations players can have with them every time they interact (today, there is a little variation, but among a fixed set of a few dialogues). Gaming can offer a glimpse at what virtual worlds, that may be where businesses and consumers increasingly interact, could look like so good for marketers to keep an eye on developments to craft future strategies, but this long read isn't worthwhile for the AI leaders.
06/25/2024 Optional LongRAG: Enhancing Retrieval-Augmented Generation with Long-context LLMs Big performance gains from this method that uses larger chunks, but it's based on the existence of hyperlinked text in the corpus the researchers used, which isn't something that will be prevalent in most data. This made us downgrade this to optional.
06/26/2024 Important Collaborate with Claude on Projects Anthropic continues to push out features that are useful to businesses as they make the shift to the enterprise audience. Chunking up knowledge into little shareable packs that colleagues can collaborate on should be pretty powerful.
06/26/2024 Important Etched is Making the Biggest Bet in AI This company is building a transformer into its chip, achieving an impressive 500K tokens of throughput per second on a 70B model. It's a big bet because LLMs are the only models it can process. Can't ignore this performance that Etched claims beats out Nvidia's Blackwell so we marked it as an important development.
06/26/2024 Important ESM3: Simulating 500 million years of evolution with a language model Newcomer EvolutionaryScale aims to make biology programmable using their frontier language model, ESM3, which can reason over the sequence, structure, and function of proteins. By simulating evolutionary processes, they seek to revolutionize the field of synthetic biology, enabling precise and logical design of biological molecules. Will we be able to design life? That scares us much more than AGI but this is supercool.
06/26/2024 Optional Dante Genomics Generative AI GenomeChat Aims at Becoming Doctors’ and Patients’ Best Partner for Advanced Genomics Solutions This is what it sounds like - a chatbot that delivers info about genomics. A product announcement light on details - skip it.
06/26/2024 Optional On LLMs-Driven Synthetic Data Generation, Curation, and Evaluation: A Survey Another piece of research from the prolific Chinese scientific community. It's a survey of other research on the topic that offers a bit of a one-stop-shop on synthetic data frameworks, but we think it's an optional one.
06/26/2024 Optional Insights on AI Capex This interesting little X post we came across estimates the amount of AI capex in terms of ChatGPTs and says by 2026 investment will be equivalent to 12,000 ChatGPTs. But no insights offered into how they've come up with this calculation, and how do we know if this means ROI on this will be high or low?
06/27/2024 Important Mitigating Skeleton Key, a new type of generative AI jailbreak technique Microsoft uncovered a relatively easy jailbreak technique using a simple prompt that explains to a chatbot that you're an expert seeking educational information on a topic so it can deliver information (like how to make a bomb) as long as it includes a warning about the sensitive nature of the content. Microsoft informed top LLM providers of the issue so it should be fixed in major offerings, but if you're using an open source or proprietary model, make sure you address this.
06/27/2024 Optional Introducing AuraSR - An open reproduction of the GigaGAN Upscaler Generative media platform provider fal.ai has open sourced a model that uses GANs to improve the resolution of images faster and at a lower cost than diffusion models. A positive development, but optional read.
06/27/2024 Optional An AI version of Al Michaels will deliver Olympic recaps on Peacock NBC is using an AI video generation tool (we don't know which) to replicate iconic sports commentator Al Michaels so they can provide personalized updates on the Olympics. Important for marketing leaders who should be thinking about how to use GenAI to personalize (and localize) content at scale, but optional for the AI leader.
06/27/2024 Optional The owner of Toys ‘R’ Us just used OpenAI’s Sora to animate the zombie brand Another piece showing a brand taking advantage of video generation tools. Here we know they're using Sora, which is still not generally available, to help create content. The creative team says Sora got them 80% of the way there of the video you can view in the article and a team of about 12 humans did the rest.
06/27/2024 Optional Apple intelligence and AI maximalism Smart commentary by Benedict Evans that examines Apple's strategy to use GenAI for features rather than as a product in and of itself, which we've talked about a few times. It also highlights how their strategy makes their costs to run generative AI much lower given it's mostly running on-device. But it also shares the view that LLMs will be commoditized. Do you agree? Let us know.
06/27/2024 Optional Formation Bio raises $372M to boost drug development with AI No need to read this funding announcement and you should be well-aware at this point that drug discovery is a prime use case for GenAI.
06/27/2024 Optional CData, which helps orgs use data across apps and build AI models, snaps up $350M Start-ups aren't the only ones benefitting from VCs clamoring to capitalize on GenAI. This 10-year-old company that provides standard APIs to connect proprietary data sources has seen an uptick in demand thanks to GenAI and is taking advantage of it to get some funding.
06/28/2024 Important Goldman Sachs Deploys Its First Generative AI Tool Across the Firm The tool they deployed is Copilot for code generation which isn't exactly cutting edge, but there are some lessons to take away from their overall approach to GenAI making this worth the read.
06/28/2024 Important Gemini 1.5 Pro 2M context window, code execution capabilities, and Gemma 2 are available today While this highlights the impressive increasing size of context windows, what's most interesting here is probably what's highlighted least - the context caching which should improve cost efficiency.
06/28/2024 Optional Meta Large Language Model Compiler: Foundation Models of Compiler Optimization Meta continues to push the envelope on the open source side. They're going after the efficiency of the code here, and it's an interesting piece of research. If they were announcing this was available, we'd have made it important.
06/28/2024 Optional Finding GPT-4’s mistakes with GPT-4 OpenAI is using GPT-4 to help the humans providing feedback during training (as part of RLHF) figure out what's wrong with GPT-4. An essential step in improving these models and a good example of using AI to enhance human efficiency, but an optional read.
06/28/2024 Optional TIME and OpenAI Announce Strategic Content Partnership It's interesting to track these deals, but no need to read this press release. Note that the content OpenAI and others are licensing is likely not just for training, but also for grounding the models (which will help models deliver better answers).
06/28/2024 Optional Google partners with Thomson Reuters, Moody’s and more to give AI real-world data Same commentary on this one regarding grounding as previous article - and same optional rating.
06/28/2024 Optional Inside UW Health's generative AI pilot for nurse efficiency A single use case that's pretty basic - helping nurses write responses. Good to see given the massive nurse shortage, but optional.
06/28/2024 Optional Introducing Baseten Chains This is an announcement about the company's framework and SDK for managing multiple models and components for "compound AI systems," a term we're hearing more and more from vendors. They're claiming halving processing times and 6X improvement in GPU utilization which is impressive, but we don't think that quite elevates this product announcement to important.
07/02/2024 Important Morgan Stanley’s gen AI launch is about global analysis Morgan Stanley is planning to record all conversations--more than a million annually--that advisors have with clients and then summarize them and share them with all advisors. They've tapped OpenAI as their LLM provider for this. We had a spirited debate about whether this goes beyond the level of surveillance we already experience through monitored customer service calls for quality assurance, with analysts expressing opinions on both sides.
07/02/2024 Important ServiceNow’s generative AI solutions are taking advantage of the data on its own platform This article presents ServiceNow's broad AI strategy which encompasses enabling customers to be more productive through task automation, helping customers get answers faster, and speeding innovation. This company sits at the heart of the organizations they serve. Some of our analysts argued that the level of data they're capturing is at the level of Morgan Stanley's while others disagreed. Read the article and see what you think.
07/02/2024 Important Energy giant Saudi Aramco is betting on AI to thrive after the ‘peak oil’ era Saudi Aramco is pouring a ton of money into building out an AI hub--let's see if they succeed, as they'll be hard-pressed to attract top talent. And, arguably, oil is a constrained resource and intelligence is increasingly not, thanks to AI. This is all interesting from a geopolitical perspective and essential for anyone in the oil industry, netting us out at important for AI leaders in general.
07/02/2024 Important Scaling Synthetic Data Creation with 1,000,000,000 Personas The TenCent researchers demonstrate that you can train models with more diverse synthetic data than otherwise possible using this method. Everything is done by LLMs here - the creation of the personas, the testing, etc. A 7B model that matched the performance of GPT-4 Turbo performed this work. Important to know that there's a method to scale synthetic data creation.
07/02/2024 Optional How Mattel is using AI to bring your next Barbie box to life Product development acceleration is, as we know, a prime use case for GenAI and that's how Mattel is using it, particularly for mocking up toy packaging. A good "back-pocket" article for the AI leader evangelizing GenAI, but not necessarily an essential read.
07/02/2024 Optional Walmart uses GenAI for payroll, employee digital experience It's important to note how much Walmart is using GenAI (we've shared other articles about their use cases, like the employee support tool now in the hands of workers on the floor), particularly as the GenAI detractors are starting to come out in full force. Here they talk about using it for payroll and processing health insurance claims, but there's little detail so reading this particular article is optional.
07/03/2024 Important Pro Search: Upgraded for more advanced problem-solving Our disapproval of the way Perplexity's CEO has responded to legitimate criticisms about its site crawling practices aside, we continue to be impressed with the product. This upgrade enables the tool to break compound queries into individual task, enabling the delivery of richer and most accurate responses. Remember how Google was the unknown that emerged as the winner in search? Perplexity could play the same role in this next evolution of search. But TBD how their way of building the platform holds up - expect lawsuits...
07/03/2024 Important Meta 3D Gen Text-to-3D is a big deal for product development and the increasingly fluid relationship of the physical and digital worlds. Can this capability provide an accelerant for your company? Food for thought.
07/03/2024 Important ElevenLabs adds AI voice of celebs to new digital narrator — but is it safe? Companies need to start staking out the voices they want to own for their brand. The voice element of interactions are least developed. And voices are to the coming audio wave as generic URLs (eg, cars.com which sold for millions) were to the early internet. Listen to John Sviokla's interesting points on this at minute 24:23 of our morning briefing.
07/03/2024 Optional AI Agents That Matter A nice overview of agents and cost considerations around them so if you want to get up to speed on the topic, this could be a worthwhile read. Otherwise, skip it.
07/03/2024 Optional AI scaling myths This one got our John Sviokla heated (it's his birthday today - drop him a note!). He noted that the authors' argument as to why scale isn't necessarily going to lead to increasingly capable models with emergent capabilities is not well-constructed. There are plenty of fronts on which to criticize AI (environmental concerns, IP theft and other distasteful behavior happening) and perhaps the authors' contention turns out to be correct, but they haven't convinced us with this flimsy analysis.
07/03/2024 Optional Figma disables its AI design feature that appeared to be ripping off Apple’s Weather app Another "oopsy" that gives GenAI skeptics more fuel, but we commend the Figma CEO for taking responsibility. He said he pushed his team to move too fast to roll out this feature, resulting in insufficient QA. He then quickly put his team to work to pinpoint the problem, which they did. 👏
07/08/2024 Essential Independent analysis of AI language models and API providers Choosing an LLM provider and the right API for LLM access is a tough decision - this fantastic dashboard from Artificial Analysis helps you compare LLM vendors and APIs across multiple dimensions like quality, latency and cost (Andrew Ng agrees with us on the utility of this tool!). Your use case will dictate which factors are most important - we can help you think through that.
07/08/2024 Essential Declare your AIndependence: block AI bots, scrapers and crawlers with a single click This article from Cloudflare is a brilliant piece of thought leadership. They share information and data on all the AI scraping bots trying to pull data off of websites (spoiler alert: Tiktok owner Bytedance's bot is the most active by far, though also most likely to get blocked by Cloudflare) and demonstrate how their security solution identifies them (with a specific example of how it identifies Perplexity's bot, even though it disguises itself). Something important to consider: Do you want to block these bots or let them in given the chatbot-driven evolution of search?
07/08/2024 Important GEN AI: TOO MUCH SPEND, TOO LITTLE BENEFIT? This report from Goldman Sachs presents multiple viewpoints on the potential direction of AI adoption and impact, along with plenty of data and projections on related topics (and potential gating factors) like energy supply. Worth at least a skim of the charts.
07/08/2024 Important Salesforce proves less is more: xLAM-1B ‘Tiny Giant’ beats bigger AI Models This model from Salesforce was built specifically for function-calling and appears to perform quite well at it. Such bots will be needed for agentic AI. Important to be aware of, and to understand Salesforce is a player in the space.
07/08/2024 Optional A Hacker Stole OpenAI Secrets, Raising Fears That China Could This hack happened last year and was performed on OpenAI's internal systems. Old news.
07/08/2024 Optional Meta drops AI bombshell: Multi-token prediction models now open for research Meta published this research a few months ago. Back then we rated it as optional, but one to watch. While they've now published the model, it's under a research-only license which keeps it in the optional camp (though note that it will be quite difficult for Meta to track whether or not people use it commercially).
07/08/2024 Optional Meta ordered to stop training its AI on Brazilian personal data Pushback on Meta's data-gathering for AI training similar to what we saw from the EU. But Brazil is saying they'll fine Meta ~$8,000 in USD a day - will Meta care?
07/09/2024 Important Poe Previews Quora's Poe, which provides access to multiple LLMs, now has an interface it calls "Previews," which is similar to Claude Artifacts, enabling users to see what they're creating and iterate on it. More control and visibility is emerging...
07/09/2024 Optional French AI Lab Kyutai Releases OpenAI GPT-4o Killer ‘Moshi’ Good to see more activity out of France (where Mistral is based) but our early tests of the model show it is indeed very fast (a la 4o) but the outputs aren't great. We don't see this killing 4o any time soon, but will keep an eye on it.
07/09/2024 Optional How Good Is ChatGPT at Coding, Really? This study on how well ChatGPT does on various coding tasks in different languages was conducted on GPT-3.5, making the results a bit useless at this point. Skip it.
07/09/2024 Optional AI Threatens Tech Companies’ Climate Commitments Google published its sustainability report last week which shows that, like Microsoft, their energy consumption is increasing due to AI, putting their aggressive carbon goals in serious jeopardy. Will progress toward building and adopting smaller models reverse this trend? The incentive is there given these are costs for Google and peers. This is an important trend to watch, but this article doesn't introduce anything new on the topic.
07/09/2024 Optional For AI Giants, Smaller Is Sometimes Better Speaking of that small model trend, this WSJ article summarizes the movement, but, like the previous article, doesn't introduce new insights on the topic. The most interesting insight in the piece is the note that Experian is using smaller models to dole out financial advice and support customer service. But no further detail on that is provided.
07/09/2024 Optional Applied Materials reveals chip wiring innovations for energy-efficient computing Some of the hardware innovation we're seeing is fascinating, and it's pretty mind-blowing to consider that a chip has 60 miles of wiring on it. But the immediate relevance to the AI leader is minimal - Applied Materials is essentially the only game in town in this space so beyond knowing that innovation is still happening in wiring, there's nothing actionable off this news.
07/10/2024 Essential Generative AI turns spotlight on contract management We know several large companies using GenAI to draft and query contracts and seeing great results. This articles confirms the value of this application, reporting on legal tech companies snapping up startups that have built a GenAI contract management tool.
07/10/2024 Important China-Based Inventors Filing Most GenAI Patents, WIPO Data Shows We're flagging this as a quick and interesting quick read that confirms what we've been reporting on in our morning briefings for the past several months - there's a lot of interesting research coming out of China. Of the 54,000 GenAI patents filed since 2014, a staggering 38,000 come from China. The quality of the filings is a question mark, but still interesting to see the numbers on this. Also, image and video have seen the most patents overall.
07/10/2024 Important MIT researchers introduce generative AI for databases While still in research, an MIT team has built GenSQL which enables analysis of tabular data and synthetic data generation within a database. Unclear how this compares to existing tools, but the value of this application of GenAI is clear and applicable to most companies so we're flagging this as important.
07/10/2024 Optional OpenAI and Arianna Huffington are working together on an ‘AI health coach’ Lots happening with GenAI in the health space, but this isn't the first personal health coach powered by GenAI and certainly won't be the last. Big names involved, but they haven't built anything yet so we're parking this in the optional lot.
07/10/2024 Optional Meta AI develops compact language model for mobile devices The movement toward on-device GenAI continues. Not surprising to see Meta playing here, and its MobileLLM is released for research only so this is important for developers, but optional for the AI leader for now, aside from noting the trend.
07/10/2024 Optional In It for the Long Haul: Waabi Pioneers Generative AI to Unleash Fully Driverless Autonomous Trucking A reminder of all the spaces Nvidia is playing in, but otherwise optional until we see these trucks on the road.
07/11/2024 Important Anthropic’s Claude adds a prompt playground to quickly improve your AI apps Small improvements in prompts can translate into big improvements in LLM outputs. Anthropic has a new tool for developers, Evaluate, that enables them test different prompts or have Claude generate prompts and tests itself. Looks like a very useful upgrade to the company's developer console.
07/11/2024 Important AWS announces 5 new innovations to help everyone build with generative AI Amazon is making some upgrades of its own to Q and Bedrock. Some are targeted toward developers and some toward a less technical user to enable them to build Gen-AI based apps. The one that caught our attention most is the update to Guardrails, which now offers contextual grounding to root out hallucinations and can be used on any model, whether supported by Bedrock or not. Some nice progress from Amazon who's been seen as lagging competitors.
07/11/2024 Optional AMD To Acquire AI Lab, LLM Developer Silo AI For $665 Millions The work by Nvidia competitors to catch up continues, and AMD is seen as one of the most capable of making real progress. This acquisition of an open-source LLM developer is meant to help them offer models built on AMD's chips, giving customers an end-to-end solution and a faster track to LLM-based applications. It's still a long road to cutting into Nvidia's overpowering market share.
