This Week’s Covers
This week’s cover is inspired by the release of Meta’s open model LLaMA 3.1, 405b, which is considered to be as good as GPT-4 and available for free to anyone, anywhere. There is a lot of tension around the ethics of releasing powerful AI to the world, so the image depicts a llama casually sauntering out of a top secret security bunker. a top secret bunker entrance, a llama walks into the fresh air –chaos 20 –ar 4:3 –style raw –personalize ytt1577 –v 6.1 The image was upscaled with Magnific.ai, and Meta’s logo was added to the concrete with Photoshop. Font is Stencil to give a feeling of military security.
The category covers stick with the llama theme. I always try to find a template to test MidJourney’s single-shot abilities to give a good image in one try. This week’s template is a llama standing next to a secret agent. large text label reads “Agents” –chaos 20 –ar 4:3 –style raw –personalize ytt1577 –v 6.1. The personalization and chaos values (and possibly the subject) resulted in a lot more illustrations that I usually see. But it was fun, and I’d give it a solid B.
Executive Summary: Week Ending 07/26/2024
This week, we are entering what may be the most transformative six months since ChatGPT was launched. For better or worse.
First, it’s important to know that there are two types of AI models—open and closed. Open models are free and can be modified, customized, and run privately, without guardrails, by anyone. Closed model engines are corporate secrets and available only through subscriptions like GPT-4.
This week, Meta’s open model LLaMA 3.1, 405b, was released to the world. It’s as good as any closed model. This means that all countries and organizations have access to a free model that closely rivals GPT-4. Notably, until GPT-5 or another frontier model is released, there will be no difference between closed and open models. Everyone has the same power at their fingertips. Good guys and bad guys. That alone is going to have a significant impact on humanity in unpredictable ways.
As a result, the tech world is split into two camps: one believes that open source is the only way forward, while the other is afraid of open source getting into the wrong hands.
Leaders like Sam Altman argue that the U.S. needs to take an active role in AI, making huge investments in infrastructure (underwriting data centers) and national security (protecting data centers).
The debate of AI regulation is driving the tech world into presidential politics more openly than ever.
Everyone’s talking about China. Mark Zuckerberg (open source champion) says we might as well make everything open source since China is stealing our secrets anyway (OpenAI’s Leopold hinted as much). Remarkably, Switzerland has announced that all government software must be open source.
Wherever the next iterations of both GPT and LLaMA arrive, they will be comparable to having a graduate student in any field at your fingertips – for pennies.
On top of this major drama, OpenAI is entering the search engine business taking direct aim at Google.
Here are this week’s executive summaries. It’s a wild week, and I highly recommend reading all of these links, themselves.
Point: Sam Altman Urges U.S. Government to Lead AI with Investments and Global Alliances
In a Washington Post op-ed, Sam Altman, CEO of OpenAI, calls for decisive action to ensure that artificial intelligence is developed under a democratic framework rather than an authoritarian one. He warns that the U.S. must maintain its current lead in AI to prevent authoritarian regimes, like those in Russia and China, from using the technology to consolidate power. Altman outlines a four-point plan, emphasizing the need for robust security, substantial government-subsidized infrastructure investment, coordinated commercial AI exportation strategy, and the creation of global safety standards for AI. He stresses that these steps are crucial not only for advancing U.S. leadership but also for spreading the benefits of AI worldwide while upholding democratic values.
https://www.washingtonpost.com/opinions/2024/07/25/sam-altman-ai-democracy-authoritarianism-future
https://archive.ph/Uaj2V
https://twitter.com/8teAPi/status/1816534103006863680
https://www.theregister.com/2024/07/25/sam_altman_ai_freedom
https://twitter.com/sama/status/1816496304257941959
Counterpoint: Mark Zuckerberg Champions Open Source AI as the Future – Key Quotes
- “Today, several tech companies are developing leading closed models. But open source is quickly closing the gap.”
- “Last year, Llama 2 was only comparable to an older generation of models behind the frontier. This year, Llama 3 is competitive with the most advanced models and leading in some areas. Starting next year, we expect future Llama models to become the most advanced in the industry.”
- “We need to control our own destiny and not get locked into a closed vendor.”
- “Open source will ensure that more people around the world have access to the benefits and opportunities of AI, that power isn’t concentrated in the hands of a small number of companies, and that the technology can be deployed more evenly and safely across society.”
- “I think it will be better to live in a world where AI is widely deployed so that larger actors can check the power of smaller bad actors.”
- “Some people argue that we must close our models to prevent China from gaining access to them, but my view is that this will not work and will only disadvantage the US and its allies. Our adversaries are great at espionage, stealing models that fit on a thumb drive is relatively easy, and most tech companies are far from operating in a way that would make this more difficult.”
https://www.facebook.com/zuck/posts/pfbid0ddkWnE2JhSCRSqw4FSzua3FxaQo5Hvz1VWxaT3HELq36Ju4BJ4GcpTycBRE9tWUHl
https://twitter.com/justjoshinyou13/status/1815839440683540800
https://www.youtube.com/watch?v=Vy3OkbtUa5k
Zuckerberg Predicts a Future Dominated by Billions of AI Agents
Mark Zuckerberg envisions a world where billions of AI agents play a central role in daily life, surpassing the number of people on the planet. In a recent interview, he predicted that every business will soon have its own AI agent to interact with customers, similar to how they currently maintain websites and social media profiles. Similarly, people will have agents to help them with tasks and information gathering. Zuck’s vision challenges the concept of a singular, dominant AI system, advocating instead for a vast ecosystem of diverse AI agents.
https://twitter.com/rowancheung/status/1816247877515042898
https://www.youtube.com/watch?v=Vy3OkbtUa5k
Marc Andreessen and Bill Gates shared similar views on the topic last year:
#386 – Marc Andreessen: Future of the Internet, Technology, and AI
(46:46) – Future of browsers
(53:09) – History of browsers
https://lexfridman.com/marc-andreessen/
In the November 2023 edition of “Gates Notes,” Bill Gates wrote “You won’t have different apps for different tasks. You’ll simply tell your device, in everyday language, what you want to do… Agents are not only going to change how everyone interacts with computers. They’re going to upend the software industry.”
