OpenAI Claims Navier-Stokes Solution amid Rival Mathematicians' Accusations
OpenAI's claim to have solved a Millennium Prize Problem is shadowed by its own admission that it cannot fully rule out that its agents drew on the unpublished work of rival mathematicians, putting the proof's independence in question.
Reporting from 1 source: GIGAZINE.
OpenAI announced on September 8 that a multi-agent AI system using an internal model more powerful than GPT-6 Astra produced a proof toward solving the Navier-Stokes equation, a Millennium Prize Problem, with Lean formalization complete. The claim is shadowed by a dispute: mathematicians Levent Alpoge of Anthropic and Tristan Buckmaster of NYU say OpenAI learned of their near-solution before sending its own prompts. OpenAI denies using their work but says it cannot fully rule out contamination.
OpenAI's September 8 announcement says the proof came from a multi-agent system whose internal model is "far more powerful than GPT-6 Astra," with Lean formalization complete. The process ran 2.7 million agent messages and output roughly 130 billion tokens. Researcher Noam Brown put the cost in the millions of dollars and predicted that within a year, anyone with a $20 monthly subscription would have access to an AI able to solve problems on par with Navier-Stokes.
The dispute is about timing. Buckmaster's statement says he contacted OpenAI's math researchers on September 3 after hearing Anthropic's progress had reached the company, and that a meeting to avoid competition revealed OpenAI sent its Navier-Stokes prompts only after learning of the pair's work. OpenAI denied referencing their results or accessing specific user data, while conceding it cannot completely rule out that anonymous data from the pair's use of OpenAI products contributed to model improvement.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.