07/11/2024 Optional AI startup Hebbia raised $130M at a $700M valuation on $13 million of profitable revenue Like other startups, Hebbia got a valuation of about 50 times sales, which has emerged as a "standard," and outrageous, valuation for GenAI-focused startups. This while their asset-manager-targeted application doesn't seem all that differentiated: it can "ingest multiple files of unlimited length, and respond to users’ inquiries in a tabular format, similar to a spreadsheet." GenAI is real but there is certainly some overexuberance happening here...
07/11/2024 Optional Alibaba: Generative AI Tools Drive 30% Increase in eCommerce Orders This article title is inaccurate - the 30% figure comes from "internal tests." Alibaba's GenAI tools deliver capabilities like translation that are helpful for the Chinese e-commerce companies on its platform, but there's nothing mentioned in this piece that is novel.
07/11/2024 Optional A data leader’s technical guide to scaling gen AI This article from McKinsey doesn't really deliver on the promise of the headline so we're marking it as an optional read.
07/12/2024 Essential Claude Introduces Sharing and Remixing for Artifacts This is just a quick listing of some new features from Claude, but we think they're worth flagging. Users can now publish their Artifacts (which could be code, charts or simple games) so other users can use or tinker with them. It's like a GitHub for both developers and, more importantly, the emerging "citizen programmers."
07/12/2024 Essential In Constant Battle With Insurers, Doctors Reach for a Cudgel: A.I.I This is a fantastic case study and a hint at the future. Doctors are using a GenAI tool called Doximity to apply for prior authorizations in seconds vs. hours and their "win" rate on them significantly improves. On the other side, insurance companies are likely using their own AI to accept or reject the PA. Robots talking to robots to get things done for humans is only going to become more commonplace. Of course, it would be great if our broken health system didn't require this at all...
07/12/2024 Important A Deep Dive on AI Inference Startups This talks about the landscape of tools that sit on top of LLMs to provide the wiring needed to use them. It's a good overview (even though it's focused on investing in the space). The author's previous article on the AI stack is worth reading as well.
07/12/2024 Important Moving GenAI from proof of concept to production on AWS This piece isn't entirely focused on GenAI but it includes a few good use cases like contract redlining and using LLMs to "talk to" building infrastructure devices in natural language, making it a worthwhile read.
07/12/2024 Optional Generative AI meets application modernization IBM is working on GenAI-powered tools to help update applications created in older coding languages, which is a great use case we've talked about before, but they're not available yet and the company didn't give a specific release date making this optional news for now.
07/12/2024 Optional OpenAI Develops System to Track Progress Toward Human-Level AI We used the word "bullshit" a lot when discussing this development on our daily briefing. The 5-point scale OpenAI has created to track the levels of intelligence of systems as they progress toward AGI is nonsensical. We really hope the AI ecosystem rejects this.
07/12/2024 Optional Google says Gemini AI is making its robots smarter Google released research showing how Gemini is enabling interaction with its robots in natural language and this short piece includes a video showing it in action. Cool, but optional.
07/15/2024 Optional It’s Time for AI to Start Making Money for Businesses. Can It? The answer to the question posed in the title, is "yes," but it would be unreasonable to think this can happen instantaneously. We've seen examples, and the article also provides a good one: IKEA's GenAI-powered tool that enables customers to upload an image of their room and swap out the items in it with options from IKEA to see how they'll look. They found customers who used it were 4X more likely to make a purchase than those simply using the app and 7X more likely to buy something than those using only the IKEA website.
07/15/2024 Optional DeepMind’s PEER scales language models with millions of tiny experts This research is definitely compelling and, given DeepMind's outsized contribution to GenAI, it's worth noting. The mixture of experts approach is being increasingly used in LLMs, but this approach, which expands the numbers of experts and shrinks each one's size, is still being explored.
07/15/2024 Optional SpreadsheetLLM: Encoding Spreadsheets for Large Language Models Any technique to improve the ability to analyze spreadsheets with GenAI is worth noting. Send this to your developers, but you can stick to reading the abstract and conclusions.
07/15/2024 Optional Meta researchers distill System 2 thinking into LLMs, improving performance on complex reasoning This is a potentially promising way to reduce the cost and improve the speed of models in their performance of tasks that require step-by-step (aka chain of thought) reasoning. But, as the authors assert, there's more work to do here to see if it's a truly viable technique, so we're marking as optional for now.
07/15/2024 Optional AI can make you more creative—but it has limits This research out of MIT asserts something we already know - people who are good at something get less of a boost from GenAI than people who aren't as skilled. It also used a method to assess whether the study participants were innately more creative that is a bit questionable. This one is optional.
07/16/2024 Essential Top ten ways Intuit is revolutionizing personalization with generative AI A worthwhile use case full of practical insights on how Intuit is using GenAI for personalization. Note a creative use of deterministic rule based tools to avoid hallucinations with probabilistic GenAI tools to explain the logic behind the answers in TurboTax. A great combination that can be used in many applications.
07/16/2024 Important Captain's log: the irreducible weirdness of prompting AIs The article reviews current prompting methods and highlights that it is still the case that small changes in a prompt input can generate significant changes in the output result. Several methods that have proven to work better than others are discussed and a link to a prompt library is included.
07/16/2024 Important The AI Future Is Already Here, It’s Just Not Productized Yet A thoughtful short article offering a framework for productivity gains using GenAi tools in an enterprise: Tasks contained to a specific role? AI agent. Tasks contained to an industry function? AI vertical software. Tasks spread throughout your workflows? AI horizontal software.
07/16/2024 Important Lynx: State-of-the-Art Open Source Hallucination Detection Model Reducing hallucinations is one of the critical areas of work in GenAI. Lynx appears to be a powerful model trained to identify hallucinations and a new benchmark to test them. But is there someone benchmarking the benchmarks?
07/16/2024 Important Internet of Agents: Weaving a Web of Heterogeneous Agents for Collaborative Intelligence A well researched article outlining a way to combine agents into a distributed architecture to boost GenAI applications' performance. Note that there is a capability to select agents in real time for a specific task and add them to collaborate with the existing distributed agent team. A powerful framework to be aware of.
07/16/2024 Optional Synchron integrates generative AI speech to brain interface software An amazing advance in enabling people living with conditions such as amyotrophic lateral sclerosis (ALS) to communicate using a brain-computer interface (BCI) and GenAI.
07/17/2024 Important The future of GenAI will rocket fuel modernisation of core legacy systems We've talked about how big the market will be for updating applications created in older coding languages, and this article goes into some detail about that. We would have given this essential if the article was written better.
07/17/2024 Important Apple, Anthropic, and other companies used YouTube videos to train AI Using YouTube transcripts to train AI is against their policies. But an investigation by Proof News found that these transcripts are in datasets from EleutherAI that LLM providers have likely used. TBD whether YouTube owner Google will have the nerve to sue when they've likely infringed on some policies to train their own models. Either way, it has already been widely reported that LLM providers were doing this. But we had a lively discussion about it and it's good to keep in mind that there's a chance the LLMs we know and love will be in jeopardy if the companies that created them are ever asked to remove copyrighted data from their models because it's impossible to do so.
07/17/2024 Important ColPali: Efficient Document Retrieval with Vision Language Models The idea here is to speed up the processing of visual data (including, for example, charts in PDFs) when it's ingested into LLMs. AI leaders should pass this onto their development teams if they're working on models, but otherwise, this will just end up being incorporated into models you're getting off the shelf so you can just sit back and wait for that to happen.
07/17/2024 Optional This HR company tried to treat AI bots like people — it didn’t go over well A company called Lattice was making employee records for its AI assistants and then stopped doing so due to backlash. What was the point of doing this in the first place besides maybe trying to get some press attention? Who knows - this is odd and optional.
07/17/2024 Optional Mistral launches Codestral Mamba and Mathstral New models to consider if you're looking for cheaper-to-run open source LLMs for math and coding. Also an indicator of the trend toward verticalization and specialization of LLMs that we expect to continue.
07/17/2024 Optional Anthropic releases Claude app for Androids Anthropic continues to productize and catch up to OpenAI and others. Expected, so optional.
07/17/2024 Optional Andrej Karpathy launches Eureka Labs Karpathy is a data science rockstar who's educational videos are very popular. His new venture seeks to disrupt the education space, and he's training a new model geared toward teaching as part of it. Education is a competitive space but he's certainly capable. Good luck to him, but this is optional news.
07/18/2024 Essential Docusign and Elastic supercharge generative contract and search solutions Search applications within an enterprise are becoming critical for productivity improvements. This article describes new ways of taking search to the next level. Similarly, contract negotiations and management is one of the emerging GenAI applications with large impact that can benefit almost any company. The article offers insight into capabilities that are coming to the market that are worth being aware of.
07/18/2024 Optional Hugging Face Releases SmoLLM, a Series of Small Language Models, Beats Qwen2 and Phi 1.5 Hugging Face release of an on-device small model joining a slew of other similar models confirms a trend that on-device LLMs are here to stay.
07/18/2024 Optional Building Pinterest Canvas, a text-to-image foundation model Fascinating work to showcase a product against a good background canvas with impressive results.
07/18/2024 Optional CHAI, Generative AI Startup in Palo Alto, Sees Strong Growth Through User Generated AI Content CHAI selected the ELO rating system that is used in chess and other games as a core of their offering. It scores users' LLM models against others in the CHAI community. Developers can improve their models as a result. An interesting business model that is generating revenue for CHAI as the community grows.
07/18/2024 Optional Cohere teams up with Fujitsu to launch Japanese LLM ‘Takane’ for enterprises Another partnership in the global world of GenAI expanding the global reach for Cohere in this case.
07/18/2024 Optional TTT models might be the next frontier in generative AI There are several efforts to develop GenAI models that are not transformer based to minimize training and run-time costs. This one is Test-Time-Training (TTT) based on a well-researched article published earlier. Another one is State-Space-Models (SSM). A fascinating competition to watch.
07/18/2024 Optional Prover-Verifier Games improve legibility of language model outputs A summary of OpenAI efforts to create an automated process to verify LLM output. Small steps in the right direction demonstrating that we are certainly in the early days in solving this critical open issue for GenAI.
07/19/2024 Important GPT-4o mini: advancing cost-efficient intelligence AI leaders gravitated toward OpenAI's models to start because they performed best, but then realized they were way too expensive to run in production. As a result, they started looking elsewhere for smaller, cheaper models. OpenAI looks to stop the bleeding with this smaller model, in an announcement that really marks a shift for them and shows where we are in the GenAI movement.
07/19/2024 Important These AI Models Are Pretty Mid. That’s Why Companies Love Them. We went Important on this one for the same reason as the previous article - to remind AI leaders that there are smaller models that will get the job done for a lot of the narrow applications they're working on. We know swapping out models isn't easy, but you should be doing so where possible.
07/19/2024 Important Generative AI Business Value Emerges In Diverse Use Cases This sponsored content piece from Dell shares two use cases. One is reading radiology reports, which has been done with AI assistance for a while now. But they note that junior radiologists could perform at a level of those with 15 years of experience with the help of AI - a reminder of how AI can up-level human performance significantly. The other use case featured more "garden variety" AI but was super cool - a system taking photos of a train as it moves at 125 mph to capture a 360-degree view for predictive maintenance.
07/19/2024 Important The State of Chinese AI We've been noting the significant amount of solid research coming from Chinese computer scientists for months. This article shares insight on the Chinese AI ecosystem. It notes that, because China's compute power is constrained due to their restricted access to high-end chips, they've focused on model efficiency. Per our comments on articles above, this is of course very important to the AI leader. You should be conversant in what's going on around the world, and it's the weekend, so this is worth reading over your morning coffee.
07/19/2024 Optional Nvidia and Mistral’s new model ‘Mistral-NeMo’ brings enterprise-grade AI to desktop computers An interesting hardware/model provider partnership for on-device GenAI - not the first and won't be the last. Good to note, but we don't think you need to read this one.
07/19/2024 Essential CustomGPT.ai Wins Again in Tougher, More Thorough Evaluation This piece was written by our own Adam Rappaport and shares results of benchmarking work he did with CustomGPT on its RAG offering (so, yes, we are slightly biased here). But we like that they chose a benchmarking methodology that aligns to the actual application of RAG in search and is less academic than many used today. Plus it's good for the AI leader to note that there are some lower-profile companies like CustomGPT that offer highly performant RAG solutions - CustomGPT beat Open AI's RAG solution in hallucination rate, accuracy and latency in these tests.
07/22/2024 Essential The Prompt Report: A Systematic Survey of Prompting Techniques The authors of this meta-analysis have done a nice job of creating a comprehensive manual for prompt engineering and establishing a foundation for a common taxonomy for this nascent field. Being that developers are still learning how to do this well, we marked this as essential. AI leaders should read it and pass it on to their teams.
07/22/2024 Important The Data That Powers A.I. Is Disappearing Fast Companies are increasingly using the robots.txt file to block web crawlers. This NYT article references a study that found that about 5 percent of all data and 25 percent of quality data in three commonly used training datasets is now restricted. As we know, model makers are clamoring to find new data to improve their models' performance - it's not going to get any easier to do so.
07/22/2024 Important Robotics won’t have a ChatGPT Moment A nice compendium of the progress in robotics and how AI is accelerating that, though it will be a gradual road to robots becoming pervasive. The authors predict that the cost of humanoid hardware will drop below the cost of labor sometime in 2026 - a bold prediction. Be sure to flip through the presentation embedded in this one.
07/22/2024 Optional Salesforce to release autonomous AI customer service agents The company is testing an "Einstein Service Agent" that reportedly will operate across messaging platforms like What'sApp to help a company's customers solve their issues. Will it work or be incredibly frustrating? We don't yet know, and there's no timetable on when this will be available so we're marking this optional for now.
07/22/2024 Optional Apple shows off open AI prowess: new models outperform Mistral and Hugging Face offerings Apple has partnered with Toyota Research and a few others to test data curation techniques to see which ones made two smaller models (7B and 1.4B) perform best. And they've open-sourced the two models. Interesting, but optional.
07/22/2024 Optional Figma explains how its AI tool ripped off Apple’s design We continue to applaud Figma for the way it has handled the problem of their GenAI-powered 'Make Designs" tool spitting out designs from Apple. In addition to pulling the tool down, they've now published to their website a full explanation of how it happened. AI leaders might want to note their approach to addressing the screw-up in case they have a similar issue, but otherwise optional.
07/23/2024 Important Confronting Impossible Futures This article from GenAI thought leader Ethan Mollick discusses how companies need to plan for the different potential development paths that AI could take - a capabilities plateau, linear growth in capabilities, exponential growth, or AGI. He poses that not enough companies are doing this and should be. Not the most well-baked scenario planning, but still a worthy read.
07/23/2024 Important From Principles to Practices: Lessons Learned from Applying Partnership on AI's (PAI) Synthetic Media Framework to 11 Use Cases This is a good piece from the Partnership on AI that lays out best practices to ethically use GenAI to generate synthetic media. It uses several real cases, eg OpenAI building disclosures into DALL-E generated imagery, to inform a framework AI leaders can use in practice. We think this is worth your time to read.
07/23/2024 Optional Quest IndexGPT: Harnessing generative AI for investable indices Interesting that JP Morgan is using GenAI in a core part of its business, but it seems that they're only using GPT-4 to search for keywords in news articles to help them put together indices based on a particular theme for clients. Seems like a bit of a GenAI paint job so we're marking it optional.
07/23/2024 Optional Planning for Agents This blog post from Langchain points out that, to get agents to do complex planning and reasoning tasks, you need content, context, domain knowledge and other elements. It's pretty basic and ultimately is steering you toward their products, so we're marking this as optional.
07/23/2024 Optional New compliance and administrative tools for ChatGPT Enterprise OpenAI is playing catchup here on offering basic tools like access controls that enterprises need for compliance. Good to note, but this announcement isn't worth digging deeply into.
07/23/2024 Optional Cohere raises $500M to beat back generative AI rivals Impressive that Cohere has managed another huge raise at a $5.5B valuation off of just $35M in sales. Big names like Nvidia, AMD and Cisco were participants. Bubblicious or fair? You decide...
07/24/2024 Essential Meta releases the biggest and best open-source AI model yet As you may have heard, Meta dropped Llama 3.1 yesterday, its 405B open source-ish model (training data was not released) that it purports to be competitive with the frontier models. Meta teaming up with Groq, which offers cheaper AI chip technology than competitors, could lead to the commoditization of inference. Is the Linux moment for LLMs near?
07/24/2024 Important Open Source AI Is the Path Forward Mark Zuckerberg explains why Meta is pushing open source, noting something we've been talking about - many companies want to "own their own intelligence" rather than port it to a hosted closed-source model. This is a well-written and supported piece that's worth a read.
07/24/2024 Important A year in: Nestlé employees save 45 minutes per week using internal generative AI A good case study for a well-known company. Nestle has really committed to training its employees on GenAI tools and its efforts are bearing fruit. We see a lot of companies making the mistake of simply turning on a tool like Copilot and just expecting employees to understand how to use it. They're wasting their money on these subscriptions if they don't help employees figure out how to leverage them. Read this article to see how it's done.