https://www.gatesnotes.com/AI-agents
Note also this comment by Wharton Professor Ethan Mollick: “I have said it before, but marketers need to think about how to advertise directly TO the LLMs. This isn’t SEO, it is something different. Especially as agents with autonomy start to appear”
https://twitter.com/emollick/status/1816662088355283133
OpenAI’s SearchGPT Targets Google’s Search Dominance with Publisher-Backed AI Tool
OpenAI has launched SearchGPT, an AI search prototype that combines real-time web data with advanced conversational AI to deliver quick, accurate answers. Rolled out to select users and publishers, it offers clear, source-linked responses in a streamlined interface. Partnering with publishers like The Atlantic and News Corp, OpenAI claims to be focused on protecting journalism while enhancing content visibility. SearchGPT represents a pivot toward integrating AI-driven search into ChatGPT, with a strong emphasis on improving content discovery and user engagement.
https://openai.com/index/searchgpt-prototype
https://twitter.com/OpenAI/status/1816536290822881780
https://www.wired.com/story/searchgpt-openai-search-engine-generative-ai
https://www.theverge.com/2024/7/25/24205701/openai-searchgpt-ai-search-engine-google-perplexity-rival
https://twitter.com/bilawalsidhu/status/1816576449224101906
OpenAI Faces Potential $5 Billion Loss Amid Financial Sustainability Concerns
OpenAI, one of the fastest-growing companies in history, may face a $5 billion loss this year and could run out of cash within 12 months, according to a report from The Information. The analysis suggests that the company’s costs for AI training and operations could skyrocket to $7 billion in 2024, with staffing expenses potentially reaching $1.5 billion. Despite its $80 billion valuation, OpenAI’s significant financial challenges raise doubts about its long-term viability. The company’s rapid expansion and the high costs of maintaining cutting-edge AI models are primary drivers behind its financial woes. With concerns mounting, OpenAI may need to secure additional funding soon to sustain its operations.
https://techstartups.com/2024/07/24/openai-bleeding-money-openai-faces-potential-5-billion-loss-this-year-and-may-run-out-of-cash-in-12-months/
https://www.datacenterdynamics.com/en/news/openai-training-and-inference-costs-could-reach-7bn-for-2024-ai-startup-set-to-lose-5bn-report
https://twitter.com/GaryMarcus/status/1816116071226868085
https://twitter.com/aaronpholmes/status/1816102562031927298
https://www.theinformation.com/articles/why-openai-could-lose-5-billion-this-year
Related post: “Nvidia will be selling at least $200B worth of b200s in 2025 Can somebody tell me where the revenue will come from? @_xjdr @teortaxesTex @doomslide” / X
https://twitter.com/angelusm0rt1s/status/1816333279374762358
AI Models Now Rival Grad Students in Skill, Approaching Near-Zero Cost
All the current AI models are performing at a level that could get them into top graduate schools, particularly in the humanities. This includes the compact, open-source Llama 3.1 70B model, which has shown significant improvements over its predecessors. Looking ahead, GPT-5 is expected to produce college-level graduates at a cost that’s practically free, which could be disruptive for many jobs next year. Meanwhile, the combination of Llama’s quality with Groq’s speed delivers instant intelligence— as demonstrated by the Llama 3.1 8B model available on Groq’s platform. <– You have to see it to believe it!
Wharton Professor Ethan Mollick: AI Requires Strategic Planning Now
Mollick argues that while skepticism about AI’s future is valid, organizations are neglecting to plan for the possibility of exponential growth. Insiders at major AI labs, including Microsoft CTO Kevin Scott, believe that AGI could be achieved within the next few years, with timelines as soon as 2027. Current tools like Claude 3.5 and GPT-4 are performing tasks—such as automating entrepreneurial processes, generating educational simulations, and creating sophisticated content—that were previously thought to be beyond AI’s reach. Mollick emphasizes that the world is changing in front of us, and strategic planning must evolve.
https://twitter.com/emollick/status/1815558443823980789
https://www.oneusefulthing.org/p/confronting-impossible-futures
Meta Unleashes LLaMA 3.1, Marking a New Era in Open-Source AI
Meta has released LLaMA 3.1, a groundbreaking frontier-class AI model with 405 billion parameters, surpassing even GPT-4 in many benchmarks. This open-source model, available to the public with permissive licensing, represents a significant shift in AI accessibility. With capabilities on par with top-tier models like Claude 3.5, LLaMA 3.1 allows for extensive customization, fine-tuning, and commercial use, enabling widespread application across industries. While this democratization of advanced AI opens the door for innovation, concerns arise over potential misuse by governments and malicious actors.
https://twitter.com/emollick/status/1815769799554961561
https://twitter.com/MatthewBerman/status/1815540031823757563
https://twitter.com/DrJimFan/status/1815816844877652195
https://twitter.com/karpathy/status/1815842603377779140
https://twitter.com/ylecun/status/1816132491637375449
https://twitter.com/emollick/status/1815461011711029581
Hot On The Heels of LLaMA 3.1, Mistral AI Unveils Flagship Mistral Large 2 Model
Mistral AI has released its latest flagship model, Mistral Large 2, featuring a 123-billion parameter architecture and a 128k context window. The model supports 11 languages, including French, German, and Chinese, and more than 80 coding languages. Notably, Mistral Large 2 rivals Meta’s Llama 3.1, delivering superior or comparable results across benchmarks like HumanEval and MultiPL-E, particularly excelling in code generation and mathematical tasks. It also significantly outperforms Llama 3.1 in multilingual tasks, while closely matching its larger counterparts in overall performance. This positions Mistral Large 2 as a competitive alternative in the large language model landscape.
