Mathematicians say OpenAI doesn't understand its own Navier-Stokes proof
Mathematicians dispute OpenAI's Navier-Stokes Lean proof, saying it differs from the cited approach and agent credit is unclear.
In early September OpenAI presented a 166-page potential Navier-Stokes solution in Lean, claiming it followed the approach of Diego Córdoba and Luis Martínez-Zoroa. Mathematicians including EPFL's Gonzalo Cao-Labora say the resemblance is only superficial, and the Clay Mathematics Institute has been reviewing the proof. NYU's Tristan Buckmaster alleged the work built on research he and Anthropic's Levent Alpöge did with public models; OpenAI denied using his data. Researcher Noam Brown said about 10,000 model instances ran for roughly 88 hours and 130 billion tokens, but he attributes under 10 percent of the result to that multi-agent setup.
- Experts say the proof differs from the cited Córdoba and Martínez-Zoroa approach.
- The 166-page argument is written in Lean and is under CMI review.
- OpenAI denied using Tristan Buckmaster's prior data.
- Noam Brown credits the base model more than the 10,000-agent swarm.
From the original article
2 min read at the-decoder.com
OpenAI submitted a 166-page solution to the Navier-Stokes problem but apparently doesn't fully understand its own proof. Several groups of experts dispute the claimed methodology, saying the actual approach differs from what OpenAI described. And because language models are black boxes, where the underlying ideas came from will likely never be clear.
By Manuel Uth · the-decoder.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.