John Schulman leaves Anthropic to join Mira Murati’s startup.

Berachain's mainnet will launch on February 6th, 2025. Welcome to Q5.

What Wintermute does thread It’s been exceptionally “fun” on twitter last few days with some accounts hitting new lows in terms of market structure understanding. I’ve written a few times in the past about what we do and how we prefer to do things, but maybe it’s a good time for a refresher:

Buckle up, Bessent might shake shit up today.

o3-mini is out! smart, fast model. available in ChatGPT and API. it can search the web, and it shows its thinking. available to free-tier users! click the "reason" button. with ChatGPT plus, you can select "o3-mini-high", which thinks harder and gives better answers.

>DeepSeek stated that Huawei has successfully adapted the V3 model using its own chips, providing developers with detailed guidelines on chip utilization. The FT previously reported that Huawei has dispatched engineers to assist clients in migrating from NVIDIA chips to Ascend.

About. "Making AI that is smarter than almost all humans at almost all things will require millions of chips, tens of billions of dollars (at least), and is most likely to happen in 2026-2027. DeepSeek's releases don't change this, because they're roughly on the expected cost reduction curve that has always been factored into these calculations." ==> Breaking this down.. 1) I agree HLAGI will probably happen in the range 2026-2030 ... not far off from Kurzweil's historical 2029 estimate... 2) I agree DeepSeek's achievement is great but not super-shocking and is a more-sudden-than-expected jump along an expected cost reduction curve 3) You may have been surprised this leap in efficiency came from China. I wasn't really. Ofc I lived 10 yrs in China and I have seen the huge effort being put into LLMs there lately, and the vast number of AI PhDs there, etc. And this is hardly the first time China has taken a US invention and increased efficiency and decreased cost. 4) There is no good reason to think HLAGI will cost tens of billions of dollars or more and need that many chips !!!! Sure if we were doing a detailed brain simulation, maybe it comes out that way (given how awkward it is to simulate wet stuff on silicon chips). But that is not the approach hardly anybody is taking to AGI these days. 5) For reasons @GaryMarcus and others have pointed out repeatedly and at length, it seems unlikely that transformer NNs or similar architectures are suited to serve as the central component of HLAGIs. I am more optimistic than Gary that they can automate a lot of the economy, after a great deal of work on vertical market specific integration. But they lack many core capabilities needed for HLAGI including the flexible creation and manipulation of grounded compositional abstractions. 6) If we look beyond the purported scaling laws of transformer NNs and look at the AGI field more broadly, at my own team's OpenCog Hyperon architecture and at the breadth of other AGI ideas presented at our annual AGI research conference (going on since 2006 now), you will find many AGI approaches that hold promise of getting to HLAGI without nearly the processor and cash cost that you suggest. 7) Regarding the political angle you take, it's a bit orthogonal to these technical points, but anyway I'll point out my own view is quite different. I think we have a better chance of getting a beneficial HLAGI if said system emerges from a broad global collaborative community effort, and is deployed on a globally distributed decentralized network, rather than if it is created by one country or company to serve its own narrow aims. With this in mind, I think it's great that multiple countries are hosting serious AI breakthroughs. I also am psyched about the potential of more lightweight LLMs like DeepSeek to run more easily on decentralized networks like the ones we have at the ASI Alliance -- it's not AGI but it's cool stuff and having it smaller and cheaper lets us roll it out more broadly to more people in a more decentralized way. The Singularity will not be Centralized!!!

My thoughts on China, export controls and two possible futures

@avichal Rookie numbers 😎

China's AI wins again Alibaba just secretly dropped Qwen 2.5 Max AI, it takes AI video generation to next level and.. it's free 10 examples: 1. woman is quarrelling with a man on a crowded street

BREAKING: Internal #DeepSeek database publicly exposed 🚨 Wiz Research has discovered "DeepLeak" - a publicly accessible ClickHouse database belonging to DeepSeek, exposing highly sensitive information, including secret keys, plain-text chat messages, backend details, and logs.

