OpenAI claims AI agents solved Navier-Stokes problem in 88 hours

OpenAI said it used a system of coordinating agents powered by its internal AI model. The agents were given tools including a cached version of the internet and the ability to run code, and they were subdivided into groups that could communicate internally. The group that produced the Navier-Stokes resolution involved on the order of 10,000 concurrent agents.

The Navier-Stokes equations describe how fluids move and are one of seven Millennium Prize Problems designated in 2000 by the Clay Mathematics Institute in Cambridge, Massachusetts. Each problem carries a $1 million prize. The institute has not commented on OpenAI's proposed solution.

OpenAI said the effort began on September 1 after it heard a rumor about progress on the problem that it later learned involved Tristan Buckmaster, a professor of mathematics at New York University, and Levent Alpöge, who works at OpenAI rival Anthropic. Buckmaster said he and Alpöge had been working in a personal collaboration on math problems including Navier-Stokes, and said Alpöge received tips that information about the pair's progress had been passed to OpenAI.

Buckmaster said OpenAI's route to the solution resembled the pair's own work and was not the direction one arrives at in a few days by giving a model the problem statement. He also raised questions about whether OpenAI models had been trained on, or had access to, sessions that the mathematicians held in OpenAI's Codex, noting that they had used several large language models in their work. He said he had not seen OpenAI's proof and did not know whether the pair's data was used.

OpenAI said its researchers and agents did not see Alpöge and Buckmaster's work through any means until it was released publicly and that no specific user data was accessed to solve the problem. The company added that it could not rule out that de-identified data derived from the pair's use of OpenAI products helped improve its models.