Key facts
- NYU mathematician Tristan Buckmaster claims OpenAI used leaked information to solve a major math problem.
- OpenAI announced it solved the Navier-Stokes existence and smoothness problem in 88 hours using 10,000 AI agents.
- The problem is one of the seven Millennium Prize problems, with a $1 million bounty.
- Buckmaster alleges OpenAI's solution was influenced by his team's progress and that OpenAI began working on the problem after learning of his work.
- OpenAI denies the allegations, stating its solution differs and no user data was accessed.
NYU mathematician Tristan Buckmaster has accused OpenAI of using unethical tactics, including leveraging leaked information, in its pursuit of solving the Navier-Stokes existence and smoothness problem, one of the seven Millennium Prize problems.
OpenAI announced Tuesday that it had solved the notoriously difficult math problem in 88 hours using approximately 10,000 AI agents. The company stated that its new AI model, trained on vast amounts of data, demonstrated a strong aptitude for mathematics, leading researchers to apply it to significant unsolved problems.
Buckmaster, however, claims that information about his team's progress, which included mathematician Levent Alpöge from Anthropic, was shared with OpenAI. He alleges that OpenAI began its work on the Navier-Stokes equations only after learning of his team's efforts, and that the company's solution may have been influenced by their approach. Buckmaster also expressed concern that OpenAI might have used his interactions with its Codex tool to inform their solution, as OpenAI reserves the right to train on user data.
Sebastian Bubeck, who leads OpenAI's mathematical research, has refuted Buckmaster's claims, calling them "false and inflammatory" and asserting that the company adhered to academic standards. Bubeck stated that OpenAI had not seen any of Buckmaster and Alpöge's work until it was publicly released and that no user data was accessed during their problem-solving efforts. He added that while unlikely, de-identified data derived from user interactions might have helped improve their models, but their proofs differ significantly.
The Navier-Stokes problem, central to understanding fluid dynamics and turbulence, carries a $1 million bounty from the Clay Mathematics Institute. Buckmaster stated his decision to go public was to counter potentially misleading announcements from OpenAI, emphasizing that he has not yet read OpenAI's full proof but felt compelled to share his account of the timing and proposals made to him.