https://twitter.com/fdaudens/status/1816144474411520317
https://venturebeat.com/ai/mistral-shocks-with-new-open-model-mistral-large-2-taking-on-llama-3-1/
https://mistral.ai/news/mistral-large-2407
https://twitter.com/GuillaumeLample/status/1816135838448972240
https://twitter.com/GuillaumeLample/status/1816135842764841009
https://twitter.com/GuillaumeLample/status/1816135846254239853
xAI Unveils World’s Most Powerful AI Supercomputer in Memphis
Elon Musk’s AI startup, xAI, has launched what it claims to be the world’s most powerful AI training supercomputer cluster in Memphis, Tennessee. The cluster is powered by 100,000 Nvidia H100 GPUs and is intended to continue to train xAI’s model, Grok. The supercomputer is estimated to have cost between $3 billion and $4 billion, and is a milestone for Memphis, marking the city’s largest capital investment from a new company.
https://www.teslarati.com/elon-musk-xai-supercomputer-cluster-100k-nvidia-h100-gpus/
Elon Musk to Propose $5 Billion Tesla Investment in His Private AI Company, xAI
Elon Musk plans to discuss a $5 billion investment from Tesla into his private AI startup, xAI, despite ongoing legal challenges. Musk, who has consistently stated that Tesla is an AI company, not a car company, founded xAI citing recruitment difficulties within Tesla. This move has sparked a lawsuit accusing Musk and Tesla’s board of breaching fiduciary duties, with allegations that Musk diverted resources meant for Tesla to xAI. The lawsuit seeks damages and a transfer of Musk’s xAI shares to Tesla. Musk intends to bring the investment proposal to a shareholder vote.
https://electrek.co/2024/07/25/elon-musk-will-discuss-tesla-investing-5-billion-private-ai-company/
https://twitter.com/elonmusk/status/1815907844434112999
Amazon Struggles to Monetize Alexa as Losses Mount
Amazon’s strategy to sell Echo speakers at low prices, aiming to generate revenue through Alexa-enabled services, has led to massive losses rather than the anticipated profits. Despite the widespread adoption of Alexa, users primarily utilize the device for basic functions like setting alarms, rather than making purchases. Amazon’s devices business, including Echo, has accumulated over $25 billion in losses between 2017 and 2021. CEO Andy Jassy is now seeking to reverse this trend by launching a paid Alexa service and reevaluating Amazon’s internal metrics that justified these losses, though skepticism remains about the plan’s effectiveness.
https://www.wsj.com/tech/amazon-alexa-devices-echo-losses-strategy-25f2581a
https://archive.ph/5VPB5
Text-to-Video Systems Could Evolve into “World Models”
Runway suggests that any sufficiently advanced text-to-video system might evolve into a “world model,” a concept where the system develops an understanding of the world that goes beyond simple text and video generation. OpenAI has hinted at similar capabilities with its own system, Sora. The emergence of these abilities is also being observed in Kling, a highly advanced text-to-video system from China. If this theory holds, the implications for AI could be significant, as these systems might gain a deeper, more intuitive grasp of the world they depict.
https://twitter.com/emollick/status/1816331031349190672
California’s AI Regulation Bill Sparks Controversy Amid Tech Industry Pushback
California State Senator Scott Wiener is at the center of the AI regulation debate with his “Safe and Secure Innovation for Frontier Artificial Intelligence Models” bill (SB 1047). The bill mandates safety testing for AI models costing over $100 million and requires the ability to shut down these models in case of safety concerns. Notably, Meta’s Llama 3.1 was released this week, is an open-source model, and exceeds this cost threshold While tech giants like Andreessen-Horowitz and Y Combinator have criticized the bill for potentially stifling innovation, Wiener argues that it is a necessary step to ensure responsible AI development without imposing overly stringent controls. He emphasizes that the bill does not require licensing or strict liability but aims to mitigate catastrophic risks. The bill has passed California’s state assembly and now awaits Governor Gavin Newsom’s approval. Despite opposition from Silicon Valley, the legislation enjoys broad support across California, reflecting a cautious but optimistic approach to balancing AI innovation with public safety. A touchy subject, to say the least.
https://www.vox.com/future-perfect/361562/california-ai-bill-scott-wiener-sb-1047
https://www.washingtonpost.com/technology/2024/07/19/biden-trump-ai-regulations-tech-industry
AI Visuals and Charts
“A picture worth billions of parameters! How open-weight models are closing the gap with closed-source ones. h/t @maximelabonne llama
Top 97 Links of The Week Organized By Category
Agents and Copilots
“Imagine an AI application that can type anywhere you can and use the full context of what’s on your screen. This is the application we all deserve (at least if you have macOS.) Check out @OmnipilotAI. It’s an app that works with every other macOS application and uses Claude
AssistantBench: Can Web Agents Solve Realistic and Time-Consuming Tasks?
Friend – An open source AI necklace
“AI assistants have been improving but they still can’t answer complex but natural questions like “Which restaurants near me have vegan and gluten-free entrées for under $25?” Today we’re launching a new benchmark to evaluate this ability. I hope this leads to better assistants!” / X
“Can AI agents solve realistic, time-consuming web tasks such as “Which gyms near me have fitness classes on the weekend, before 7AM?” We introduce AssistantBench, a benchmark with 214 such tasks. Our new GPT-4 based agent gets just 25% accuracy!”
Anthropic
Anthropic’s crawler is ignoring websites’ anti-AI scraping policies – The Verge
Apple AI
Apple shows off open AI prowess: new models outperform Mistral and Hugging Face offerings | VentureBeat
Apple’s first foldable iPhone could arrive in 2026 – The Verge
“Apple is accelerating its move away from reliance on Qualcomm / Apple正加速擺脫對Qualcomm的依賴” / X
Artificial General Intelligence (AGI)
“LLMs are alien beasts. It is deeply troubling that our frontier models can both achieve silver medal in Math Olympiad but also fail to answer “which number is bigger, 9.11 or 9.9”? The latter question broke the internet recently because none of GPT-4o, Claude-3.5 or Llama-3 could
Audio
Udio – Introducing v1.5
You can now split fully-mixed Udio tracks into four separate stems: Vocals, Bass, Drums, and everything else. This enables advanced users to remix their tracks with external tools, or use elements from Udio songs in their music.
