MATHEMATICIANS REJECT OPENAI'S ACCOUNT OF ITS OWN PROOF — 'IT IS A DIFFERENT APPROACH'

Several groups of mathematicians dispute OpenAI's account of its 166-page potential solution to the Navier-Stokes problem, presented in early September. OpenAI said the proof followed the approach of Spanish mathematicians Diego Córdoba and Luis Martínez-Zoroa, but EPFL mathematician Gonzalo Cao-Labora said the similarities are only surface-level and that, in the details, it is a different approach. The proof is written in Lean, which verifies each logical step, and the Clay Mathematics Institute has been reviewing it since its release.

The dispute also covers provenance and scale. Before publication, NYU's Tristan Buckmaster accused OpenAI of building on work he and Anthropic's Levent Alpöge had done with publicly available models, which OpenAI denied, and a definitive answer may be impossible because the models are black boxes. OpenAI also deployed about 10,000 model instances for roughly 88 hours and 130 billion tokens, yet researcher Noam Brown put less than 10 percent of the success down to the multi-agent setup, saying nobody knows what the swarm actually contributed.