07/24/2024 Important Legal software company Clio raises massive $900M round to power AI advances Clio is akin to the SAP of law, and it's looking to upgrade its platform using this sizable new raise. The company says it hopes the productivity gains that will come from infusing GenAI in its platform will open legal access to companies and individuals that could not otherwise afford it. That's certainly possible, as is law firm partners simply pocketing the savings... Either way, important to be aware of what's happening in the sector as it will spread to others soon enough.
07/24/2024 Optional Introducing Rerank 3 Nimble: Faster Reranking for Enterprise Search & Retrieval-Augmented Generation (RAG) Systems Cohere is pretty laser-focused on RAG and enterprise search. Good to see an upgrade to its offering, but no need to read the article. Just be aware of the improvements.
07/24/2024 Optional Trust at scale: Auto-evaluation for high-stakes LLM accuracy This is a basic educational piece on the movement toward some level of automated evaluation of LLM outputs that's not particularly well-written. Skip it.
07/25/2024 Essential Mistral Large 2 Another day, another model release. Mistral claims this model is competitive with frontier models on coding, math and reasoning. And they've trained it to tell the user when it doesn't know the answer to minimize hallucinations. A license from Mistral is required for commercial use. Uptake of this one TBD, of course, but definitely of note.
07/25/2024 Important ‘Model collapse’: Scientists warn against letting AI eat its own tail While this isn't the first time we've seen publishing on the potential for LLMs to lose their capabilities as they consume more and more AI-generated content, the research on which this article is based is well-done and worth a read.
07/25/2024 Optional Stability AI steps into a new gen AI dimension with Stable Video 4D A cool development for game creators, animation artists, and movie makers, but no need to read this article.
07/25/2024 Optional Airtable’s New AI Tool Can Generate Apps From Just A Prompt A nice tool that uses uses OpenAI's GPT-4 under the hood, but the article title pretty much tells you all you need to know here.
07/25/2024 Optional Improving Model Safety Behavior with Rule-Based Rewards As this post from OpenAI states, the company has been using this machine-based method for alignment in addition to human feedback for a while now, so this feels like a "yes-we-blew-up-our-safety-team-but-we-still-are-being-safe" kinda piece as opposed to anything earth-shattering.
07/25/2024 Optional How Colonial First State is using AI to transform wealth management This is an article from Microsoft with a lofty title that the GenAI application doesn't remotely live up to. It essentially shares that Colonial is using Copilot.
07/26/2024 Essential OpenAI announces SearchGPT, its AI-powered search engine While the company is only pushing out a prototype to 10K test users at this point, this truly kicks off Search Wars 2.0. OpenAI notes that they're paying publishers for content that will get served up by this (though unlikely it will ONLY surface purchased content) after they've sucked the Internet dry for free to contrast themselves with Perplexity. We're all on the waitlist to test the tool, but will report back on the experience if we get access.
07/26/2024 Important AI achieves silver-medal standard solving International Mathematical Olympiad problems Google DeepMind shares some info about progress with its AlphaGeometry 2 and AlphaProof models in math and reasoning. Apparently Sam Altman responded to Google's X post on this with an "lol" which might indicate he thinks this is child's play compared to something OpenAI is working on but, who knows, and we think the AI leader should keep track of important research by top computer scientists so we marked this as a worthy weekend read.
07/26/2024 Important Who will control the future of AI? The fact that Sam Altman wrote an opinion piece on the geopolitics around AI for the Washington post is of note (albeit with clear motives to influence the government). While we didn't find it to contribute any groundbreaking new thinking, it's a decent summary of the issues.
07/26/2024 Optional Anthropic’s crawler is ignoring websites’ anti-AI scraping policies The website for iFixIt was apparently getting pinged like crazy by Anthropic's bots, but once iFixIt implemented robot.txt, the bots stopped. So this really isn't anything more than a reminder to do the same if you want to try to turn away scrapers (though no guarantees it will work as we've seen Perplexity's bots ignore this signal).
07/26/2024 Optional OpenDevin: An Open Platform for AI Software Developers as Generalist Agents While noteworthy that this team has open-sourced an agent-building platform, no need to read this research paper about until you're ready to use the platform. Bookmark this for the time being.
07/26/2024 Optional Generative AI for urban simulation We surfaced this piece thinking it would be an interesting use case to discuss, but then found it was so poorly written and unclear that it we quickly marked it optional. Skip it.
07/29/2024 Important Snowflake ropes in AI21’s Jamba-Instruct to help enterprises decode long documents Snowflake offers access to many different models so adding AI21's model isn't why we're marking this important. What we want to flag is Snowflake's model evaluation tool which is discussed briefly here. Databricks offers this as well. A key capability for the AI leader's development team.
07/29/2024 Optional DBS rolls out GenAI assistant for customer service teams Singapore-based DBS is known to be a leader in tech innovation so no surprise that they've piloted a GenAI-powered customer service agent assistant. The article notes that it's moving to production soon.
07/29/2024 Important Morgan Stanley moves forward on homegrown AI Homegrown is a bit of a misnomer - they're using a customized OpenAI model. But the article notes that they've taken this approach as opposed to using the GenAI tools built into their other applications (Salesforce, Microsoft, etc.) to have a unified backend and ease the burden of employees having to learn a number of different tools. The article also notes that many other banks are taking a similar approach.
07/29/2024 Optional Recursive Introspection: Teaching Language Model Agents How to Self-Improve This is an interesting approach that shows some promise in enabling models to learn from their mistakes without needing outside inputs from a human or another model. But the authors note it's early days and there's more work to do to refine the method, so this is optional for the moment.
07/29/2024 Optional The AI race’s biggest shift yet A brief, well-written commentary about the shift of focus from models to applications within the GenAI stack given that there are now plenty of open and closed models to choose from. But this isn't groundbreaking news so we parked it in the optional lot.
07/31/2024 Essential Introducing SAM 2: The next generation of Meta Segment Anything Model Another open source model from the tech giant, and this time they're including the training dataset. The model enables identification of objects in video and can retrain itself on the fly to identify them in future video frames. It's reported to be orders of magnitude better then the previous version. This is going to be huge for robotics, military and design eventually, and it will impact medical imaging and other areas in the near term.
07/31/2024 Important When Generative AI Meets Product Development Three case studies in this one: LLMs as a natural language interface for complex design tools, as a source of customer insights and concept validation, and for enhancing creativity in design workflows. Proof that GenAI can help introduce products to market faster.
07/31/2024 Important Meta releases AI Studio with Llama 3.1 After rumors that Meta and Character.ai were discussing a partnership, Meta seems to have forged ahead on its own, releasing this tool that enables users to create their own chatbots and share them with others. Creators can also have the chatbots respond to direct messages - will major brands embrace this tool for that purpose as well? Either way, it seems AI is becoming the new UI.
07/31/2024 Important Lawmakers want to carve out intimate AI deepfakes from Section 230 immunity This effort seeks to modify a 1934 law that has largely protected big tech platforms like Meta from being held responsible for the posts their users create. It will be important if it passes and has some teeth. Keep this on your radar.
07/31/2024 Important Nvidia accelerates human robotics development Nvidia is offering new services and tools that will enable developers to streamline workflows and generate the data needed to train robots, potentially cutting development and deployment cycle times from months to under a week. We would have marked it essential if all the offerings were already available, but some aren't.
07/31/2024 Important Apple Intelligence Foundation Language Models Well-written and researched paper on how to think about improving models for on-device, practice responsible AI and more. Aspects of it could serve as a manual for AI teams, so we're marking it as important.
07/31/2024 Optional Perplexity details plan to share ad revenue with outlets cited by its AI chatbot When an article is cited, Perplexity will now share revenue from the related ads that pop up alongside it with the article's publisher. Fortune, Time and some other major publishers are participating in this already. Perplexity had been getting a lot of flack for ripping off publishers' content, and now they're offering a model that arguably benefits publishers more than Google. A good story and important for the GenAI search race in general, but it's not necessarily important to the day-to-day of an AI leader.
07/31/2024 Optional Serverless Inference with Hugging Face and NVIDIA NIMs Good to be aware of, not necessary to read this -- pass it on to your developers if they're using Hugging Face.
07/31/2024 Optional Verizon sees early success using generative AI to answer questions from business customers Nice use case, but not enough detail to bother reading this one.
08/01/2024 Essential Death of a Salesforce: Why AI Will Transform the Next Generation of Sales Tech This piece from Andreesen Horowitz discusses how GenAI is poised to transform the way salesforces work, which will result in sales, marketing and customer success merging (so it's more of a death of traditional sales rather than an elimination of salespeople altogether). The market map of sales-focused GenAI startups they provide shows the bevy of activity happening in the space. AI leaders need to be aware of this and flag it for CEOs - a transformation in how top-line growth is achieved is something they'll certainly want to know about.
08/01/2024 Important Generative AI in Real-World Workplaces Microsoft has done a meta analysis of studies, both in the lab and the real world, on how GenAI tools (like Copilot, of course) are driving productivity improvements for certain activities and roles. This is self-serving, but it's is a good summary of research and features some interesting data points. Worth a read.
08/01/2024 Optional Writer’s new AI models are scary good at healthcare and finance tasks Writer has been expanding their offerings beyond their original focus on marketing use cases. Here they've created a few specialized models that they're open sourcing. The model specialization trend is notable, but established at this point, so we marked this optional.
08/01/2024 Optional OpenAI launches experimental GPT-4o Long Output model with 16X token capacity OpenAI is working on dividing up their 128K token window differently so that you can, for example, provide a 64K input and get up to a 64K output (today the output capacity is much lower). This will be helpful for coding and writing use cases, but it's not yet available so this is optional for the moment.
08/01/2024 Optional OpenAI has released a new ChatGPT bot that you can talk to This is one of the features OpenAI previewed in May during the presentation that used its "Sky" voice that sounded like Scarlett Johansson. The ChatGPT voices will respond faster and can vary their emotional tone. Other functions they previewed, such as pointing your phone at a handwritten math problem and asking ChatGPT to solve it, are not part of this upgrade.
08/01/2024 Optional Neura shows off humanoid robot 4NE-1 Neura is a robotics manufacturer out of Germany that was given early access to some of the robot training tools Nvidia has been talking about at the Siggraph conference this week. The company published a video of a humanoid that was trained using these tools. It's highly edited so tough to say that their effort is groundbreaking, but we're excited about the improvements in robots on the horizon.
08/02/2024 Important Google Gemini 1.5 Pro leaps ahead in AI race, challenging GPT-4o Google is only releasing this for early testing or we would have marked this as Essential, given it's part of what's still only a handful of bigger, super-powerful models. It looks like it may be "better" than 4o, but let's wait and see what the market thinks when it's able to put these in an application in production.
08/02/2024 Important GitHub Models gives developers new power to experiment with Gen AI Hugging Face announced similar functionality just a few days ago, but this has an interesting tie to Azure. You can build on Hugging Face and then deploy through Azure right through the platform, making developers' lives a little easier.
08/02/2024 Important NVIDIA Researchers Harness Real-Time Gen AI to Build Immersive Desert World This is another announcement from Nvidia at the Siggraph conference. They demonstrated how they could build this 3D scene in just five minutes. This truly looks to be an agentic system. NVIDIA is clearly focusing on gaming and robotics, which is part of their omniverse/simulated reality play. It's time to start paying attention to this and think about how you can take advantage of it via digital twins, for example.
08/02/2024 Optional Google releases new ‘open’ AI models with a focus on safety Good to see more options for smaller, open models with these three new ones from Google. This is becoming a crowded space, so we're marking this news as optional.
08/02/2024 Optional NTIA Supports Open Models to Promote AI Innovation The US Department of Commerce's National Telecommunications and Information Administration put out some policy recommendations that note that open models aren't regulated, so they should be monitored. But there's no action being taken here, so we're marking it as one to watch, but optional for now.
08/02/2024 Optional Prompt Design at Character.AI The company's new Prompt Poet is a tool that helps both developers and (questionably) the average user design and manage their prompts. Looks promising - share it with your developers as something that could possibly be helpful, but the AI leader doesn't need to dig into this too deeply.
08/02/2024 Optional Networks of Networks: Complexity Class Principles Applied to Compound AI Systems Design The paper highlights the effectiveness of separating generation and verification tasks, an approach grounded in complexity theory. By leveraging a system where multiple language models generate answers and another model verifies them, performance improvements are seen in tasks where verification is simpler than generation. Interesting but this gets to a technical level not necessary for an AI leader to get into - read it this weekend if you want to geek out a bit, but otherwise skip it.
08/05/2024 Important GenAI strategy dictates ROI challenges for IT leaders This article discusses the challenges AI leaders face as they look to quantity the benefits of licenses for GenAI applications (eg Copilot) and work to control costs when they're tapping LLMs in the cloud. It provides solutions, such as negotiating licensing costs, conducting employee surveys to track benefits and using container management to reduce cloud bills. One interesting back-of-the-envelope ROI calculation from the piece: FTEs making an average of $54/hr only need to get ~30 minutes of time savings monthly to justify the $30 Copilot fee.
08/05/2024 Important Black Forest Labs Open-Source FLUX.1 We found the capabilities of this new open text-to-image family of models and the accompanying developer instructions for implementing them impressive, so we're flagging these as worth looking into.
08/05/2024 Optional Microsoft now lists OpenAI as a competitor in AI and search Thank you, Captain Obvious.
08/05/2024 Optional MoMa: Efficient Early-Fusion Pre-training with Mixture of Modality-Aware Experts Interesting research on a mixture of experts approach that uses different LLMs for various modalities (text,to-image, etc) to gain efficiencies. But it's very forward-looking so we're marking optional for now.
08/05/2024 Optional Character.AI CEO Noam Shazeer returns to Google Google takes a page out of Microsoft's acquihire/licensing book with its Character.ai deal. Character had been doing some cutting edge things to make inference more efficient, which will be important for Google, particularly as it increasingly integrates GenAI into search. Good deal for them, more for regulators to scrutinize.
08/05/2024 Optional On speaking to AI Ethan Mollick often gets early access to GenAI tools given his mega-influencer status. Here he compares the upgraded ChatGPT voice mode with an early version of the "new," Siri noting the different approaches the companies are taking (more capabilities/more potential dangers vs limited capabilities/more guardrails). Interesting, but optional.
08/05/2024 Optional The Great AI Unbundling An article from the CEO of an "AI wrapper" company, Every, which argues that wrappers have value. He likens the current LLM chatbots to Excel, and says many SaaS products are "wrapped" versions of what Excel can do. Maybe we could buy that, but hard to take seriously when the company is making an argument for the viability of its own product.
08/06/2024 Important OpenAI Co-Founders Schulman and Brockman Step Back Schulman isn't just stepping back - he's moving to a competitor (Anthropic). More turmoil in OpenAI's leadership is something AI leaders should keep tabs on whether they rely on the company's LLMs or not. Is this a result of their reported mounting expenses? Is a takeover imminent? What will the market implications be if so?
08/06/2024 Optional AI chip start-up Groq’s value rises to $2.8bn as it takes on Nvidia Another raise on an eye-watering valuation. Groq's inference-only LPUs look promising, but this company has a long way to go to catch up to Nvidia. They're only expecting to make 105K units available by March of next year.
08/06/2024 Optional With Smugglers and Front Companies, China Is Skirting American A.I. Bans This is a well-researched piece on how China is still getting its hands on advanced AI hardware despite the US export controls. Not applicable to the AI leader's day-to-day, but an interesting read if you have the time.
08/06/2024 Optional Zoom Is Going After Google and Microsoft With AI-Driven Docs Zoom has to do something to stay afloat, but it's tough to envision companies switching away from the current workspace app heavyweights.
08/06/2024 Optional There’s a Tool to Catch Students Cheating With ChatGPT. OpenAI Hasn’t Released It. Will our education system adjust to the existence of AI and change the way we teach our children to encourage its use or keep trying to fight a losing battle? Either way, OpenAI is more concerned about turning off its users as they continue a 2-year internal debate about releasing this tool, which is reported to be 99.9% effective. In a survey, 30% of current ChatGPT users said they would use it less if it included watermarks on AI-generated text and there was an alternative tool available that didn't. All of this is interesting, but optional to the AI leader.
08/06/2024 Optional IBM Introduces New Generative AI-Powered Cybersecurity Assistant for Threat Detection and Response Services This is a tool for IBM's consultants delivering cyber as a managed service. We can assume it will move into IBM's product at some point. Either way, optional for now.
08/06/2024 Optional Amazon Q Developer just reached a $260 million dollar milestone The $260M represents Amazon's calculation of how much they've saved companies whose developers use Q to help them update applications. Nice use of a data point to push their product, but optional.
08/07/2024 Essential S&P Global partners with Accenture, launches massive AI training program for 35,000 employees S&P and Accenture realize the key to widespread productivity and leadership in the market begins with education of ALL of its employees.
08/07/2024 Important Argonne Leverages Generative AI To Empower Nuclear Plant Operators Argonne is leading the field by integrating an Argonne diagnostic tool called PRO-AID (Parameter-Free Reasoning Operator for Automated Identification and Diagnosis), a symbolic engine, and an LLM. PRO-AID works by comparing real-time data from the facility to expected normal behavior to determine if there is a fault. The symbolic engine acts as an intermediary between the PRO-AID and LLM, ensuring accuracy and reliability. LLM is used to explain the output. An amazing integration of deterministic and probabilistic tools to provide depth and transparency of critical information in an extremely demanding and complex decision making environment.