“Available now: Stems 🥳 Pro & Premier users can now separate the vocals and instrumentals from songs, which will give you more control over how to use Suno. 1. Go to Library or Create, and click the vertical “…” on a song row 2. Click “Get Stems” 3. If you’re not there
“Added audio 🤯 You can listen to 73,500+ book summaries in 8 to 12 minutes each Atomic Habits:
Daily Bots Demo – Fun demo of a variety of characters to talk to
ElevenLabs
ElevenLabs: Introducing our new Turbo 2.5 model – Hindi, French, Spanish, Mandarin and 27 other languages just got 3x faster. This unlocks high-quality low-latency conversational AI for nearly 80% of the world. For the first time, we support Vietnamese, Hungarian and Norwegian text to speech. And English is now 25% faster compared to Turbo v2.
Augmented and Virtual Reality (AR/VR)
How AI Brought 11,000 College Football Players to Digital Life in Three Months – WSJ
“There have been rapid advancements in AI text-to-video in the past couple of weeks, with multiple players now making pretty impressive systems. This is “giant mech walking through a flowing river” (I selected the best of two videos) from Kling, Runway and Luma.
Business and Enterprise AI
Bhutan’s first AI startup is seven college kids in a dorm – Rest of World
“Your 9-to-5 job is dying. By 2034, it’ll be extinct. That’s Reid Hoffman’s latest prediction – the founder of LinkedIn who predicted the rise of social media in 1997. Here’s what he said next:
“Such a great story on @restofworld: Bhutan’s AI pioneers defy odds! 🇧🇹💻 NoMindBhutan, the country’s 1st AI startup, run by 7 college students: – Chatbots national bank & airline – Thriving despite no access to major cloud servers or international payments platforms” / X
FTC is investigating how companies are using AI to base pricing on consumer behavior | TechCrunch
Ethics/Legal/Security AI
The AI job interviewer will see you now – Rest of World
“the openresearch team releases the first result from their UBI study:
An artist combines AI and unsecured webcams to make mischief – The Verge
Video game performers to strike over AI concerns | AP News
“We’re excited to join forces with @FlyAerodome, the leader in Drone-As-First-Responder technology, in a strategic partnership that will integrate Aerodome’s DFR solution with Flock’s crime-solving platform. Explore the possibilities of Flock + Aerodome:
Ukraine rushes to create AI-enabled war drones | Reuters
Google AI
Reddit now blocks all search engines other than Google
Google is updating the Play Store with AI-powered app reviews and curated spaces
Imagery
Adobe
Adobe – Adobe Unveils Powerful New Innovations in Illustrator and Photoshop Unlocking New Design Possibilities for Creative Pros
Adobe rolls out new generative AI features to Illustrator and Photoshop – The Verge
MidJourney
“this sref is a movie –sref 2611855837
International AI
Inside the United Nations’ AI policy grab – POLITICO
Meta AI
““The AI race is quickly changing. Focus is shifting away from the models themselves to the products they power, as evidenced by the events of this week.” – @alexeheath
Microsoft AI
“Microsoft researchers just published new research introducing ‘SpreadsheetLLM’ and ‘SheetCompressor’ It encodes spreadsheet data into a format that can be used by LLMs to process and understand spreadsheets better
Multimodality
OpenAI
OpenAI is rolling out voice capabilities soon. – The Verge
OpenAI
OpenAI Has Talked to Broadcom About Developing New AI Chip — The Information
“OpenAI asks New York Times to disclose reporters’ notes in ‘vindictive’ legal move
US lawmakers send a letter to OpenAI requesting government access
Open Source AI
Hugging Face
“The @huggingface Hub serves over 6 petabytes and nearly 1 billion requests daily! And AI is just getting started.🚀 Huge shoutout to our amazing infra and Hub team! 👏
Meta/Llama
Meta attacks OpenAI’s business model as the AI race shifts – The Verge – https://www.theverge.com/2024/7/26/24206274/the-ai-race-big-shift-models-to-products
The Llama 3 Herd of Models | Research – AI at Meta
“Exclusive: Meta just released Llama 3.1 405B — the first-ever open-sourced frontier AI model, beating top closed models like GPT-4o across several benchmarks. I sat down with Mark Zuckerberg, diving into why this marks a major moment in AI history. Timestamps: 00:00 Intro
Meta releases Llama 3.1 open-source AI model to take on OpenAI – The Verge
Llama 3.1
Introducing Llama 3.1: Our most capable models to date
“Meta shakes up the AI space with Llama 3.1! 🦙💥 Key highlights: 1. 🏆 405B model claims to match or beat GPT4 & Claude 3.5 2. 🔓 New license allows using outputs to train other LLMs Tech specs: • 🌍 8 languages supported (English, French, German, Hindi, Italian, Portuguese,” / X
Download Llama
“Starting today, open source is leading the way. Introducing Llama 3.1: Our most capable models yet. Today we’re releasing a collection of new Llama 3.1 models including our long awaited 405B. These models deliver improved reasoning capabilities, a larger 128K token context
“We’ve also updated our license to allow developers to use the outputs from Llama models — including 405B — to improve other models for the first time. We’re excited about how this will enable new advancements in the field through synthetic data generation and model distillation” / X
“Potential benchmark leaks for a new series of Llama 3.1 models, including a 405 bn param version. Unconfirmed, however the 70Bb one matching GPT-4 levels. Specially, at its size of being 6x smaller. Also to note that this is the base model not instruct. And many of these
“Things people are overlooking from Llama 3.1 release – More permissive license: allows training on model outputs – Prompt Guard: first BERT-based model that can classify prompt injection and jailbreaking – Multilingual: German, French, Hindi, Thai, …
“Build your own mixture-of-agents using LlamaIndex! In this video @1littlecoder introduces “mixture of agents” – a novel approach using multiple local language models to potentially outperform single models, even surpassing GPT-4 on some benchmarks. It includes a step-by-step
“Now we know which expert to consult. Macro economics problem? Ask Llama. Micro is more of a Claude or ChatGPT thing.