I'm from the EU. Silicon Valley says AI innovation is impossible in Europe. Too much red tape. Too little funding. But these 8 European startups are changing the narrative:

DeepSeek's new R1 reasoning model is dragging down the NASDAQ. It dropped 6 days ago but it seems Wall Street is only now digesting what it means. I'm no equity analyst, but a few things I've been thinking about. DeepSeek is a huge deflationary shock to the price of intelligence. R1 is outcompeting OpenAI's O1 model for likely less than 1/20th the cost, and they are doing it with only 32B active parameters (GPT-4 likely used ~220B active parameters according to @SemiAnalysis_). They also fully open sourced all of their models, the distillations, and a comprehensive paper detailing how they did it. Intelligence is now way cheaper than we thought. This is great for all consumers of AI—meaning you and me. So why is the NASDAQ tanking? Remember, the NASDAQ is an index of producers, not consumers. The price of oil plummeting is bad news for oil companies, but it's great for those of us who drive. The fact that NVIDIA and all of the hyperscalers are so overrepresented in the NASDAQ these days means the stock market is structurally long the price of intelligence. So who benefits from this deflationary shock? I think there is one company in particular that is best positioned now. It's now been more than 2 years since the release of ChatGPT, and it's clear that no lab has that much of an edge. It only takes a few months for Google, OpenAI, Anthropic, and now DeepSeek to copy each other and trade spots on the leaderboard. This is partly because these companies all publish research (researchers want glory) and even for stuff that's unpublished, these organizations leak like sieves. Engineers want to know how things work. It's quite literally the most interesting question in the world: what is intelligence made of? Labs are just not able to hide this without military grade secrecy (and none of the best talent wants to work for the military). So we're stuck in this status quo. Everyone is trading places at the top of the leaderboard, nobody has a clear long-term edge, and DeepSeek and Meta are intent on open sourcing their models, which causes closed models to continually depreciate. Even with all this AI spend, there don't seem to be any durable moats. So who does have a structural moat here? Look at OpenAI. Sora is already behind the state of the art on video (Kling and Veo are racing ahead). Dall-E is OK but no longer best in class. They are now betting hard on Operator, which is their agentic model. Operator is supposed to be able to book flights, order food, do agentic stuff for you. But it has significant problems aside from the coherence of the model itself: If you are working directly with one of their partners like Instacart, Operator gets full access. But much of the open web appears to be blocking Operator, and that may get exacerbated if the web is crawling with Operator instances. You also have to keep handing control back and forth to log in and out of services, solve Captchas—it's all quite cumbersome and finnicky. Take Google on the other hand. Gemini is quietly #1 on @lmarena_ai. They are #1 on image generation with Imagen. They are ahead on video with Veo. They aren't doing anything agentic yet—Google is usually the last mover on the sexier stuff—but once they do, they have a huge structural advantage. Google's webcrawler bots already have full license to touch everything on the web. They already have access to your Gmail, calendar, they can easily traverse the web and have cached most of it (DeepResearch shows how easy this is for Google), and they also have the crown jewel of untapped data: Youtube. And, of course, they are uniquely positioned to drive agents directly on Androids. Although Google is spending a ton on compute, and they are still a hyperscaler, Google is net short intelligence. They are a consumer of AI in order to serve their customers. DeepSeek and this intelligence deflation is long term good for Google, as it means their own spend will go down. It's cool to hate on Google these days, but I think Google ends up being the long-term winner here if DeepSeek-R1 spells a secular trend. That said, don't count out OpenAI. They are still the strongest product company, and they've earned trust from consumers and enterprises for always being 3 months ahead of the rest of the market. They basically invented the entire test-time compute paradigm, and o3 is a real breakthrough which has yet to drop. If intelligence is the most valuable resource in the world, being 3 months ahead of the competition is enough to earn themselves a big premium, and huge enduring trust from their customers. So yes, the biggest loser here is NVIDIA. If China is a real player (and NVIDIA is not allowed to export to China), and DeepSeek is massively deflating the price of intelligence, and they were able to do all of this on nerfed H800 chips, then NVIDIA is in trouble. You want to be in the game of selling intelligence. NVIDIA is in the game of selling FLOPs. If the ratio of FLOPs to intelligence goes down, down goes NVIDIA stock. So it goes. And of course, we have to say it: congrats to @deepseek_ai team in wiping out a trillion dollars of equity value from the NASDAQ. That's six OpenAIs in a single day, vaporized. Not bad. 👍