Andrew Curran
@AndrewCurran_
🏰 - I write about AI, mostly. Expect some strange sights.
OpenAI and Oracle entered an agreement to develop 4.5 gigawatts of additional Stargate data center capacity bringing total capacity under development to over 5 gigawatts running over 2 million chips, significantly advancing progress toward their $500 billion commitment announced…
On IMO P6 (without going into too much detail about our setup), the model "knew" it didn't have a correct solution. The model knowing when it didn't know was one of the early signs of life that made us excited about the underlying research direction!
One piece of info that seems important to me in terms of forecasting usefulness of new AI models for mathematics: did the gold-medal-winning models, which did not solve IMO problem 6, submit incorrect answers for it? 🧵
These agentic teams based models like Grok Heavy, the Gemini Deep Think that just won gold, and the next gen from OpenAI are all going to use about fifteen times more tokens than current systems. This is why Pro plans are north of $200. Essentially; Jensen wins again.
Claude Opus, coordinating four instances of Sonnet as a team, used about 15 times more tokens than normal. (90% performance boost) Jensen has mentioned similar numbers on stage recently. GPT-5 is rumored to be agentic teams based. The demand for compute will continue to increase.