“Llama 3.1 is eating away at any advantage these closed models have… literally you have access to state of the art model through together api or fireworks for $5/1M. Then you have access to those weights and can fine tune the model and deploy the smaller version yourself… using” / X
“If this image is real comparing Llama-3.1 405/70/8b against gpt4o – we have SOTA Frontier Models available Open Source now:
“Llama-3 400b benchmarks got leaked on Reddit. It looks like it beat GPT-4o, which is fantastic. Not sure if they also beat Sonnet 3.5, but I think they come pretty close We will know more once we run the model on Livebench AI. It would be super cool if an open-source model
“Llama-3 405b is officially dropping tomorrow. Seeing rumors that it’s already out on the Internets. Size: 820 GB.” / X
“Our Llama 3.1 405B is now openly available! After a year of dedicated effort, from project planning to launch reviews, we are thrilled to open-source the Llama 3 herd of models and share our findings through the paper: 🔹Llama 3.1 405B, continuously trained with a 128K context
“Mark is an incredible CEO. What particularly stood out to me from our conversation was how deeply he thinks about open source and how it’ll benefit Meta (and the world) in the long term. A few of my favorite moments/quotes: On the future of AI agents: “I think we’re going to” / X
Alex Cheema – e/acc on X: “2 MacBooks is all you need. Llama 3.1 405B running distributed across 2 MacBooks using @exolabs_ home AI cluster https://t.co/MLm47UR0B7” / X – https://twitter.com/ac_crypto/status/1815969489990869369
“The mix of speed + smarts on display has me shook. Rare AI demo to have this effect 🤯 This is what iteration at the speed of thought looks like. Like, how can this tech not absolutely transform knowledge work?
“Wow. After only one day of Llama 3.1 405b, French startup Mistral AI dropped LARGE 2. It’s ANOTHER open-source flagship AI model that scores close to Llama 3.1 405b and even surpasses it on coding benchmarks while being much smaller at 123b. Benchmarks vs. Llama 3.1 405b: –
Qwen
“Synthetic data can beat its teacher! The AI-MO team released their winning dataset with an additional fine-tuned @Alibaba_Qwen 2 model that approaches or surpasses @OpenAI GPT-4o and @AnthropicAI Claude 3.5 in match competitions. 👀 There was a sentiment that fine-tuned models
Podcasts/YouTube/Op-Eds
Why Big Tech Wants to Make AI Cost Nothing
“To help explain the weirdness of LLM Tokenization I thought it could be amusing to translate every token to a unique emoji. This is a lot closer to truth – each token is basically its own little hieroglyph and the LLM has to learn (from scratch) what it all means based on
Open Source AI Is the Path Forward | Meta
Publishing
Is This The End? Scripps is Looking for Director of Newsroom AI – https://www.adweek.com/tvspy/is-this-the-end-scripps-is-looking-for-director-of-newsroom-ai/
Get paid or sue? How the news business is combating the threat of AI
“I have said it before, but marketers need to think about how to advertise directly TO the LLMs. This isn’t SEO, it is something different. Especially as agents with autonomy start to appear” / X
“The last four customer calls I had all said they heard about Formula Bot from ChatGPT. To be ahead of Microsoft is 🤯
“A lot of models were at least partially trained on GPT-4 outputs (against their policy). It is why so many identify as GPT-4 when pushed and tell the same jokes as GPT-4. A subtle effect of this new policy is the next generations of models will have a more Llama-like personality” / X
Anthropic’s crawler is ignoring websites’ anti-AI scraping policies – The Verge
AI paid for by Ads – the gpt-4o mini inflection point
“Aspen Institute’s wake-up call for journalism: Embrace AI or risk obsolescence 📰🤖 “Every new technology comes with risks—it’s how the media industry responds that determines how (or whether) news providers can prevail.” — @Vivian A must-read for all in media, especially with
The Bookseller – News – Academic authors ‘shocked’ after Taylor & Francis sells access to their research to Microsoft AI
Robotics and Embodiment
Dog-like robot jams home networks and disables devices during police raids — DHS develops NEO robot for walking denial of service attacks | Tom’s Hardware
Science and Medicine
AlphaProof, a New A.I. from Google DeepMind, Scores Big at the International Math Olympiad – The New York Times
AI achieves silver-medal standard solving International Mathematical Olympiad problems – Google DeepMind
Google claims math breakthrough with proof-solving AI models | Ars Technica
“I told you OpenAI solved this [the 9.11 > 9.9 issue] last year… It’s crazy discipline that they kept it under wraps with even most of their employees on the business side not knowing what going on. For all the talk of lousy opsec…”
“If anyone is interested, OpenAI did work similar to the new AlphaProof 2 years ago, at a smaller scale, and has written a paper on it.
“We have just released the ✨NuminaMath datasets: the largest collection of ~1M math competition problem-solution pairs, ranging in difficulty from junior challenge to Math Olympiad preselection. These datasets were used to win the 1st Progress Prize of the AI Math Olympiad and
AI predicts droughts a year in advance | PreventionWeb
“AI System Achieves Silver Medal-level score in IMO The International Mathematical Olympiad (IMO) is the oldest, largest & most prestigious competition for young mathematicians. Every year, countries send their top young mathematicians to take a 6 problem test spanning two days.