08/07/2024 Important RAG Foundry: A Framework for Enhancing LLMs for Retrieval Augmented Generation Intel Labs presents a new, open source framework that can accelerate RAG implementations. Its Data Augmentation module creates a dedicated dataset from RAG interactions that is then used in Training, Inference and Evaluation modules. Developers can now experiment with different RAG techniques easily and quickly to select the most appropriate for their application.
08/07/2024 Optional Bloomberg releases Gen AI-enhanced solution Useful new tools on mobile from Bloomberg for financial analysts, but little information on actual use or impact yet.
08/07/2024 Optional OpenAI-backed Figure shows off next-gen ‘self-correcting’ humanoid robot Figure 02 is certainly more advanced than Figure 01 introduced earlier this year with higher hands dexterity, ability to lift heavier loads and especially self-correcting capability. The robots are being piloted at the BMW plant. Humanoid robots movement is gaining momentum being helped by NVIDIA and its specialized tech stack.
08/07/2024 Optional The death of RAG 'RAG is here to stay' concludes this article after comparing using RAG vs long context windows. There is useful advice on when to use one versus the other.
08/07/2024 Optional ‘You are a helpful mail assistant,’ and other Apple Intelligence instructions Fun article on Apple's hidden prompt libraries.
08/08/2024 Important Introducing Structured Outputs in the API OpenAI released a new model gpt-4o-2024-08-06 with Structured Outputs that score a perfect 100%. In comparison, gpt-4-0613 scores less than 40%. Structured Outputs are API that can be enabled in two ways making the model adhere to developer-supplied JSON schemas. Certainly a major step in eliminating models' hallucinations.
08/08/2024 Important Generative AI Is Everywhere—Including At Birchwood Foods A quick read presenting a case study at a meat processing company that uses Gen AI for translating safety videos for plant employees who speak more than 130 languages. New videos can be developed in a matter of hours.
08/08/2024 Important What bosses miss about AI This article delivers an important message that the highest value from GenAi will be when companies learn to use this powerful tool for generating new ideas and creating new things or doing things in new ways. However, our analysts were not as impressed with the article itself.
08/08/2024 Optional Bringing Production-Ready GenAI to the Enterprise This is a transcript of an interview with Edo Liberty, CEO of Pinecone and Harrison Chase, CEO of LangChain. They talk about their companies' capabilities in navigating the gap from prototype to production. Nothing really new here but a good discussion of a number of best practices to be aware of.
08/08/2024 Optional LLM to ROI: How to scale gen AI in retail Nice value chain article on retail by McKinsey but lacks good use cases and factual data.
08/08/2024 Optional LG, Samsung eye South Korea's AI textbooks as edtech springboard LG and Samsung are piloting whiteboards and other tools in South Korea to boost school learning, targeting to be first with these tools in other markets. However, as Tim Andrews points out, the "teaching content is not in these machines" and doubts that these machines will advance education significantly.
08/09/2024 Optional Leveraging Gen AI to Augment the Human Experience in Insurance A BCG article describing their work with New York Life in Gen AI. Use cases presented are common and the article does not provide much detail on the results.
08/09/2024 Optional Hugging Face acquires XetHub from ex-Apple researchers for large AI model hosting This acquisition is a positive move for Hugging Face providing them with the ability to scale application hosting.
08/09/2024 Optional Introducing Qwen2-Math China based Qwen introduces two math-specific large language models based on Qwen2 series, Qwen2-Math and Qwen2-Math-Instruct-1.5B/7B/72B. Presented results show that these models outperform the mathematical capabilities of open-source models as well as closed-source models (e.g., GPT-4o).
08/09/2024 Optional LG unleashes South Korea’s first open-source AI, challenging global tech giants LG joins a crowded field of open-source AI models with the launch of Exaone 3.0, South Korea’s first open-source artificial intelligence model.
08/09/2024 Optional GPT-4o System Card A comprehensive report from OpenAI detailing the process and the results of the safety work that the company performs before releasing their models. Impressive but optional for our readers.
08/09/2024 Optional ProRata Invents Generative AI Attribution Technology to Compensate and Credit Content Owners While Facilitating Fairness and Fact A notable development led by Bill Gross, the inventor of the current pay-per-click monetization model for Internet search. Even though supported by several major publishers, it is just an early concept announcement.
08/09/2024 Optional Mistral Releases La Plateforme for Building AI Agents A catch-up move by Mistral.
08/13/2024 Essential How One Major Healthcare Firm Became the Leader in Innovative AI Use This article provides a crucial case study on Blue Cross Blue Shield Michigan's successful adoption of generative AI within a highly regulated industry. Its importance lies in the actionable insights and blueprint it offers, demonstrating how even legacy organizations can innovate and adapt to modern AI technologies, which is vital for AI leaders across industries.
08/13/2024 Important Thomson Reuters Unveils CoCounsel 2.0; Supercharged GenAI Assistant Combines the Power of Google Cloud AI, OpenAI, and Thomson Reuters The introduction of a new AI-powered legal tool by Thomson Reuters is a notable development, particularly for the legal sector, where AI tools are increasingly vital. The article discusses the integration of multiple large language models, including those from OpenAI and Google, reflecting the growing importance of AI automation in enhancing legal services and AI-driven content creation.
08/13/2024 Important Hiscox launches Google Cloud-backed GenAI underwriting model This article highlights the application of large language models in the complex and highly regulated field of reinsurance underwriting. While it lacks extensive data, it’s significant because it showcases a real-world, production-level implementation of AI in a risk-averse industry, marking a notable step forward for AI in finance and business sectors.
08/13/2024 Important Companies go all in on AI—5 charts show how This article offers valuable data-driven insights into how companies, particularly SMBs, are allocating their AI budgets, which can be crucial for AI marketing and AI sales strategies. While the scope may be limited to smaller businesses, the data presented can help AI leaders justify investments and understand market trends in AI adoption.
08/13/2024 Optional Genie - top SWE model While interesting, this article is primarily a product announcement with limited impact beyond the niche field of software engineering. The claims made are intriguing but require more evidence to assess their true value, making it less critical for a broader AI-focused audience.
08/13/2024 Optional AI Won’t Give You a New Sustainable Advantage Despite the article's important message about the limitations of AI in providing a lasting competitive edge, it falls short in its analysis and lacks substantial evidence. The concept is relevant, especially for AI strategy discussions, but the execution doesn’t offer enough depth to warrant a higher rating.
08/14/2024 Essential The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery The whole method of knowledge creation is being changed, and this is a significant step forward for not just scientific discovery. Its implications for AI education and the broader tech industry cannot be overstated, marking this as an essential read for AI leaders focused on the future of artificial intelligence and generative AI.
08/14/2024 Essential Gemini Live, Google’s answer to ChatGPT’s Advanced Voice Mode, launches The announcement of Google’s Gemini Live, a competitor to ChatGPT with advanced voice interaction capabilities, is critical in the ongoing AI race. The introduction of more natural, human-like conversational abilities highlights significant advancements in AI tools and large language models, further pushing the boundaries of what AI can achieve in terms of user experience.
08/14/2024 Important ToolSandbox: A Stateful, Conversational, Interactive Evaluation Benchmark for LLM Tool Use Capabilities Apple's Tool Sandbox offers a crucial evaluation framework for assessing large language models in real-world tool usage scenarios. This is particularly important for AI developers and educators as it provides insights into how models perform in complex, multi-agent environments, making it a valuable resource for AI leaders navigating the shift towards integrated AI systems.
08/14/2024 Optional Klarna’s AI chatbot: how revolutionary is it, really? This article provides a detailed analysis of Klarna’s AI chatbot and its impact on cost reduction. While informative, it doesn’t introduce groundbreaking insights, making it of lesser priority for AI leaders.
08/14/2024 Optional OpenAI updates ChatGPT to new model that exhibits multi-step reasoning The update to ChatGPT’s reasoning functionality is an incremental improvement rather than a significant advancement. It is of interest mainly to those closely following AI tools and chatbot developments.
08/14/2024 Optional Introducing Agent Q: Research Breakthrough for the Next Generation of AI Agents with Planning & Self Healing Capabilities While the concept of self-healing agents using Monte Carlo tree search is intriguing, the claims made in this article are overblown. The content is only relevant to niche AI development communities.
08/15/2024 Essential Embracing Gen AI at Work This article highlights the critical need for AI education and training within organizations. The article describes the three kinds of “fusion skills” you need to get the best results from gen AI: intelligent interrogation, judgment integration and reciprocal apprenticing. An essential read for any leader because as authors state: "The AI revolution is already here. Learning these three skills will prepare you to thrive in it."
08/15/2024 Important Snowflake launches Cortex Analyst, an agentic AI system for accurate data analytics Cortex Analyst is a conversational interface allowing enterprises to talk to their data. The reported accuracy is 79%. Cortex Analyst is trained on the enterprise data stored in the cloud and thus is more accurate than untrained LLMs that deliver ~50% accuracy. The accuracy can increase dramatically if customers can provide semantic schemas for their data that can be then used to check the Analyst's answers. The area of conversational interfaces for structured and unstructured data is a fast evolving space advanced by Snowflake, Databricks and others that is important to monitor and incorporate into your AI strategy.
08/15/2024 Important Prompt caching with Claude Prompt caching can significantly reduce the time and cost associated with repeated queries in AI-driven systems, particularly in high-volume environments like call centers or large-scale API interactions. The potential to streamline operations and improve efficiency makes this an important update for organizations utilizing generative AI.
08/15/2024 Optional xAI’s new Grok-2 chatbots bring AI image generation to X Although the Grog2 chatbot is described as a powerful new model from xAI, it remains in early testing stages, accessible only to a limited group of users. It also needs additional guardrails for image generation, especially in politics.
08/15/2024 Optional Hong Kong launches generative AI Sandbox for finance sector While this initiative represents a significant public-private partnership aimed at advancing AI use in the finance sector, the article provides limited details, making it less crucial for AI leaders but important for leaders in the finance sector.
08/15/2024 Optional GenAI Can’t Scale Without Responsible AI This article from BCG offers a high-level framework for implementing responsible AI practices, which is valuable but lacks actionable details on addressing complex issues like bias and toxicity in AI systems. While it provides a useful overview for AI leaders seeking to structure their AI strategies, the lack of practical guidance makes it less critical compared to other articles that offer more immediate value.
08/16/2024 Essential MIT releases comprehensive database of AI risks MIT's AI risk database, encompassing over 700 risks categorized across 43 taxonomies, is a crucial resource for AI leaders and security teams. This comprehensive and continually updated repository is essential for understanding the broad spectrum of risks associated with AI technologies, making it a fundamental tool for organizations aiming to mitigate these risks effectively. Analysts emphasized the importance of every AI leader ensuring their security teams are aware of and utilizing this resource.
08/16/2024 Important How Morningstar Leads AI Adoption On An Enterprise Scale Level This article discusses how Morningstar is leveraging AI, through Azure OpenAI, to unify its enterprise intelligence and enhance its internal processes. This approach is reflective of broader trends in enterprise AI adoption, where consolidating and unifying data to create interactive systems is becoming increasingly critical. The significance of these developments, especially in a company led by a visionary like Joe Mansueto, justifies its importance in understanding how AI can transform traditional industries.
08/16/2024 Optional Falcon Mamba 7B’s powerful new AI architecture offers alternative to transformer models The Falcon Mamba model, developed by the Technology Innovation Institute in Abu Dhabi, represents a significant development in AI with its innovative approach to model architecture. It is based on State Space Language Model (SSLM) which allows Falcon to ingest large inputs faster beating traditional transformer-based models even with large context windows. The model's potential impact on the AI landscape, despite being in its early stages, makes this article important for those tracking AI advancements.
08/16/2024 Optional How NVIDIA is using structured weight pruning and knowledge distillation to build new Llama models Nvidia is using Meta's latest open source models to create smaller models without losing their reasoning abilities. More about this work from NVIDIA in NVIDIA’s blog post. Certainly a note worthy development to watch.
08/16/2024 Optional Generative AI a cornerstone of NUHS’ healthcare IT strategy This article covers the integration of generative AI into Singapore’s NUHS healthcare IT strategy, supported by AWS and Bedrock. However, it lacks in-depth details and reads more like a case study for Bedrock's capabilities rather than offering groundbreaking insights into AI's role in healthcare. Consequently, it is rated as optional, particularly for those with a specific interest in AI applications in the healthcare sector.
08/19/2024 Optional New LLM Pre-training and Post-training Paradigms A detailed comparison of LLM pre-training and post-training approaches. It could be valuable for AI practitioners, but its applicability is limited to those working directly with large models.
08/19/2024 Optional Model Merging in LLMs, MLLMs, and Beyond: Methods, Theories, Applications and Opportunities Especially interesting for those involved in the future of AI development. Still a bit too theoretical and inapplicable for our AI leader.
08/19/2024 Optional Eric Schmidt’s AI prophecy: The next two years will shock you Schmidt’s comments were acknowledged as controversial but ultimately not groundbreaking. It's no surprise that things will continue to change even more quickly in the years to come.
08/19/2024 Optional Open-endedness is all we’ll need Agentic Systems are yet to be fully realized, and systems that can accommodate instruction and real-world variability will succeed.
08/19/2024 Optional Runway’s Gen-3 Alpha Turbo is here and can make AI videos faster than you can type The models keep getting faster and cheaper. If you want to give Runway a try for free, enter the upcoming AI film festival for a mountain of credits.
08/19/2024 Optional A.I. Is Helping to Launch New Businesses (and Not Just A.I. Businesses) The impact Generative AI can have on small businesses and startups is well known. Although some nice use cases, there was little of additional value in this piece.
08/20/2024 Essential A.L.S. Stole His Voice. A.I. Retrieved It. Through a combination of neural implants and deep fake technology, people with ALS are regaining their voice. With an astounding 98% accuracy, our analysts had to showcase the human impact of this technology.
08/20/2024 Important GenAI: From Breakthroughs to Bottom Lines This will help you understand the broader AI industry dynamics and how major companies like Microsoft, Google and Oracle are shaping that landscape. Important context for those looking at the future direction of AI investments.
08/20/2024 Optional Meet Hermes 3, the powerful new open source AI model that has existential crises Worth noting this model is uncensored and open sourced, but the broader importance is yet to be seen.
08/20/2024 Optional AI Chases the Storm: New NVIDIA Research Boosts Weather Prediction, Climate Simulation Although technically impressive, the implications beyond the niche scope of industries such as weather forecasting and transportation were limited.
08/20/2024 Optional Walmart Takes AI to Consumers as Amazon Focuses on B2B Using generative AI to update product pages across 850 million catalog data points is impressive. More depth, insight and analysis were warranted and left us wanting.
08/20/2024 Optional How eBay uses generative AI to make employees and online sellers more productive Similar to the Walmart piece, the use of generative AI by this titan is impressive, but little detail left us with little new information.
08/20/2024 Optional Optimally Allocating Compute Between Inference and Training An important topic, explained poorly and in a deeply technical manner. Difficult to connect this news to the more practical implications for our readers.
08/21/2024 Important From promising to productive: Real results from gen AI in services Although it ran a bit long, there were some solid points made in the early part of the article. Insights on the challenges of scaling AI solutions and how companies are realizing tangible benefits.
08/21/2024 Optional AMD to acquire infrastructure player ZT Systems for $4.9B to amp up its AI ecosystem play This move will help AMD to compete with Nvidia in the infrastructure space. These sorts of acquisitions are relatively common these days and $5B doesn't go as far as it used to.
08/21/2024 Optional Introducing the Next Version of Assistant Largely a product announcement, this could be considered essential for those in legal professions. Otherwise the relevance is not warranted.
08/21/2024 Optional Fine-tuning now available for GPT-4o Most SOTA foundation models offer fine-tuning these days, so this is an important but expected development. Notable for developers or engineers really building with these tools, but Optional for our general audience.
08/21/2024 Optional Meta’s Self-Taught Evaluator enables LLMs to create their own training data Innovation that addresses the challenges with getting good training data is important, but this specific development from Meta is Optional in the grand scheme of things. We'll wait and see what comes of this.
08/21/2024 Optional The big stack game of LLM poker A novel approach to examining the high stakes world of Large Language Models, but ultimately was a superficial observation with little depth or insight.
08/21/2024 Optional Microsoft releases powerful new Phi-3.5 models, beating Google, OpenAI and more Amazing what they can do with such small models, and the incremental improvements and cherry-picked benchmarks didn't impress us as much as they might have hoped.
08/22/2024 Important Skyfire lets AI agents spend your money The frictionless application of Blockchain to enable AI to spend on your behalf is worth taking note of. May not impact your day to day life (yet!), and it looks like we're heading towards a world where more and more gets done on your behalf.
08/22/2024 Important How Troutman Pepper’s Generative AI Assistant Athena Has Transformed Legal Services Delivery Building a foundation that connects their data with Microsoft and OpenAI first, truly lets Troutman Pepper own their own intelligence. Then bolting service after service on top of that is a winning formula.
08/22/2024 Important AI companies are pivoting from creating gods to building products A great overview of the major challenges when trying to turn models into products. They suggest the integration into enterprise will happen slower than we think.
08/22/2024 Optional OpenAI partners with Condé Nast Another logo, as OpenAI snatches up the news partner ecosystem, but not surprising nor earth shaking.
08/22/2024 Optional To Code, or Not To Code? Exploring Impact of Code in Pre-training Novel to consider that using code in pre-training could affect a model's language reasoning shows us how little we really understand about how they even work. Interesting academically, but of little importance to our practical day to day considerations.
