Technology
OpenAI's Math Breakthrough Claim Sparks Credit Dispute With Rival Researchers
OpenAI announced it had produced an AI-generated solution to the Navier-Stokes equation, a 200-year-old mathematical problem describing fluid behavior in substances like water and air, and one of the unsolved Clay Millennium Prize problems worth $1 million each. However, the announcement has been overshadowed by accusations from NYU mathematician Tristan Buckmaster, who alleges OpenAI accelerated its efforts after learning about work he was conducting with Anthropic researcher Levent Alpöge, and that OpenAI attempted to influence how credit for the discovery would be assigned.
Sebastien Bubeck, a mathematician and AI researcher at OpenAI, said during a press briefing that the company began training a new AI model with advanced mathematical abilities on August 28. After hearing rumors that Anthropic was making progress on the Navier-Stokes problem, OpenAI decided to commit additional resources to the effort. According to Bubeck, more than 1,000 AI agents worked on the problem for over 50 hours, with the number eventually scaling up to as many as 10,000 agents before the company arrived at a solution. "I thought there must be a mistake somewhere," Bubeck said. "And on Sunday morning we had the final solution, Lean-formalized and everything." Lean is a programming language used to formally verify mathematical proofs.
OpenAI said the project required significantly more computing power than any previous mathematical effort by the company, with costs reaching "in the millions of dollars," according to Mark Chen, OpenAI's head of research.
On Monday, Buckmaster and Alpöge published documents outlining key advances related to the Navier-Stokes problem, saying they had used AI tools including Claude and Codex in their work. Buckmaster also stated that he learned last week that OpenAI had become aware of their research and had begun devoting substantial resources to the same problem. He said he asked OpenAI representatives whether the company had accessed his and Alpöge's Codex logs, and was told the model "didn't look up user data," though he claims OpenAI did not answer questions about how its model was trained. Buckmaster further alleged that OpenAI proposed several options, including one in which he could publish a paper crediting an internal OpenAI model with solving the Navier-Stokes problem while omitting Alpöge's name.
Bubeck disputed this characterization on social media, denying that OpenAI had proposed removing Alpöge from the credit. OpenAI CEO Sam Altman also publicly defended his team's work. Buckmaster, Alpöge, and Anthropic did not immediately respond to requests for comment.
During the briefing, Bubeck and other OpenAI executives denied ever examining Buckmaster and Alpöge's Codex prompts to inform their own research. "We, whether it's the researchers or the agents, did not see any of their work until it was released publicly last night," Bubeck said. In a blog post, OpenAI acknowledged, "while unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models."
Bubeck added, "I want to be extremely clear that we recognize the priority of Levent Alpöge and Tristan Buckmaster's work on unforced Euler, and we have nothing but congratulations to them on this monumental achievement that they have made. To be clear, we did not use their prompt or proof to prompt our models or direct our agents." OpenAI mathematician Ven Chandrasekaran also noted that the company's solution differed substantially in approach from the one developed by Buckmaster and Alpöge.
The dispute over credit for solving one of mathematics' most significant open problems appears likely to continue, and as AI plays a larger role in mathematical research, similar disputes may become increasingly common.