Microsoft collaborates with Mass General Brigham and University of Wisconsin–Madison to further advance AI foundation models for medical imaging – Stories
Video
Kling
“”puppies holding a press conference, cat on a giant screen” in Kling, Runway, and Luma (which doesn’t really hit the mark here)
“Just a reminder that it is pretty trivial to generate fake viral videos of anything you want, especially if you don’t care about exact details. Kling: “shaky iPhone footage of a fox on a surfboard in the waves” – actually pretty impressive waves, reflections & consistency.
“Animated with @Kling_ai @KlingAIOfficial just gave it an image and it did a fantastic job on it
“Newspaper Mummy Skateboarding on the Street. Aivideo created by kling. #kling #klingai @Kling_ai #midjourney #AIart #aivideo #sport ⛳️img : midjourney 📽️vid: kling 🎼music: capcut 🎬edit: capcut
KLING AI: Next-Generation AI Creative Studio
Luma
“For my latest short, my main weapon was the new Keyframing ‘Endframe’ feature from @LumaLabsAI You can put 2 images into endframe. And With some prompt guidance, you can get a more accurate scene. Here, I made 2 separate shots of the character transition from happy then sad.
Tech Papers, Training, and Development
“Synthetic data can beat its teacher! The AI-MO team released their winning dataset with an additional fine-tuned @Alibaba_Qwen 2 model that approaches or surpasses @OpenAI GPT-4o and @AnthropicAI Claude 3.5 in match competitions. 👀 There was a sentiment that fine-tuned models
“A new paper suggests too much training on AI-produced content causes AI models to break. This is an ongoing discussion, with lots of research and discussion about when/if synthetic training data works. So a helpful paper, but likely not the final word.
“There is still no benchmark for LLM hallucination rates. Few benchmarks have comparisons to humans There are no common benchmarks that cover use cases in innovation, writing, persuasion, human interaction, education, creativity, etc. Yet LLMs are often built towards benchmarks” / X
How to Create High Quality Synthetic Data for Fine-Tuning LLMs
“Patronus AI announced the release of ‘Lynx’, a new open-source hallucination detection model They claim that it outperforms existing AI models such as GPT-4, Claude-3-Sonnet, and more An important challenge to solve
The Rest: AI News of The Week
Don’t let the volume overwhelm you. Have fun and skim these. The links are organized by topic, sorted from ‘coolest’ to ‘least cool’, and each topic is clearly defined with a headline. I’ve added a description and glossary of what the topics mean, beneath each label, in plain language. I do the work so you don’t have to! When you visit the pages, note that the links and descriptions are often pulled directly from tweets or articles, so it’s not always my voice. Pause when you see something that interests you. Reach out to me any time. I enjoy sharing and discussing these items.
Agency/Agents/Copilots News of the Week: Agency is when AI can do things for you (like Googling an actress name or fetching the latest weather forecast). An agent is one step further, when AI given autonomy to take action on your behalf (“Alexa, book a reservation for three at Peak in Hudson Yards for Friday night”). A co-pilot is an assistant (like spell check or autofill).
This week’s latest agent news: https://ethanbholland.com/2024/07/26/agents-and-copilots-ai-news-week-ending-07-26-2024/
Amazon News of The Week: Individual company products will often be placed in the categories they match (image, audio, agents, robots, etc). Occasionally, I’ll dedicate space to a company’s news if it’s broad or a major product release.
This week’s latest Amazon AI news: https://ethanbholland.com/2024/07/26/amazon-ai-news-week-ending-07-26-2024/
Anthropic News of the Week:
Anthropic is a company that builds LLMs like OpenAI, Mistral, Meta, etc. Their main AI brand is Claude. As with Amazon and Apple, individual Anthropic company posts will often be placed in the categories they match (image, audio, agents, robots, etc). Occasionally, I’ll dedicate space to a company’s news if it’s broad or a major product release.
This week’s Anthropic news: https://ethanbholland.com/2024/07/26/anthropic-news-week-ending-07-26-2024/
Apple News of the Week: As with Amazon, individual Apple company products will often be placed in the categories they match (image, audio, agents, robots, etc). Occasionally, I’ll dedicate space to a company’s news if it’s broad or a major product release.
This weeks’ latest Apple AI news: https://ethanbholland.com/2024/07/26/apple-ai-news-week-ending-07-26-2024/
Artificial General Intelligence (AGI) News of the Week: Artificial General Intelligence, in a nutshell, is when artificial intelligence is able to beat humans at everything (including embodying physical forms and completing physical tasks). It’s usually a thought catalyst for predictions, like when AGI will occur. 10 years? 25 years? 100? AGI is an event horizon that is tough to define, tough to imagine, and tough to predict. OpenAI defined AGI in its charter as “highly autonomous systems that outperform humans at most economically valuable work”. OpenAI has a section of its website dedicated to AGI. Google’s DeepMind published my favorite report on the five levels of artificial intelligence on the way to AGI (see also here).
This week’s latest Artificial General Intelligence (AGI) news: https://ethanbholland.com/2024/07/26/artificial-general-intelligence-agi-news-week-ending-07-26-2024/
AI Audio News of the Week: In this case, AI audio can mean a few things. The first is “generative audio” which refers to creating sounds with AI, much like ChatGPT writes words or MidJourney creates images. For example, asking for the “sound of waves crashing on the beach” would be text to sound. Another example would be an AI ‘watching’ a video and adding sound to it, like a foley artist would add footsteps or a creaking door to a movie scene. Lastly, AI audio can refer to microphones that only pick up certain speaker’s voices or headsets that cancel out all voices but your friends. This week’s latest AI audio news: https://ethanbholland.com/2024/07/26/audio-news-week-ending-07-26-2024/
Autonomous Vehicles/Driverless Cars News of the Week: Driverless car news doesn’t always get its own category, because it’s so close to robot embodiment. I go with my gut each week around what to place in each category. My recommendation would be to follow Robotics/Embodiment also, as the two fields are converging.