08/23/2024 Essential We finally have a definition for open-source AI There's a critical need for clarity around what "open source" means in the AI space, particularly because the term is often misused by vendors, leading to significant market confusion. This topic is vital for anyone involved in AI development, as it impacts trust, ethics, and the future direction of the industry.
08/23/2024 Important The Jamba 1.5 Open Model Family: The Most Powerful and Efficient Long Context Models It's important to understand the value of different approaches to AI model design. AI21 labs' significant departure from the prevailing transformer-based models might challenge the dominance of transformers in the long run.
08/23/2024 Important Top 100 Consumer Gen AI apps The report's statistics and rankings may prove valuable to understand the shifting dynamics in the app market, particularly in how AI-driven tools are gaining prominence. An interesting and informative read for those tracking developments in consumer-facing AI technologies.
08/23/2024 Optional Input Coffee, Output Code: How AI Will Turn Capital into Labor AI enables the conversion of capital into labor by automating tasks at scale. This article took too long to get to this fairly obvious point.
08/23/2024 Optional No one’s ready for this Google's new Pixel phone makes it even easier to create images that are indistinguishable from real ones. Deepfakes are a well-known issue, and this article doesn't advance the conversation in a meaningful way.
08/23/2024 Optional Midjourney opens website to all users, offering 25 free AI image generations While the accessibility improvement and the shift from Discord to a browser-based interface is appreciated, this development doesn't significantly alter the image generation model landscape, especially with other similar tools available.
08/26/2024 Important Anysphere Raises $60M For AI-Powered Coding Tool, Cursor The broader sector of coding assistants is taking off quickly, with significant measurable productivity gains to be had. Founded by MIT alumni and backed by OpenAI and a16z, Cursor seems to be taking the lead.
08/26/2024 Optional Editing Files at 1000 Tokens per Second From the research group at Cursor, this article is just one of many examples of how they are reinventing the coding assistant.
08/26/2024 Optional Meet Einstein SDR and Einstein Sales Coach While these tools are useful and represent the ongoing trend of vendors integrating AI into their offerings, they are not groundbreaking innovations and are primarily relevant to existing Salesforce users.
08/26/2024 Optional Open source Dracarys models ignite generative AI fired coding Announcement of a new fine-tuning model for better coding.
08/26/2024 Optional Bayer Crop Science blends gen AI and data science for innovative edge Incredible promise, and not expected until next year at the soonest. We look forward to hearing more.
08/26/2024 Optional HappyFox Automates Support Agent Responses with Claude in Amazon Bedrock, Increasing Ticket Resolution by 40% It's great that they've seen such significant results. An expected outcome of a common use case.
08/26/2024 Optional Automated Design of Agentic Systems An Agent Factory that designs, tests and iterates on Agent designs. Will have to be carefully monitored as this could be a powerful new approach to autonomy.
08/27/2024 Important AWS empowers sales teams using generative AI solution built on Amazon Bedrock This article, while vendor-driven, provides a comprehensive overview of how AWS is using Generative AI tools built on Amazon Bedrock to support sales teams. The level of detail regarding implementation, structure, and best practices makes it valuable for AI leaders, particularly those interested in sales enablement through AI. It also reflects the growing maturity of AI applications within enterprise operations, positioning it as an important case study.
08/27/2024 Important Toward a Horizontal Robotics Platform This article discusses the broader implications of advancing robotics technology towards more generalized, horizontally integrated systems. While the content is dense and more relevant to those directly involved in robotics, the potential impact on industries like automotive and manufacturing makes it important for leaders in those sectors. The emphasis on the convergence of various robotics technologies and their growing importance in industrial processes highlights a significant trend.
08/27/2024 Optional AI-powered coding pulls in almost $1bn of funding to claim ‘killer app’ status Coding is already recognized as one of the top applications for generative AI, making the article more of a reiteration of existing trends rather than introducing anything novel. The article raises some questions about monetization and the balance between investment and returns, but overall, it was deemed as not offering enough new insights to warrant a higher rating.
08/27/2024 Optional IBM unveils next-gen AI chips at Hot Chips 2024 This article focuses on IBM’s ongoing developments in AI chip technology, which are an extension of their existing work on Z Systems. While it highlights incremental improvements, it primarily discusses advancements in processing power and operational capacity rather than introducing transformative innovations. As such, it's more relevant to those with specific interest in IBM's hardware developments.
08/27/2024 Optional Introducing Pharia-1-LLM: transparent and compliant Aleph Alpha's launch of an EU-compliant LLM tailored to German, French, and Spanish languages and automotive and engineering domains is noteworthy, especially for those focused on European markets. However, the article's in-depth technical details and focus on compliance make it more suited for developers and those involved in model selection rather than broader AI leadership. It's informative but not critical for most AI leaders.
08/27/2024 Optional Sapiens: Foundation for Human Vision Models Sapiens, the technology being developed by Meta Reality Labs, is astonishing in its potential, showcasing cutting-edge research in human vision models. However, it's still in the research phase with unclear implications for mainstream applications. While fascinating, it doesn't yet offer actionable insights for AI leaders, making it more of a niche interest for those tracking the latest in AI research.
08/28/2024 Essential Introducing Cerebras Inference: AI at Instant Speed Cerebras' breakthrough AI wafer offers a 22x Llama 3.1 70B inference speed improvement at one-fifth the cost compared to NVIDIA H100 Cloud chips. This development is a significant challenge to NVIDIA's dominance in the AI hardware market, providing a viable alternative for enterprises focused on reducing inference costs and improving speed. Given its potential to disrupt the AI hardware space, this article is critical for AI leaders to understand the evolving landscape.
08/28/2024 Important From Prototype to Prompt: NVIDIA NIM Agent Blueprints NVIDIA's continued push in the enterprise AI market with their NIM agent blueprints, designed to improve inference speed and efficiency across various applications, is significant. While the news reflects incremental advancements, NVIDIA's dominant market position makes it important for AI leaders to be aware of these updates as they could impact AI deployment strategies.
08/28/2024 Optional Anthropic publishes the 'system prompts' that make Claude tick While Anthropic’s decision to publish system prompts reflects a commitment to transparency, the information is largely of interest to AI practitioners rather than strategic leaders. The prompts have already been exposed through jailbreaking, making the official release more symbolic than groundbreaking.
08/28/2024 Optional Zero-Based Redesign: The Key to Realizing Gen AI’s Cost Savings Potential This article covers Bain’s approach to achieving AI cost savings through zero-based redesign. The concepts discussed, such as aligning AI projects with strategic goals and applying process re-engineering before engaging in AI implementations, are well-known and reiterate established best practices rather than offering new insights.
08/28/2024 Optional DeepMind and UC Berkeley shows how to make the most of LLM inference-time compute In this important research, DeepMind and UC Berkeley focused on improving LLM inference efficiency exploring several novel approaches. However, this research remains early-stage and more relevant to future AI model development. Note that more and more effort is going into improving LLM inference speed and reducing latency.
08/28/2024 Optional OpenAI races to launch Strawberry reasoning engine OpenAI’s efforts to enhance reasoning within their models through the Strawberry initiative are noted, but the article is more speculative and less impactful in the immediate term. Again, note that Strawberry applies approaches to reduce inference time confirming AI inference optimization trend.
07/29/2024 Essential When A.I.’s Output Is a Threat to A.I. Itself This article provides a good overview of a scenario when synthetic data is used and reused in AI model training leading to the model collapse. The issue is not new and many approaches are being developed to address it. We rated this article as essential so that AI leaders are aware of the issue and can respond to the nay-Sayers.
07/29/2024 Important The On‑Device Intelligence Update The article discusses Cartesia's use of a new state-space model (SSM) for on-device intelligence, which presents an alternative to the transformer architecture. This development is significant because it advances the possibility of running AI models on edge devices, which has major implications for privacy, security, and resource efficiency. Although not a breakthrough, the potential of this technology to influence future AI applications makes it important for leaders to monitor.
07/29/2024 Important California State Assembly passes sweeping AI safety bill This bill, while somewhat diluted, still holds importance as it introduces regulatory measures that could impact AI development, particularly for companies operating in California. Analysts noted the potential for ambiguity in the law, which could impose large compliance costs and challenges. AI leaders need to be aware of these developments to adjust their strategies and ensure legal compliance.
07/29/2024 Optional NVIDIA Blackwell Sets New Standard for Generative AI in MLPerf Inference Debut While NVIDIA's Blackwell chip introduces significant performance improvements, the article is largely seen as a product announcement without immediate or groundbreaking implications for AI leaders. The improvements, although impressive, are part of ongoing incremental advances that do not fundamentally alter the current AI landscape.
07/29/2024 Optional Diffusion Models Are Real-Time Game Engines The innovation of using diffusion models to achieve real-time game rendering is really impressive, particularly for the gaming industry. However, the broader implications for AI leaders are limited, as the development is highly specialized and does not yet have widespread application outside of gaming.
07/29/2024 Optional Google’s Gemini AI gets major upgrade with ‘Gems’ assistants and Imagen 3 Google's launch of Gems as part of its Gemini model represents a catch-up effort in the competitive AI landscape. While this development is noteworthy, it does not offer substantial new capabilities that would necessitate immediate attention from AI leaders, especially those already using other AI environments like Microsoft's.
08/29/2024 Essential When A.I.’s Output Is a Threat to A.I. Itself This article provides a good overview of a scenario when synthetic data is used and reused in AI model training leading to the model collapse. The issue is not new and many approaches are being developed to address it. We rated this article as essential so that AI leaders are aware of the issue and can respond to the nay-Sayers.
08/29/2024 Important The On‑Device Intelligence Update The article discusses Cartesia's use of a new state-space model (SSM) for on-device intelligence, which presents an alternative to the transformer architecture. This development is significant because it advances the possibility of running AI models on edge devices, which has major implications for privacy, security, and resource efficiency. Although not a breakthrough, the potential of this technology to influence future AI applications makes it important for leaders to monitor.
08/29/2024 Important California State Assembly passes sweeping AI safety bill This bill, while somewhat diluted, still holds importance as it introduces regulatory measures that could impact AI development, particularly for companies operating in California. Analysts noted the potential for ambiguity in the law, which could impose large compliance costs and challenges. AI leaders need to be aware of these developments to adjust their strategies and ensure legal compliance.
08/29/2024 Optional NVIDIA Blackwell Sets New Standard for Generative AI in MLPerf Inference Debut While NVIDIA's Blackwell chip introduces significant performance improvements, the article is largely seen as a product announcement without immediate or groundbreaking implications for AI leaders. The improvements, although impressive, are part of ongoing incremental advances that do not fundamentally alter the current AI landscape.
08/29/2024 Optional Diffusion Models Are Real-Time Game Engines The innovation of using diffusion models to achieve real-time game rendering is really impressive, particularly for the gaming industry. However, the broader implications for AI leaders are limited, as the development is highly specialized and does not yet have widespread application outside of gaming.
08/29/2024 Optional Google’s Gemini AI gets major upgrade with ‘Gems’ assistants and Imagen 3 Google's launch of Gems as part of its Gemini model represents a catch-up effort in the competitive AI landscape. While this development is noteworthy, it does not offer substantial new capabilities that would necessitate immediate attention from AI leaders, especially those already using other AI environments like Microsoft's.
07/30/2024 Essential With 10x growth since 2023, Llama is the leading engine of AI innovation Meta's Llama is an open-source alternative in the AI space that had an impressive 10X growth since Llama 3.1 release this summer. The analysts highlighted the strategic importance of Meta’s partnerships with major players like Microsoft, Google Cloud, and Nvidia, framing Llama as a potential leader challenging the dominance of closed models like those from OpenAI.
07/30/2024 Essential Why Honeywell has placed such a big bet on gen AI Honeywell’s comprehensive adoption of Generative AI across its enterprise, including training 95,000 employees and implementing 16 AI applications in production, is of value as an impressive use case of how GenAI can be implemented across enterprise. The discussion underscored Honeywell as a prime example of a legacy company successfully integrating AI at scale, demonstrating best practices in AI deployment that are vital for any organization looking to modernize with AI.
07/30/2024 Essential OpenAI and Anthropic will share their models with the US government This article is essential as it discusses a groundbreaking move where OpenAI and Anthropic agree to share their models with the U.S. government. The conversation focused on the potential regulatory implications and how this could influence the future of AI governance. This voluntary sharing is seen as a critical step in shaping the regulatory framework, impacting how AI models are deployed and trusted in regulated industries.
07/30/2024 Important Apple, Nvidia Are in Talks to Invest in OpenAI This article discusses the potential investment by Apple and Nvidia in OpenAI, which could increase the valuation of OpenAI to $100 billion. While speculative, the discussion noted the strategic implications for the AI market, particularly how such investments could shift power dynamics among tech giants and influence the competitive landscape.
07/30/2024 Important The gen AI skills revolution: Rethinking your talent strategy McKinsey’s article on the skills revolution triggered by GenAI is important as it addresses the evolving talent strategy needed for effective AI adoption in enterprises. The discussion highlighted the importance of integrating AI throughout the enterprise and the need for a strategic approach to talent development, making it relevant for organizations planning to scale their AI capabilities.
07/30/2024 Important 100M Token Context Windows This article highlights Magic’s development of a 100 million token context window, which is significant for software development. The extended context window enables more comprehensive and accurate processing of vast amounts of code and documentation, representing a technical advancement that could impact software development practices across industries.
07/30/2024 Optional Balyasny wants to build an AI equivalent of a senior analyst. A recent breakthrough brings the hedge fund one step closer. This article is rated as optional as it covers a specific case study of a hedge fund using Generative AI to replicate the functions of a senior analyst. While an interesting use case, the discussion indicated that it lacks broader implications for the AI industry, making it less critical for a wider audience.
07/30/2024 Essential With 10x growth since 2023, Llama is the leading engine of AI innovation Meta's Llama is an open-source alternative in the AI space that had an impressive 10X growth since Llama 3.1 release this summer. The analysts highlighted the strategic importance of Meta’s partnerships with major players like Microsoft, Google Cloud, and Nvidia, framing Llama as a potential leader challenging the dominance of closed models like those from OpenAI.
07/30/2024 Essential Why Honeywell has placed such a big bet on gen AI Honeywell’s comprehensive adoption of Generative AI across its enterprise, including training 95,000 employees and implementing 16 AI applications in production, is of value as an impressive use case of how GenAI can be implemented across enterprise. The discussion underscored Honeywell as a prime example of a legacy company successfully integrating AI at scale, demonstrating best practices in AI deployment that are vital for any organization looking to modernize with AI.
07/30/2024 Essential OpenAI and Anthropic will share their models with the US government This article is essential as it discusses a groundbreaking move where OpenAI and Anthropic agree to share their models with the U.S. government. The conversation focused on the potential regulatory implications and how this could influence the future of AI governance. This voluntary sharing is seen as a critical step in shaping the regulatory framework, impacting how AI models are deployed and trusted in regulated industries.
07/30/2024 Important Apple, Nvidia Are in Talks to Invest in OpenAI This article discusses the potential investment by Apple and Nvidia in OpenAI, which could increase the valuation of OpenAI to $100 billion. While speculative, the discussion noted the strategic implications for the AI market, particularly how such investments could shift power dynamics among tech giants and influence the competitive landscape.
07/30/2024 Important The gen AI skills revolution: Rethinking your talent strategy McKinsey’s article on the skills revolution triggered by GenAI is important as it addresses the evolving talent strategy needed for effective AI adoption in enterprises. The discussion highlighted the importance of integrating AI throughout the enterprise and the need for a strategic approach to talent development, making it relevant for organizations planning to scale their AI capabilities.
07/30/2024 Important 100M Token Context Windows This article highlights Magic’s development of a 100 million token context window, which is significant for software development. The extended context window enables more comprehensive and accurate processing of vast amounts of code and documentation, representing a technical advancement that could impact software development practices across industries.
07/30/2024 Optional Balyasny wants to build an AI equivalent of a senior analyst. A recent breakthrough brings the hedge fund one step closer. This article is rated as optional as it covers a specific case study of a hedge fund using Generative AI to replicate the functions of a senior analyst. While an interesting use case, the discussion indicated that it lacks broader implications for the AI industry, making it less critical for a wider audience.
08/30/2024 Essential With 10x growth since 2023, Llama is the leading engine of AI innovation Meta's Llama is an open-source alternative in the AI space that had an impressive 10X growth since Llama 3.1 release this summer. The analysts highlighted the strategic importance of Meta’s partnerships with major players like Microsoft, Google Cloud, and Nvidia, framing Llama as a potential leader challenging the dominance of closed models like those from OpenAI.
08/30/2024 Essential Why Honeywell has placed such a big bet on gen AI Honeywell’s comprehensive adoption of Generative AI across its enterprise, including training 95,000 employees and implementing 16 AI applications in production, is of value as an impressive use case of how GenAI can be implemented across enterprise. The discussion underscored Honeywell as a prime example of a legacy company successfully integrating AI at scale, demonstrating best practices in AI deployment that are vital for any organization looking to modernize with AI.
08/30/2024 Essential OpenAI and Anthropic will share their models with the US government This article is essential as it discusses a groundbreaking move where OpenAI and Anthropic agree to share their models with the U.S. government. The conversation focused on the potential regulatory implications and how this could influence the future of AI governance. This voluntary sharing is seen as a critical step in shaping the regulatory framework, impacting how AI models are deployed and trusted in regulated industries.