This week’s autonomous vehicle news: https://ethanbholland.com/2024/07/26/autonomous-vehicles-news-week-ending-07-26-2024/
Augmented and Virtual Reality (AR/VR) News of the Week: Augmented reality is when you see images or information on top of the real world. A car windshield with a heads-up display of the speed. Or glasses that have facial recognition and overlay the names of everyone in view. Virtual reality is when you are transported into another place, usually wearing goggles, but a flight simulator could also be considered virtual reality.
This week’s latest AR/VR news: https://ethanbholland.com/2024/07/26/augmented-and-virtual-reality-ar-vr-news-week-ending-07-26-2024/
Business/Enterprise News of the Week: This broad category is for stories that impact corporations and large scale AI implementation. Enterprise refers to a type of AI that is often custom built for a business or leverage an API to connect secure data to an AI model.
This week’s latest enterprise AI news: https://ethanbholland.com/2024/07/26/business-and-enterprise-ai-news-week-ending-07-26-2024/
Chips and Hardware AI News of the Week: Most of the chip news is NVIDA usually, yet more and more Meta, Google, and OpenAI are starting toward their own manufacturing. I have to make the call whether to put Meta, Google, and OpenAI’s chip news under this section or their company sections. Lately, I’m putting each company’s chips news into the company category, rather than the chips category. This is the rest of the chips headlines.
This week’s latest chips and hardware news: https://ethanbholland.com/2024/07/26/chips-hardware-and-infrastructure-week-ending-07-26-2024/
Education AI News of the Week: There is a lot of buzz around the impact of AI in education. This section focuses both on the risks and rewards of how AI can impact learning. It’s broader than just K-12 and includes things like skills, trade, professional, and higher education. This is not about how to learn AI, it’s about AI’s impact on learning.
This week’s latest education news: https://ethanbholland.com/2024/07/26/education-ai-news-week-ending-07-26-2024/
Ethics/Legal/Security AI News of the Week: This section focuses on the impact AI is having on ethics (deep fakes, war, trust, false information, plagiarism, job loss, income), legal (rights, laws, regulations), and security (hacking, phishing, national interests, safety). For huge news stories like the NY Times suing OpenAI, I usually put them under the main section or give them their own page.
This week’s latest AI ethics/legal/security news: https://ethanbholland.com/2024/07/26/ethics-legal-security-ai-news-week-ending-07-26-2024/
Google AI News of the Week: Individual company products will often be placed in the categories they match (image, audio, agents, robots, etc). Occasionally, I’ll dedicate space to a company’s news if it’s broad or a major product release.
This week’s latest Google AI news: https://ethanbholland.com/2024/07/26/google-ai-news-week-ending-07-26-2024/
Imagery News of the Week: AI imagery covers “generative AI” image tools. This usually text-to-image, where a user enters a prompt (“a polar bear walking through NYC”) and a tool like Dalle or MidJourney generates an image in the likeness of the description. This is different than AI vision, where an AI “looks at” an image and can derive context, details, and contents. AI vision is a subset of AI called multimodality. Imagery, in this case, is for image creation and modification/editing. Adobe Photoshop’s AI tools would fall into this category. I’ll also include things like automatic masking and object removal, even though that’s in between imagery and vision… but practically speaking it fits into editing.
This week’s latest AI image news: https://ethanbholland.com/2024/07/26/imagery-news-week-ending-07-26-2024/
International AI News of the Week: A lot of international news will get cross listed in the chips, security, or open-source categories, however it’s nice to have a separate category for worldwide AI news.
This week’s latest international AI news: https://ethanbholland.com/2024/07/26/international-ai-news-week-ending-07-26-2024/
Meta AI News of the Week: This is a space dedicated for Meta specific AI advancements and news stories.
This weeks Meta AI news: https://ethanbholland.com/2024/07/26/meta-ai-news-week-ending-07-26-2024/
Microsoft AI News of the Week: This is a space dedicated for Microsoft specific AI advancements and news stories.
This weeks Microsoft AI news: https://ethanbholland.com/2024/07/26/microsoft-ai-news-week-ending-07-26-2024/
Mobile AI News of the Week: In April, 2024 I added a dedicated category for mobile. Prior, I put all most the mobile news into either the company (Apple v. Google v. Microsoft) or locally run AI. It also ended up in the chips and hardware section, or the consumer products category. There is enough mobile news to at least start cross linking it all in one place. This week’s latest mobile AI news: https://ethanbholland.com/2024/07/26/mobile-news-week-ending-07-26-2024/
Multimodal AI News of the Week: This is a broad topic for an single AI model that demonstrates an ability to interact with more than one modality (imagery, video, audio, text). Often multimodal news will end up in one of these categories. I’m playing it by ear on a case by case basis. Please be patient with my organizational challenges.
This week’s multimodal AI news: https://ethanbholland.com/2024/07/26/multimodality-news-week-ending-07-26-2024/
OpenAI: OpenAI is the leading force in the AI boom of 2023 and now 2024. This section focuses on news that is specific to OpenAI. This section will compete with all of the other sections (imagery, vision, ethics, etc) because OpenAI is so broad. I won’t be able to consistently pick when to put things under OpenAI or other sections, so bear with me.
This week’s latest OpenAI news: https://ethanbholland.com/2024/07/26/openai-news-week-ending-07-26-2024/
Open Source Models: An open source AI model refers to a class of artificial intelligence models with public source code. They can be inspected, copied, installed, and customized on private computers. In contrast, a closed source model is proprietary and owned by a company that you pay to use (like PowerPoint or Photoshop). One of the most famous open source language models is a French model called Mistral. Its code is completely publicly available, and anyone can download it and customize it. On one hand, open source is a transparent and powerful way to democratize AI, but on the other hand, open source models circumvent the guard rails and copyright protections that private companies implement. Open source models are the wild west of artificial intelligence, but also the potential saving grace (depending on who you ask). It’s a bit like gun control debates but for computing power.
This week’s latest open source news: https://ethanbholland.com/2024/07/26/open-source-ai-news-week-ending-07-26-2024/
Perplexity News of the Week:
Perplexity is renowned for its advanced search and information retrieval technologies. In 2024, they introduced “Perplexity Pages,” a tool transforming AI-driven research into detailed, shareable web pages. However, in 2024, the company also faced allegations of content theft, with claims that its AI-generated articles improperly replicate work from other sources. This week’s latest Perplexity news: https://ethanbholland.com/2024/07/26/perplexity-news-week-ending-07-26-2024/
Podcast/YouTube Clips of the Week: This is for more general interviews and explainer videos and podcasts that provide access to leadership, demos of new products, and walkthroughs and tutorials. Videos focused on specific topics will live in the topic category (i.e. images), but broader videos will live here.
This week’s latest podcasts and YouTube clips: https://ethanbholland.com/2024/07/26/podcasts-youtube-op-eds-week-ending-07-26-2024/
Publishing AI News of the Week: These are stories about AI’s impact on the publishing industry. From copyright and crawling to the death of page views or even the end of browsers.
This week’s latest publishing AI news: https://ethanbholland.com/2024/07/26/publishing-news-week-ending-07-26-2024/
RAG Retrieval-Augmented Generation News of the Week: RAG allows a language model to “reference an authoritative knowledge base outside of its training data sources before generating a response” (via Amazon). Historically RAG was prone to hallucinations, however new methods are improving the reliability. There is enough news about RAG, that I want to start tracking it separately for my own use.
This week’s latest RAG (Retrieval-Augmented Generation) AI news: https://ethanbholland.com/2024/07/26/rag-retrieval-augmented-generation-news-week-ending-07-26-2024/
Robotics/Embodiment News of the Week: This is the most intense area of AI. Embodiment refers to putting an AI inside of a machine. It’s “embodying” the object and therefore giving a robot agency in the real world. An example would be using a large language model as an interface to a complex coding task. Just as you ask “Alexa, play Bad Blood by Taylor Swift on Spotify” using plain language, with embodiment you could ask a robot to “Go to the laundry basket and bring me all of the red shirts”. The language model in the robot would translate your request into the proper code to go get the red shirts. The robot was never trained on the task. Another type of embodiment would be training a robot using virtual reality simulations. Using an simulation, a robot could be trained on thousands of scenarios until the real world can be swapped out and the robot doesn’t “notice”. This section also includes factory automation and human prosthetics. There will be some overlap with other categories like autonomous vehicles. I first learned about embodiment from Alan Thompson. I highly recommend his video explainer: https://youtu.be/peLqYP9BAUg?si=2FzrvDlw-qaQFaCx.
This week’s latest robot and embodiment AI news: https://ethanbholland.com/2024/07/26/robotics-and-embodiment-news-week-ending-07-26-2024/
Science/Medicine AI News of the Week: AI’s strength is learning patterns. This applies nicely to medical diagnosis and identifying trends. When combined with data and AI vision, this means AI is good at looking at x-rays. Language models are helping with patient interface, and robotics and augmented reality are advancing surgery. Powerful enterprise models like Google’s Alphafold can master protein folding. Other models can read ancient scrolls without opening them.
This week’s latest AI science and medicine news: https://ethanbholland.com/2024/07/26/science-and-medicine-news-week-ending-07-26-2024/
AI Video News of the Week: AI video in this case refers to generative video. Much like imagery meant generative imagery. This usually text-to-video, where a user enters a prompt (“a wizard walking out of a flaming building”) and a tool like Pika or Runway generates an video in the likeness of the description. It also covers animation of still images, where an image is given motion (like a photo of a waterfall appearing to have flowing water). As with images, this is different than AI vision, where an AI “looks at” an image or video and can derive context, details, and contents. Video, in this case, is video creation and modification/editing.
This week’s latest AI video news: https://ethanbholland.com/2024/07/26/video-news-week-ending-07-26-2024/
X/Twitter/Grok: Grok is one of several AI’s developed by X, and it’s a bit blended in with Telsa and other Elon Musk technology. Not every week will have a Grok section, but like Meta, Google, Apple, and OpenAI, X will be in the news enough to have its own section.
This week’s latest X news: https://ethanbholland.com/2024/07/26/twitter-x-grok-week-ending-07-26-2024/
Technical and AI Developer News of the Week: Everything that is too technical for general consumption goes here. These are stories I think are important, but might be inaccessible and confusing. It’s also a space for developer news and deep dives into how AI works, under the hood.
This week’s technical and dev AI news: https://ethanbholland.com/2024/07/26/tech-papers-training-and-development-week-ending-07-26-2024/
Credits/Sources
Most of these weekly links come from just a few prolific oversharing sources. Please follow them, as they work hard to find the news each week and they make it a lot easier for me to compile.
- Robert Scoble: https://x.com/Scobleizer
- Ethan Mollick: https://www.linkedin.com/in/emollick/
- Alan Thompson: https://lifearchitect.ai/
- Theoretically Media: https://www.youtube.com/@TheoreticallyMedia
- The Rundown: https://www.therundown.ai/
- Bilawal Sidhu: https://twitter.com/bilawalsidhu/
- TLDR: https://tldr.tech/ai
- Jeremiah Owyang: https://twitter.com/jowyang
- Nick St. Pierre: https://twitter.com/nickfloats
- Dr. Jim Fan: https://twitter.com/DrJimFan
- All About AI: https://www.youtube.com/@AllAboutAI
- Marshall Kirkpatrick: https://aitimetoimpact.com/
- AI News (Smol Talk): https://buttondown.email/ainews/archive/
- Andrej Karpathy: https://x.com/karpathy
- Brett Adcock: https://x.com/adcock_brett
- Florent Daudens: https://x.com/fdaudens
- Ate-a-Pi: https://x.com/8teAPi
- Francesco Marconi: https://x.com/fpmarconi
For previous issues, please visit the archives!
Thanks for reading!
Leave a Reply