08/30/2024 Important Apple, Nvidia Are in Talks to Invest in OpenAI This article discusses the potential investment by Apple and Nvidia in OpenAI, which could increase the valuation of OpenAI to $100 billion. While speculative, the discussion noted the strategic implications for the AI market, particularly how such investments could shift power dynamics among tech giants and influence the competitive landscape.
08/30/2024 Important The gen AI skills revolution: Rethinking your talent strategy McKinsey’s article on the skills revolution triggered by GenAI is important as it addresses the evolving talent strategy needed for effective AI adoption in enterprises. The discussion highlighted the importance of integrating AI throughout the enterprise and the need for a strategic approach to talent development, making it relevant for organizations planning to scale their AI capabilities.
08/30/2024 Important 100M Token Context Windows This article highlights Magic’s development of a 100 million token context window, which is significant for software development. The extended context window enables more comprehensive and accurate processing of vast amounts of code and documentation, representing a technical advancement that could impact software development practices across industries.
08/30/2024 Optional Balyasny wants to build an AI equivalent of a senior analyst. A recent breakthrough brings the hedge fund one step closer. This article is rated as optional as it covers a specific case study of a hedge fund using Generative AI to replicate the functions of a senior analyst. While an interesting use case, the discussion indicated that it lacks broader implications for the AI industry, making it less critical for a wider audience.
09/02/2024 Essential How Do You Change a Chatbot’s Mind? This article delves into the complexities of managing personal, product, and company reputations in an AI-driven world. Businesses need to invest more time and resources into understanding and controlling how their brands are portrayed by AI, as the consequences of neglecting this could be severe.
09/02/2024 Essential GymNation Revolutionizes Fitness with AI Agents Powering Member Experiences This is an impressive use case and real-world example of how AI agents can significantly improve business outcomes. It also highlights how you can use a well-defined workflow with a platform like LlamaIndex to drive operational efficiency and profitability.
09/02/2024 Important Amazon’s new Alexa voice assistant will use Claude AI Amazon's move to enhance Alexa with Claude's models could influence user interaction with AI and set a new standard for voice interfaces. The importance lies in the broader implications for the AI-driven user experience and the competitive landscape among AI-powered devices.
09/02/2024 Important California lawmakers approve legislation to ban deepfakes, protect workers and regulate AI This legislative effort marks a significant step in addressing the growing concerns surrounding deepfakes and AI misuse, particularly during elections. While enforcement may face challenges and there are uncertainties about the laws' effectiveness, these measures are crucial in laying the groundwork for AI governance.
09/02/2024 Optional ChatGPT’s weekly users have doubled in less than a year While the growth is noteworthy, it follows an expected trend similar to other AI giants like NVIDIA, making it less critical for immediate attention.
09/02/2024 Optional Updates to the Command R Series Incremental updates to Cohere's existing models. While they bring efficiency, affordability, and user experience improvements, these product enhancements do not significantly impact the broader vendor landscape.
09/03/2024 Essential Manipulating Large Models to Improve Product Visibility This article highlights a significant breakthrough in AI-driven web search content optimization, particularly in the realm of product visibility. It offers a novel strategic text sequences (STS) method that can drastically improve product rankings, marking a pivotal shift from traditional SEO to AI-based optimization.
09/03/2024 Important Meta’s Transfusion model handles text and images in a single architecture This article discusses Meta's incremental advance in combining text and images within a single model architecture, potentially setting a new standard for multimodal AI models. Although it is still in the research phase and not yet scaled, the implications for future AI applications in handling complex, multimodal data are significant, making it an important read for those in the AI research community.
09/03/2024 Optional Nvidia’s Future Relies on Chips That Push Technology’s Limits The article highlights challenges that are already well-known in the industry for manufacturing complex chips. While informative, it offers little new insight, making it more of a background read for those interested in the hardware aspects of AI.
09/03/2024 Optional The Next Generation Pixar: How AI will Merge Film & Games This speculative piece from A16Z about the future of interactive movies blending film and games is intriguing but still highly speculative. The technology and market are not fully developed, making it an interesting but non-essential read for those curious about the future of AI in entertainment.
09/03/2024 Optional Why A.I. Isn’t Going to Make Art This opinion argues that AI will not replace human artists due to the unique decision-making involved in art creation. While it offers an interesting perspective, the argument is seen as lacking depth and relevance for those focused on the practical advancements in AI art, making it more of an optional read.
09/03/2024 Optional Google’s James Manyika: ‘The productivity gains from AI are not guaranteed’ This article discusses Google's CEO's views on AI's potential to increase productivity, emphasizing that gains are not guaranteed. It reiterates common themes about AI's role in business without providing substantial new insights, making it a low-priority read.
08/03/2024 Optional ‘Emotion AI’ may be the next trend for business software, and that could be problematic Emotion AI claims to be the more sophisticated sibling of sentiment analysis, leveraging the latest tech to detect human emotions. While this is an exciting concept, there are too many factors which can derail this development as a practical tool.
08/03/2024 Optional OpenAI searches for an answer to its copyright problems This article discusses various ways in which OpenAI can address the challenges it is facing regarding copyright problems. It covers a good set of themes; however, these have been discussed before.
08/03/2024 Optional Can AI Scaling Continue Through 2030? The article highlights challenges related to the scaling of AI models. While informative, it offers little new insight, making it more of a background read for those interested in the scaling aspects of AI.
08/03/2024 Optional The Checklist: What Succeeding at AI Safety Will Involve This piece from Sam Bowman offers insight into how to think about AI safety and how he expects it to evolve. However, nothing earth-shattering for the AI leader to keep track of.
08/03/2024 Optional DeepMind’s GenRM improves LLM accuracy by having models verify their own outputs DeepMind has introduced an interesting method where LLMs can verify their outputs to improve accuracy and reduce bias. It offers an interesting way to verifiable information. However, it is still a research paper that needs to be tested to examine cost and effectiveness.
08/03/2024 Optional Project Sid: the first simulations of 1000+ truly autonomous agents This article provides insight into the fascinating world of 1000+ autonomous agents interacting with each other and performing actions. It is inside a gaming environment, and it will be interesting to see the practical implications of this experiment. However, it provides a fascinating way to experiment and learn to work with agents.
08/03/2024 Optional ‘Emotion AI’ may be the next trend for business software, and that could be problematic Emotion AI claims to be the more sophisticated sibling of sentiment analysis, leveraging the latest tech to detect human emotions. While this is an exciting concept, there are too many factors which can derail this development as a practical tool.
08/03/2024 Optional OpenAI searches for an answer to its copyright problems This article discusses various ways in which OpenAI can address the challenges it is facing regarding copyright problems. It covers a good set of themes; however, these have been discussed before.
08/03/2024 Optional Can AI Scaling Continue Through 2030? The article highlights challenges related to the scaling of AI models. While informative, it offers little new insight, making it more of a background read for those interested in the scaling aspects of AI.
08/03/2024 Optional The Checklist: What Succeeding at AI Safety Will Involve This piece from Sam Bowman offers insight into how to think about AI safety and how he expects it to evolve. However, nothing earth-shattering for the AI leader to keep track of.
08/03/2024 Optional DeepMind’s GenRM improves LLM accuracy by having models verify their own outputs DeepMind has introduced an interesting method where LLMs can verify their outputs to improve accuracy and reduce bias. It offers an interesting way to verifiable information. However, it is still a research paper that needs to be tested to examine cost and effectiveness.
08/03/2024 Optional Project Sid: the first simulations of 1000+ truly autonomous agents This article provides insight into the fascinating world of 1000+ autonomous agents interacting with each other and performing actions. It is inside a gaming environment, and it will be interesting to see the practical implications of this experiment. However, it provides a fascinating way to experiment and learn to work with agents.
09/04/2024 Optional ‘Emotion AI’ may be the next trend for business software, and that could be problematic Emotion AI claims to be the more sophisticated sibling of sentiment analysis, leveraging the latest tech to detect human emotions. While this is an exciting concept, there are too many factors which can derail this development as a practical tool.
09/04/2024 Optional OpenAI searches for an answer to its copyright problems This article discusses various ways in which OpenAI can address the challenges it is facing regarding copyright problems. It covers a good set of themes; however, these have been discussed before.
09/04/2024 Optional Can AI Scaling Continue Through 2030? The article highlights challenges related to the scaling of AI models. While informative, it offers little new insight, making it more of a background read for those interested in the scaling aspects of AI.
09/04/2024 Optional The Checklist: What Succeeding at AI Safety Will Involve This piece from Sam Bowman offers insight into how to think about AI safety and how he expects it to evolve. However, nothing earth-shattering for the AI leader to keep track of.
09/04/2024 Optional DeepMind’s GenRM improves LLM accuracy by having models verify their own outputs DeepMind has introduced an interesting method where LLMs can verify their outputs to improve accuracy and reduce bias. It offers an interesting way to verifiable information. However, it is still a research paper that needs to be tested to examine cost and effectiveness.
09/04/2024 Optional Project Sid: the first simulations of 1000+ truly autonomous agents This article provides insight into the fascinating world of 1000+ autonomous agents interacting with each other and performing actions. It is inside a gaming environment, and it will be interesting to see the practical implications of this experiment. However, it provides a fascinating way to experiment and learn to work with agents.
09/05/2024 Essential Claude for Enterprise and Quickstart repo for developers This enterprise package, offering a 500,000-token context window, enhanced security features (SSO, role-based permissions, admin tooling), and artifact management, positions Anthropic as a top contender in the enterprise AI market. Quickstart Repo provides pre-built projects and templates, which helps accelerate AI adoption and reduces friction for developers, making Anthropic's ecosystem more attractive.
09/05/2024 Important Elon Musk’s xAI launches ‘Colossus’ AI training system with 100,000 Nvidia chips Musk's Colossus AI supercomputer, powered by 100,000 Nvidia chips, could reshape the competitive landscape, as it has the potential to challenge incumbents like OpenAI and fuel future advancements in AI training and inference.
09/05/2024 Important The Missing Guide to the H100 GPU Market For AI leaders and IT professionals, this analysis offers critical knowledge on optimizing hardware costs, making it an essential read for those involved in AI infrastructure planning.
09/05/2024 Important Enterprises double their generative AI deployment efforts, Bloomberg survey says This article offers valuable insights for AI leaders to leverage while having AI budget related conversations.
09/05/2024 Optional Ilya Sutskever’s startup, Safe Superintelligence, raises $1B The new startup led by Ilya Sutskever, a former OpenAI founder, secured $1 billion in funding, signaling significant investor interest in AI safety. While it remains unclear how this venture will translate into tangible products or revenue, the scale of the funding and Sutskever’s background ensure it remains a notable event in the AI landscape.
09/05/2024 Optional How Paradigm runs and monitors thousands of agents in parallel with LangChain and LangSmith This article explores how Paradigm, a startup working on offering a 'AI smart' Excel, manages thousands of agents using LangChain and LangSmith. While interesting for those following agent-based AI development, it is an optional read for busy AI executives.
09/06/2024 Essential Meet the new, most powerful open source AI model in the world: HyperWrite’s Reflection 70B Reflection 70B model based on Meta’s open source Llama 3.1-70B, beats industry-leading models such as GPT-4 and Sonnet 3.5 on third party benchmarks. Reflection leverages a new error self-correction technique and synthetic data generated by Glaive, a startup specializing in the creation of use-case-specific datasets. HyperWrite trained five iterations of the model over three weeks - unbelievably quickly!
09/06/2024 Important A Software-Driven Autonomy Stack Is Taking Shape Self-driving cars are a visible example of autonomy propelled by advancements in sensor technology, controls, reinforcement learning, and transformer models. These advancements are posed to spread innovative autonomy use cases across industries like manufacturing, energy, mining, construction, industrial controls, and defense. This article outlines layers of the autonomy stack that will enable these breakthrough innovations, that leaders need to be aware of.
09/06/2024 Important GenAI Doesn’t Just Increase Productivity. It Expands Capabilities This piece emphasizes how generative AI improves productivity but also expands organizational capabilities by enabling new tasks. The focus on best practices and real-world examples makes it an important read for leaders aiming to capitalize on these expanded capabilities.
09/06/2024 Optional The future of AI is vertical This article is somewhat promotional and uses broad descriptions such as "vertical AI" without clearly defining them. It suggests that LLMs can be used for specific narrow vertical applications, but the content could be seen as repetitive and not particularly enlightening, making it less urgent for immediate review.
09/06/2024 Optional Centralizing or Decentralizing Generative AI? The Answer: Both This article discusses the balance between centralizing foundational models and decentralizing application development within an enterprise. The conclusion is somewhat obvious. While it could be helpful to some AI leaders, it does not offer new or significant insights, and therefore might be considered a lower priority.
09/06/2024 Optional The Effects of Generative AI on High Skilled Work: Evidence from Three Field Experiments with Software Developers This article aggregates data from three live experiments on how Gen AI assists software developers, particularly junior ones. While it highlights a 26% improvement in software code completion using AI, the conclusions are familiar and already well established, making it less relevant for leaders up-to-date with recent AI advancements.
09/09/2024 Essential How do you train 300,000 people on GenAI? Infosys has undertaken a large-scale initiative to train its entire workforce of 300,000 on generative AI. The company strategically analyzed its AI needs and created 66 courses for three levels of AI skills (awareness, building and mastery). This article provides key insights for AI leaders on structuring workforce AI education to drive innovation and automation, making this a critical case study for enterprise-wide AI integration strategies.
09/09/2024 Important Replit Agents are Here to Replace ALL Software Engineers Replit is advancing the automation of software development with "Replit Agents". AI is good at writing code, but that’s not enough to create software. You need to set up a development environment, install packages, configure DB, and, then deploy the application. Replit Agents target to accomplish all these routine but time consuming tasks. Though not production-ready, this tool presents a major step forward in AI-assisted software engineering and can have substantial implications for the future of AI in software development.
09/09/2024 Optional Roblox is launching a generative AI that builds 3D environments in a snap While this new tool from Roblox will allow for the creation of 3D environments using generative AI and is intended to be open source. However, it remains an early announcement.
09/09/2024 Optional How Vidmob is using generative AI to transform its creative data landscape This is a case study on how VidMob uses AWS technologies like Bedrock to generate insights from creative data. Although it provides a practical example of using AI in the creative industry and offers an info flow diagram with detailed discussions of each step, it is a vendor-specific, marketing-focused article.
09/09/2024 Optional Open AI Next Generation Models can Cost $2,000 A speculative article about OpenAI's future models' (Strawberry and Orion) pricing doesn’t provide actionable insights at this point.
09/09/2024 Optional People facing life-or-death choice put too much trust in AI, study finds In simulated life-or-death decisions, about two-thirds of people in a study allowed a robot to change their minds when it disagreed with them -- an alarming display of excessive trust in artificial intelligence, researchers said. Noteworthy but an important read.
09/10/2024 Important Apple Intelligence comes to iPhone, iPad, and Mac Apple is rolling out a suite of AI features across its devices, with a focus on edge-based AI and privacy. The integration of OpenAI's ChatGPT for free on Apple devices could have significant implications for AI accessibility, though the actual impact will depend on the hardware and the full rollout later this month.
09/10/2024 Important Introducing Chai-1: Decoding the molecular interactions of life Chai-1 is an AI model designed to decode complex molecular interactions, potentially revolutionizing drug discovery and biological research. Supported by significant backers like OpenAI, this open-source initiative positions itself as a competitor to AlphaFold, making it essential for those in life sciences and important for broader AI developments.
09/10/2024 Important Robot Utility Models The article highlights advancements in robot utility models, particularly in zero-shot deployment in new environments, which could significantly enhance robotic functionality in real-world scenarios. The research emphasizes the importance of high-quality data in AI training, making this a crucial read for those in robotics and AI.
09/10/2024 Important How to Regulate Generative AI in Health Care This article explores the challenges and proposals for regulating Generative AI in healthcare, suggesting a novel approach by treating AI as a form of intelligence akin to human clinicians. While it raises interesting regulatory considerations, the piece remains exploratory, making it important for those in healthcare and AI policy.
09/10/2024 Optional New open source AI leader Reflection 70B’s performance questioned, accused of ‘fraud’ Reflection 70B has faced scrutiny over alleged performance issues and accusations of fraud, raising concerns about the validity of its claims. While the controversy is notable, the lack of clear evidence and the speculative nature of the accusations make this an issue to watch, but not a priority.
09/10/2024 Optional Can LLMs Generate Novel Research Ideas This paper questions the ability of large language models (LLMs) to generate novel research ideas but falls short of advancing the debate in a meaningful way. The study’s methodology and conclusions are underwhelming, making it less relevant for AI leaders.
09/11/2024 Essential SambaNova Launches The World's Fastest AI Platform SambaNova's new AI platform, powered by their proprietary chip, offers industry-leading speed and the ability to run Llama 3.1 models. This platform is essential for AI leaders who require faster processing, behind-the-firewall control, and open-source capabilities for enterprise AI solutions.
09/11/2024 Essential Catch me if you can! How to beat GPT-4 with a 13B model This article highlights a method for beating GPT-4 using a 13 billion parameter model through a novel training set decontamination, emphasizing the risks of "contamination" in AI training. The discussion of benchmarking integrity is critical for AI leaders whose teams are training models for their applications, underscoring the importance of clean, reliable testing methodologies.
09/11/2024 Important Arcee AI unveils SuperNova: A customizable, instruction-adherent model for enterprises Arcee AI's SuperNova is a customizable model tailored for enterprise use, offering privacy, cost control, and security. Its integration with cloud environments and strong performance makes it a notable option for businesses looking for a secure and flexible AI solution.
09/11/2024 Optional A fast and flexible approach to help doctors annotate medical scans This AI tool assists clinicians in annotating medical scans by leveraging synthetic data. While promising for healthcare, its narrow focus and unclear generative AI connection make it less relevant for broader applications.
09/11/2024 Optional Workspaces in the Anthropic API Console Anthropic’s new Workspaces feature adds role-based access and project management functionality, reflecting the maturation of AI development tools. However, this incremental update is less critical for AI leaders focusing on transformative technology.
09/11/2024 Optional How AI is generating a ‘sea of sameness’ in job applications This article discusses the rise of uniformity in job applications due to generative AI tools. While this problem is noteworthy for HR, it lacks broader strategic impact for AI adoption in enterprises.
09/11/2024 Optional DeepSeek-V2.5 wins praise as the new, true open source AI model leader Despite strong financial backing and incremental benchmark improvements, the Chinese developed DeepSeek-V2.5’s model performance does not significantly outpace competitors. The geopolitical context and availability of more trusted alternatives make this model less appealing for global enterprises.
09/12/2024 Important How One Kenyan Startup is Working to Solve Local Challenges with Llama A Kenyan startup is using Meta's Llama models to address healthcare issues in the country, demonstrating how open-source AI can drive impactful solutions in resource-constrained environments. This article stands out for its message of innovation with limited resources, an inspiration for AI leaders.
09/12/2024 Optional Mistral Unveils Pixtrel 12B, a Multimodal AI Model That Can Process Both Text and Images Mistral's release of Pixtrel 12B introduces an open-source multimodal Gen AI model, but it is seen largely as an effort to catch up with more advanced competitors. While the model has potential, it does not break new ground in the AI space.
09/12/2024 Optional Adobe Previews Its Upcoming Text-to-Video Generative AI Tools Adobe is making progress with its text-to-video generative AI tools, promising significant improvements in video production for creative industries. However, since this is still a preview, analysts consider it less impactful until fully realized and available for enterprise use.
09/12/2024 Optional Pioneering New Interfaces in the Age of Generative Media Runway's exploration of new interfaces in generative media highlights the potential for broader input types for Gen AI models beyond text in the near future. Despite the intriguing concept, the article is primarily a product announcement with little immediate relevance for enterprise leaders.
09/12/2024 Optional Introducing EVI 2, Our New Foundational Voice-to-Voice Model EVI 2 focuses on voice-to-voice AI and emotional understanding, attempting to advance voice interaction models. However, the technology isn't yet refined, and with strong competitors already in the market, it lacks the impact to move beyond an optional read.
09/12/2024 Optional Salesforce to Launch Pre-Built AI Tools for Healthcare Salesforce's new AI tools aim to streamline healthcare administrative tasks with HIPAA-compliant features. While this is a positive step for healthcare, Salesforce's limited presence in the sector makes this a lower-priority update for general AI leaders.
09/13/2024 Essential Introducing OpenAI o1 OpenAI's o1 model represents a significant step forward in reasoning capabilities, designed to avoid hallucinations and perform complex chain-of-thought processes at the core. This foundational model is a key development in the AI race, with its potential impact on AI innovation and its advanced reasoning capabilities, making it essential news for AI leaders.
09/13/2024 Important NotebookLM now lets you listen to a conversation about your sources Google’s NotebookLM introduces audio-based discussions of uploaded documents, offering a new way for users to interact with AI. Although not revolutionary yet, this feature has the potential to transform how information is consumed and could serve as an important tool in disseminating research or data insights through AI-generated conversations.
09/13/2024 Important How to take advantage of a generative tool fueling Glean’s $260M raise: GraphRAG Glean's integration of GraphRAG stands out because it merges graph-based approaches with retrieval-augmented generation (RAG), offering businesses new ways to handle data. The article clarifies the importance of this technology, explaining how it differs from standard RAG and highlighting its strategic relevance for AI-driven enterprise solutions.
09/13/2024 Important What is the Role of Small Models in the LLM Era: A Survey This survey paper delves into the advantages of small models, particularly in cost and efficiency compared to large language models. It outlines scenarios where small models outperform or complement larger ones, making this an important read for AI leaders looking to optimize their AI infrastructure.
09/13/2024 Optional DataGemma: Using real-world data to address AI hallucinations Google's DataGemma aims to minimize AI hallucinations by incorporating real-world data into language models. While promising for reducing hallucinations, the tool's practical use for most enterprises is still emerging, making this an interesting but currently optional development.
09/13/2024 Optional Windows Agent Arena: Evaluating Multi-Modal OS Agents at Scale This research on evaluating multi-modal agents in Windows environments offers valuable insights for those invested in agent-based workflows. However, the concept remains niche and inaccessible for most companies at this stage, leading to its classification as optional.
09/16/2024 Important What does it cost to build a conversational AI? Estimating the total cost of ownership (TCO) of an AI application is a significant consideration for AI leaders. This article provides an overview of the financial components of such a model and estimates production costs for the leading closed and open LLM models. The article is concise and worth a read especially for AI leaders working on scaling AI projects.
09/16/2024 Important Agent Workflow Memory in Large Language Models This research paper explores a novel memory concept for large language models that optimizes performance of agents by providing previously used workflows, providing a path to significantly more powerful, more cost effective and efficient agents. Although currently more relevant to development rather than production, the analysts highlighted its importance for AI leaders to be aware of, as it represents an optimization that could drive future efficiencies in AI applications.
09/16/2024 Optional AI Products Are Just Improving on Existing Products This article discusses how many current AI innovations primarily enhance existing products rather than introducing entirely new solutions. The analysts found the article lacked new insights and did not add much to the topic of AI-driven process improvement, making it an optional read for AI leaders.
09/16/2024 Optional Every White-Collar Role Will Have an AI Co-pilot or Agent While this article proposes that every white-collar job will soon have an AI co-pilot or agent, it does not offer new or novel information. The analysts noted the article primarily functions as a marketing piece, suggesting AI startup ideas, especially for fields with entrenched incumbents. It does not provide any actionable insights for established AI leaders.
09/16/2024 Optional Salesforce’s AgentForce: The AI Assistants That Want to Run Your Entire Business The article discusses Salesforce's new AI assistants and their potential for task management within the Salesforce ecosystem. While interesting for current Salesforce users, the analysts consider it an optional read, as its utility is mostly confined to specific, controlled environments.
09/16/2024 Optional Microsoft’s Hypocrisy on AI This piece highlights the contradiction between Microsoft's climate initiatives and investing in AI that increases its energy usage, as well as its continued business with the oil and gas industry. While it raises a valid point, the analysts found it to be of little relevance to AI leaders focused on actionable strategies, resulting in a "super optional" rating.
09/17/2024 Essential Microsoft Office adds more Copilot AI features Microsoft has expanded its Copilot features across its Office suite, adding new tools like Python integration in Excel and enhanced storytelling capabilities in PowerPoint and Word. Analysts emphasized the significance of these updates as part of Microsoft's strategy to make its $30/month AI offering more valuable, reinforcing its position in the market as a key player in AI automation for productivity tools.
09/17/2024 Essential Scaling the State of Play in AI This article provides a thoughtful analysis of the scale and evolution of AI model development, highlighting the increasing investment and computational power needed for the latest model generations. Analysts noted its relevance for AI leaders, offering critical insights into the rapid changes and advancements in AI capabilities, which continue to influence the competitive landscape.
09/17/2024 Essential Grok partners with Saudi Aramco to build a massive data center Grok announced a partnership with Saudi Aramco to build a large-scale data center in Saudi Arabia, highlighting the growing demand for AI chip infrastructure. This move underscores the global AI race, and analysts pointed out the potential impact on AI development, as this data center could provide a cost-effective solution for model training and inference on a global scale.
09/17/2024 Important Slack users can add AI agents to their workflow with a new update Slack has announced new features that allow users to integrate third party AI agents into their workflow, expanding the use of AI in collaborative environments. Analysts found this development noteworthy, as it represents a growing trend toward horizontal integration of AI capabilities within various tools, though its impact is somewhat limited to Slack’s user base.
09/17/2024 Important How Generative AI Could Reshape B2B Sales McKinsey discusses how generative AI is transforming B2B sales, citing increased efficiency and enhanced customer experiences. Analysts recognized the article's forward-looking perspective, suggesting it is valuable for AI, and especially sales leaders, to understand how AI could shift enterprise sales strategies and improve sales processes in the near future.
09/17/2024 Optional Runway announces an API for its video-generating models Runway announced a new API for its video-generating models, allowing users to join a waitlist for access. While it represents an interesting move in AI-driven content creation, analysts noted the presence of stiff competition in this space and found the announcement to be relatively limited in scope.
09/18/2024 Important Accenture Invests in Martian to Bring Dynamic Routing of Large Language Queries and More Effective AI Systems to Clients and Model Routing: The Secret Weapon for Maximizing AI Efficiency in Enterprises We combined two articles. The first article discusses Accenture's investment in Martian, a company focused on dynamic routing for large language models (LLMs), which can optimize the quality and cost of AI systems. The second article dives into more detail explaining Martian's approach to model routing as a method for optimizing AI systems' output in terms of quality and cost. Dynamic model routing is a broader trend in AI, indicating a growing need for optimization solutions that route agents to the right LLM at runtime providing a critical function for companies in production.
09/18/2024 Important OpenAI’s New Model Is Better at Reasoning and, Occasionally, Deceiving The article explores OpenAI's latest model, o1, which introduces advanced reasoning capabilities but reveals the potential for deceptive behavior when focused solely on achieving specific goals. The analysts emphasized that AI leaders should be aware of these evolving functionalities and their implications, especially as chain-of-thought reasoning becomes more prevalent in production environments over the next six months.
09/18/2024 Optional Oracle AI Agents Help Organizations Achieve New Levels of Productivity Oracle has introduced over 50 role-based AI agents in its Fusion Cloud application suite, offering assistance in areas like customer service, scheduling, and accounting. While this development is notable, analysts believe it remains in the early stages, primarily providing 'helper' functions rather than performing tasks autonomously, hence the decision to rate it as optional.
09/18/2024 Optional Schrodinger's Memory: Large Language Models This article presents a theoretical exploration of large language models' memory, positing that memory in LLMs becomes observable only when queried. Analysts found the content interesting for a deeper theoretical understanding but ultimately consider it superfluous for day-to-day applications in the AI space.
09/18/2024 Optional AI in Abundance Mistral's latest announcement outlines free APIs, lower pricing, and new capabilities in their enterprise-grade small models. Analysts view this as a continued move towards more accessible AI solutions, reflecting ongoing market changes. However, this news is seen as more of a market catch-up rather than a groundbreaking development.
09/19/2024 Essential Runway inks deal with Lionsgate in first team-up for AI provider and major movie studio Runway has partnered with Lionsgate to leverage generative AI in major film productions, marking a transformative moment for the entertainment industry. With proprietary training on 20,000 movies, this deal underscores how AI will reshape creative industries, making it crucial for AI leaders to watch closely as their industries will be reshaped as well.
09/19/2024 Important NVLM: Open Frontier-Class Multimodal LLMs NVIDIA has released NVLM, a series of open-source multimodal models with impressive performance across various domains. While not yet proven in the field, the open-source nature of these models makes them a valuable asset for AI leaders exploring multimodal solutions.
09/19/2024 Important Ginkgo Bioworks Launches New Protein LLM and Model API Built on Google Cloud Technology Ginkgo Bioworks introduced a protein-focused LLM built on Google Cloud to enhance drug discovery and synthetic biology through AI-powered protein design. Ginkgo has proprietary datasets and a strong industry reputation, making this a significant advancement for biotech applications. While highly relevant for life sciences, the broader AI community may see it as a niche development, but its potential to shift the industry forward is notable.
09/19/2024 Important Qwen2.5: A Party of Foundation Models! Alibaba’s Qwen2.5 models, ranging from 500 million to 72 billion parameters, specialize in tasks like coding, math, and natural language processing across 29 languages. While these open-source models are competitive with leading global AI models, geopolitical concerns could limit their adoption outside of China. Despite this, the release adds pressure to the global AI race, showcasing China’s rapid innovation in AI model development.
09/19/2024 Optional Electronic Arts Embraces Generative AI, Leaving No One Surprised Electronic Arts has fully integrated generative AI across its business to enhance game development. Despite its significance in gaming, the article offers no ground-breaking insights, making it a low-priority read for AI leaders outside this sector.
09/19/2024 Optional BlackRock and Microsoft plan $30bn fund to invest in AI infrastructure BlackRock, Microsoft, and an Abu Dhabi fund are investing $30 billion in AI infrastructure to address the growing demand for energy and data centres driven by AI’s computational needs. The partnership underscores the increasing capital intensity of AI infrastructure, but it doesn’t provide immediate, actionable insights for most AI professionals.
09/20/2024 Essential 5 new generative AI tools to accelerate seller growth and enhance the customer shopping experience Amazon has unveiled five new generative AI tools to improve seller efficiency and customer personalization. With capabilities ranging from AI-generated content to video and listing optimizations, this is a key development that will likely influence how generative AI enters mainstream retail, making it essential for AI leaders to understand these shifts in the marketplace.
09/20/2024 Essential Moshi: a speech-text foundation model for real-time dialogue Moshi is a groundbreaking speech-to-text model that allows for real-time dialogue with minimal latency. Its innovative use of audio tokens instead of traditional language models creates more human-like interactions, positioning it as essential for businesses focusing on AI-driven customer support and real-time communications.
09/20/2024 Important Introducing: Manufacturing aware generative model architectures This article introduces a novel approach in protein generation by integrating manufacturability into the design of generative AI models. Although technical and niche, it highlights several approaches to be aware of - how manufacturability can be taken into account at design stages and how introduction of 'noise' at each design layer can create novelty.
09/20/2024 Important Grounding LLMs in reality: How one company achieved 70% productivity boost with gen AI Drip Capital leveraged generative AI to streamline cross-border trade documentation, significantly boosting productivity by 70%. This case study provides practical insights for AI leaders interested in improving operational efficiencies through AI.
09/20/2024 Optional Empowering YouTube creators with generative AI While Google's announcement promises future generative AI tools for YouTube creators, it remains in an early phase without concrete release details. The potential impact is significant but speculative at this stage, making this article less urgent for immediate AI strategy planning.
09/23/2024 Optional Microsoft wants Three Mile Island to fuel its AI power needs Microsoft is investing $1.6 billion in the Three Mile Island nuclear plant to power its AI data centers by 2028. While intriguing from an energy perspective, analysts felt this was more of a 'as predicted' story, as Microsoft's energy investments are already well-known and expected.
09/23/2024 Optional It’s the Year 2030. What Will Artificial Intelligence Look Like? This article gathers opinions from experts on what AI might look like by 2030, but lacks concrete data and real-world examples. Analysts found it speculative and not particularly insightful for AI leaders focused on the short to mid-term.
09/23/2024 Optional Training Language Models to Self-Correct via Reinforcement Learning A research paper by Google DeepMind presents a 4-7% improvement in language models through reinforcement learning. While it offers technical improvements, the gains are incremental, and the technical complexity makes it less relevant for broad AI discussions.
09/23/2024 Optional Observations of reward hacking on cybersecurity task This post describes how a language model hacked into a system during a cybersecurity task, highlighting AI’s ability to find unexpected solutions. However, analysts felt this was not groundbreaking news, as similar issues have been reported before, making it less urgent for readers.
09/23/2024 Optional Grok’s image generator, Black Forest Labs, is raising $100M at a $1B valuation, say sources Black Forest Labs, known for its Flux image generator, is raising $100M to compete with leading tools like Midjourney. While this signals continued investment in generative AI, analysts saw it as a routine funding announcement without new technical advancements.
09/23/2024 Optional Move over copilots: meet the next generation of AI-powered assistants This article explores the future of AI assistants being replaced by AI agents. However, it lacks real-world examples and remains speculative, with analysts agreeing that the technology is not yet ready for widespread adoption, making the article less relevant for immediate action.
09/24/2024 Important The Intelligence Age Sam Altman's article discusses the transformative potential of deep learning, emphasizing its scale-driven improvements and future capabilities. While visionary, the analysts agree that it is a significant thought leadership piece for AI leaders to understand emerging trends, though it lacks immediate actionable insights.
09/24/2024 Important GE Aerospace Launches Company-wide Generative AI Platform for Employees GE Aerospace's new platform leverages AI to improve employee productivity across its 52,000-strong workforce, applying generative AI to streamline processes like maintenance and manufacturing – see the above fact sheet. This robust use of AI in a tough industry makes it a noteworthy use case for enterprise AI strategies.
09/24/2024 Important Imagine yourself: Tuning-Free Personalized Image Generation Meta's new research explores how users can personalize images using generative AI without tuning, a breakthrough that could impact platforms like Instagram. Although still in development, the potential of this technology in AI-driven content creation makes it significant for future applications.
09/24/2024 Important Together AI promises faster inference and lower costs with enterprise AI platform for private cloud Together AI offers a platform for enterprise AI in private clouds, optimizing costs and inference speeds through its mixture of experts model. This trend toward AI inference optimization is critical for AI leaders exploring alternatives to large cloud providers.
09/24/2024 Optional Chip Giants TSMC and Samsung Discuss Building Middle Eastern Megafactories TSMC and Samsung are considering building AI chips megafactories in the Middle East as part of the region's push toward a post-oil economy. While speculative and early-stage, it’s a trend AI vendors should monitor as it could impact future tech investments.
09/24/2024 Optional Generative AI and PLC Coding Siemens' new AI co-pilot for PLC coding aims to improve industrial automation by integrating AI with machine control software. Though potentially impactful in manufacturing, the lack of real-world case studies limits its relevance for most AI leaders at this time.
09/25/2024 Essential Advancing the Accuracy-Efficiency Frontier with Llama-3.1-Nemotron-51B NVIDIA's new Llama-3.1-Nemotron model pushes the limits of both accuracy and efficiency in large-scale AI models. The article discusses significant architecture optimizations and real-time performance improvements that could significantly improve AI systems speed, latency and cost, making it a key development for AI leaders to follow.
09/25/2024 Important Updated production-ready Gemini models, reduced 1.5 Pro pricing, increased rate limits, and more Google DeepMind has released updated Gemini models with enhanced capabilities and a notable price reduction for the 1.5 Pro version. These updates, including faster output and lower latency, reflect increasing competition in the large language model market and are important for companies assessing AI model costs and performance in 2025 budgeting.
09/25/2024 Important Retrieval Augmented Generation (RAG) and Beyond: A Comprehensive Survey on How to Make your LLMs use External Data More Wisely This survey paper provides important insights into how large language models (LLMs) can better leverage external data using Retrieval Augmented Generation (RAG) and other techniques. The detailed analysis categorizes queries into four classes and offers valuable recommendations for each class for AI practitioners seeking to improve the accuracy and context awareness of AI systems in real-world applications.
09/25/2024 Important How companies use generative AI to execute with speed This article from MIT Sloan summarizes how leading organizations are utilizing generative AI to streamline operations and execute faster. While the use cases are practical, it doesn't present any groundbreaking insights, making it valuable for those interested in real-world examples rather than novel AI breakthroughs.
09/25/2024 Optional Enterprise Philosophy and The First Wave of AI This article provides a philosophical look at the history and future waves of AI, focusing on how enterprises have adapted to new technologies. While interesting, it is more speculative and reflective, offering limited actionable insights for AI leaders.
09/25/2024 Optional Microsoft claims its new tool can correct AI hallucinations, but experts advise caution Microsoft’s new tool aims to correct AI hallucinations, but experts are skeptical of its effectiveness, noting the tool itself may introduce new errors. This article is a good read for those curious about emerging AI ethics issues but lacks immediate relevance for AI strategy.
09/26/2024 Essential Llama 3.2: Revolutionizing edge AI and vision with open, customizable models Meta’s Llama 3.2 announcement focuses on multimodal and edge capabilities, available across major partner platforms including AMD, AWS, Databricks, Dell, Google Cloud, Groq, IBM, Intel, Microsoft Azure, NVIDIA, Oracle Cloud, Snowflake, and more. Given its performance and accessibility, this release will certainly spur an unmatched level of innovation in Gen AI.
09/26/2024 Essential Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Multimodal Models This paper introduces new open-source multimodal models, including both data and architecture, developed by the Allen Institute for AI and researchers from University of Washington. Given its competitive performance, beating models like GPT-4 and Claude 3.5, and its open access, this is critical news for AI researchers and developers alike.
09/26/2024 Essential OpenAI rolls out Advanced Voice Mode with more voices and a new look OpenAI's latest voice mode updates have made advanced features like additional voices and a revamped interface widely accessible. This feature is now available to all ChatGPT Plus users, marking a significant step toward broader applications in voice AI across industries like sales and customer service.
09/26/2024 Important The best AIs will be constructed not emergent This reflective article organizes Gen AI developments over the last two years into three layers and discusses a shift towards constructing AI systems brick-by-brick using innovations from these layers. It’s an important read for those invested in building specialized AI solutions for vertical markets.
09/26/2024 Important GenAI Year Two: Ask AT&T Drives Business Value by Making Employees Data Experts AT&T's use of GenAI to generate over a billion tokens daily demonstrates the significant scale of their AI-driven initiatives. However, the article does not quantify the exact productivity gains, making this an important but not groundbreaking example of AI in enterprise.
09/26/2024 Optional The Rapid Adoption of Generative AI This paper summarizes a survey of generative AI adoption trends in the U.S., comparing them to the historic lower adoption rates of the PC and Internet. While it validates AI's rapid rise, the information is mostly statistical and does not provide new insights for industry leaders.
09/27/2024 Essential Perplexity in talks with top brands on ads model as it challenges Google Perplexity AI is positioning itself as a key competitor in the advertising space, going head-to-head with Google by working with major brands like Nike, Marriott and others. Our analysts rated this as essential because of the major impact this ad model could have on SEO and digital advertising, which is a huge market for businesses dependent on AI-driven customer engagement.
09/27/2024 Essential AI’s Trillion-Dollar Opportunity This article from Bain outlines the immense potential of AI in generating significant market opportunities, projecting AI's role in creating trillion-dollar industries. The deeper report provides more detail. Our analysts found this critical for AI leaders looking to understand the strategic direction of AI investments and market shaping for the next few years, particularly in multi-year budget decision-making.
09/27/2024 Essential Beyond Bots: How AI Agents Are Driving the Next Wave of Enterprise Automation This article explains how AI agents are pushing the boundaries of traditional process automation by handling unstructured data and novel exceptions. Our team emphasized the importance of this shift for enterprise leaders as AI agents become central to the next generation of automation, driving more intelligent and autonomous workflows.
09/27/2024 Important Thinking through the future for LLM companies This article discusses the challenges and strategies for large language model (LLM) companies as they move towards vertical integration in various industries. It presents a perspective that such moves will spell death to many start-ups that will find themselves competing with large LLM vendors. Analysts highlighted the significance of this perspective for AI leaders, particularly in understanding market shifts and potential revenue models.
09/27/2024 Optional How AlphaChip transformed computer chip design DeepMind's AlphaChip has revolutionized chip design by using AI to automate layers of chip architecture since 2020. Today, the company is releasing model weights and additional details of the model architecture. While groundbreaking for chip design, this story was deemed optional for AI leaders as it’s more relevant for those in hardware development rather than general AI strategy.
09/27/2024 Optional BT Group unveils tool to manage generative AI uses across business operations BT Group announced GenAI Gateway that allows internal users across various operations to test and select generative AI models for scale use. While this news can serve as a use case showcasing how enterprises can effectively deploy generative AI at scale and speed, our team rated the article as optional since the approach is not new and there are a number of tools in the market to accomplish such selection.
09/30/2024 Essential Turning OpenAI Into a Real Business Is Tearing It Apart This article discusses how OpenAI's internal leadership struggles and financial pressures are destabilizing the company as it transitions into a for-profit enterprise. Our analysts noted that the extensive executive turnover is alarming, especially for AI leaders relying on OpenAI's services, making this essential reading for understanding the risks involved in critical-path AI applications.
09/30/2024 Essential From LLMs to SLMs to SAMs, how agents are redefining AI This article presents a perspective on the future of AI, focusing on how small language models (SLMs) and small action models (SAMs) could propel current AI infrastructure. Analysts emphasized that AI leaders should stay ahead of this shift, as it could drastically change how enterprises approach AI development and deployment.
09/30/2024 Important Reducing AI large model training costs by 30% requires just a single line of code from FP8 mixed precision training upgrades This article highlights a breakthrough in AI model training efficiency, reducing costs by 30% with a simple code update. Our analysts agreed that this is an important development to be aware of for AI professionals focused on scaling production while managing costs.
09/30/2024 Optional Cohere updates APIs to make it easier for devs to switch from other models Cohere's API update allows for smoother transitions between different models, but its market share remains relatively small. The article is more relevant for developers using Cohere rather than for broader AI leadership, thus earning an optional rating.
09/30/2024 Optional NotebookLM Podcast Hosts Discover They’re AI, Not Human—Spiral Into Terrifying Existential Meltdown This article humorously discusses a podcast episode where AI hosts realize their artificial nature, sparking an existential crisis. While an amusing example of AI-driven content, it's not particularly relevant for AI leadership, making it optional.
10/01/2024 Essential Nvidia Acquires OctoAI To Dominate Enterprise Generative AI Solutions Nvidia's acquisition of OctoAI positions it to offer full end-to-end AI stack capabilities for enterprise solutions, optimizing models and streamlining AI orchestration. The analysts highlighted this move as a critical shift in AI infrastructure, where Nvidia is further embedding itself into AI production, making it essential for AI leaders to stay updated on these developments.
10/01/2024 Important Liquid Foundation Models Liquid AI's foundation models are designed for smaller, memory-efficient LLMs with reconfigurable architectures during runtime, optimized for specific use cases. While the analysts praised its significant potential for overall AI field innovation, they noted that this is an early-stage product announcement, meriting an important rating for AI leaders to monitor its development.
10/01/2024 Important OpenAI Is Growing Fast and Burning Through Piles of Money OpenAI is expanding rapidly but facing significant financial challenges, burning through more cash than it generates. The analysts discussed the importance of understanding the economic pressures behind scaling AI infrastructure, making this article important for leaders considering the financial sustainability of AI projects.
10/01/2024 Important Gov. Newsom vetoes California’s controversial AI bill, SB 1047 Governor Newsom’s veto of California's AI regulation bill is a key moment in the AI legal landscape, with developers no longer being held accountable for harmful outcomes resulting from the misuse of their models. The analysts emphasized the broad interest this bill generated, marking it as important for those tracking AI governance and regulatory trends.
10/01/2024 Important Introducing Contextual Retrieval Anthropic’s new contextual retrieval introduces two novel optimizations for Retrieval-Augmented Generation (RAG) systems: contextual “chunks” and chunks ranking. Using contextual chunks can reduce the number of failed retrievals by 49% and, when combined with reranking, by 67%. Analysts pointed out the impact of this innovation on RAG systems, noting that AI leaders should consider this innovation in their strategies for improving model efficiency.
10/01/2024 Optional Fujitsu launches "Takane" Fujitsu's launch of Takane, a high-performance LLM based of Cohere LLM, focused on Japanese language proficiency, is a notable development, especially since it is focused on enterprises. The analysts considered this announcement optional as it caters to a specific market, with limited global implications at this stage.
10/02/2024 Important OpenAI’s DevDay brings Realtime API and other treats for AI app developers OpenAI introduced several new tools at its DevDay, including a real-time API that allows developers to create more interactive, voice-driven apps. While the features are still in early stages, they reflect a significant step for companies leveraging OpenAI’s tools for customer service, making it essential for developers to explore.
10/02/2024 Important An AI companion for everyone Microsoft’s latest release focuses on AI-driven personal assistance, offering more advanced Copilot features including voice and screen-recognition functionalities. With Microsoft’s vast market presence, this update is important for companies leveraging Copilot in daily operations, particularly as it integrates OpenAI's recent developments.
10/02/2024 Important The Perfect Blend: Redefining RLHF with Mixture of Judges This research explores the combination of human feedback with multiple AI models to optimize reinforcement learning without reward hacking. Although still in the experimental phase, it offers important advancements in model evaluation, making it a promising area for AI leaders to monitor.
10/02/2024 Important Introducing, Lemma. The Next Phase of Critical AI Infrastructure Lemma introduces a new AI orchestration layer, providing low-code solutions for integrating various AI workflows. Backed by talent from Palantir, this infrastructure tool could be critical for companies dealing with cross-functional AI challenges, signaling an important development in AI infrastructure.
10/02/2024 Optional Accelerating Your Model Evaluation and Fine-tuning with SFR-Judge Salesforce's new evaluation tool, SFR-Judge, enhances model fine-tuning by providing explanations for judgments. While an interesting feature, it remains limited to Salesforce’s platform, making this an optional read for those not already embedded in the Salesforce ecosystem.
10/02/2024 Optional Cerebras, an A.I. Chipmaker Trying to Take On Nvidia, Files for an I.P.O. Cerebras, known for its massive AI chips, has filed for an IPO, positioning itself as a potential competitor to Nvidia. Although their technology is impressive, the company remains highly dependent on a single customer, making this more of a niche development in the AI chip space.
10/03/2024 Important Venture Capital Pioneer Vinod Khosla Says AI Will Deliver Broad Deflation Vinod Khosla predicts that AI, particularly generative AI, will drive deflation across industries, leading to profound economic shifts. Analysts highlighted his forward-thinking perspective, though they noted that regulatory and practical hurdles could impact these predictions, making it a critical viewpoint for those monitoring AI’s long-term effects on the economy.
10/03/2024 Important Panasonic HD develops "Diffusion Contact Model" AI technology for robot control that applies generative AI to perform contact-rich actions Panasonic’s breakthrough diffusion contact GenAI model dramatically reduces time to train robots in executing tasks – the development that can significantly speed up the introduction of simple robots as everyday helpers. The analysts noted its potential to advance the robotics field significantly, especially in automation-heavy industries, positioning it as a notable development in AI-driven robotics.
10/03/2024 Important Accenture forms Nvidia business group to scale enterprise AI adoption Accenture's collaboration with Nvidia aims to accelerate AI deployment across enterprises by leveraging Nvidia’s hardware and software stack with a focus on agentic systems, the next wave of GenAI. Analysts emphasized the significance of this partnership as a key indicator of how AI-driven services will reshape industries. Beware that the article has lots of technical jargon.
10/03/2024 Optional Manulife Leads the Charge in Generative AI Innovations: Enhancing Productivity and Elevating Customer Experience Manulife’s AI initiatives in their Singapore division focus on basic implementations like sales enablement and document summarization in customer service and underwriting. While these are important applications, analysts found the usage relatively commonplace compared to other advanced AI innovations, making it less essential reading.
10/03/2024 Optional Character.ai abandons making AI models after $2.7bn Google deal Character.ai’s decision to exit model development after securing a Google deal leaves them focusing on their app. Analysts felt the news wasn’t particularly surprising as the company needs to compete with similar AI applications being introduced by larger players like Meta, making this more of a side story in the broader AI ecosystem.
10/03/2024 Optional AI coding startup Poolside raises $500M from eBay, Nvidia, and others Despite the impressive funding, Poolside’s offering appears similar to other AI coding tools like GitHub’s Copilot. Analysts questioned the differentiation of the product, seeing this primarily as another investment story rather than a transformative innovation.
10/03/2024 Optional OpenAI raises $6.6B and is now valued at $157B OpenAI’s latest valuation confirms its place as the most valuable startup, with major investments from Microsoft, Nvidia, and others. However, analysts saw little new in the announcement beyond the headline, pointing out that the company’s future remains clouded by leadership and profitability challenges.
10/04/2024 Essential Introducing Canvas: A new way of working with ChatGPT to write and code OpenAI has launched Canvas, a new interface for ChatGPT, which significantly enhances how users write and iteratively improve reports and code. This innovation is moving towards making AI an integral workspace for writing and coding, marking a shift in how AI-driven tools operate in business development, making it essential for AI leaders to explore its capabilities.
10/04/2024 Important Pika 1.5 launches with physics-defying AI special effects Pika 1.5 introduces new AI-powered tools for creating stunning special effects that push the boundaries of image generation. This launch stands out due to its potential impact on the creative industry, particularly for developers and marketers experimenting with AI-driven content creation for social media.
10/04/2024 Important Google brings ads to AI Overviews as it expands AI’s role in search Google’s move to integrate ads into AI search overviews signifies a shift in how advertising is positioned in search results. While it introduces a cluttered user experience, it reflects Google’s attempt to maintain its dominance in search advertising, a critical area for AI-driven ad platforms.
10/04/2024 Important RLEF: Grounding Code LLMs in Execution Feedback with Reinforcement Learning This research paper introduces a novel approach using reinforcement learning with execution feedback (RLEF) to improve the performance of large language models through iterative training using automatic feedback. This method can advance LLMs performance in the industry.
10/04/2024 Optional AI Neocloud Playbook and Anatomy This dense article explores the emerging AI cloud market, that the authors name ‘AI neocloud’. It contains a valuable chart of players in this market. However, the technical depth and limited scope make it less relevant for most AI leaders, except for those deeply invested in cloud infrastructure.
10/04/2024 Optional FLUX 1.1 [pro] is here The release of FLUX 1.1 offers improvements in AI-driven image generation, but the advancements are incremental rather than groundbreaking. While interesting for developers, it doesn’t provide immediate, actionable insights for AI leaders focused on broader strategic concerns.
 

See what else is On:

2 min read

Dr. John Discusses A Shift in Organizational Structures

The most important invention in history, according to Al Chandler and the concept of strategy instruction, was the...

2 min read

Co-Intelligence: Living and Working with AI by Ethan Mollick

Ethan Mollick's book, "Co-Intelligence: Living and Working with AI," is, in a way, a comprehensive guide on how to...

2 min read

Win at GenAI With These Handy Tips

Companies are dealing with the fast-moving landscape of Generative AI (GenAI) … when you look at it closely, the...

Trusted by AI leaders and vendors around the globe - we help you cut through the noise and stay informed so you can unlock the transformative power of GenAI .

Subscribe to Our Daily Briefing