OpenAI says AI solved Navier-Stokes amid credit dispute

OpenAI said Tuesday that an unreleased artificial intelligence model solved the Navier-Stokes problem in days, a claim that would answer one of mathematics' Millennium Prize Problems if confirmed and has triggered a dispute over credit.

The Navier-Stokes problem concerns equations describing how fluids such as air and water move. The equations are used in aircraft design, weather forecasting and the study of blood flow. OpenAI said its work showed that the equations can break down over time, answering the question set out by the Clay Mathematics Institute.

Each Millennium Prize Problem carries a $1 million reward, and only one has been solved since the list was published in 2000. OpenAI said it would not claim the money if its finding is confirmed.

Under the prize rules, a solution has to be published in a peer-reviewed journal and survive two years after gaining acceptance in the mathematical community before the institute considers it. Clay Mathematics Institute president Martin Bridson said the evaluation would be “deliberately unhurried” and “absolutely rigorous.”

OpenAI said it began training a new model in late August and later directed it at the six remaining open Millennium Prize Problems after seeing online rumours that a competitor had solved some of them. OpenAI researcher Sebastien Bubeck said 10,000 AI agents worked on Navier-Stokes at the final stage, passing messages between one another.

OpenAI chief research officer Mark Chen said the computing costs ran in the millions of dollars. According to OpenAI, the agents reached a solution on Saturday, about 88 hours after the project began.

The announcement came hours after Tristan Buckmaster, a mathematician at New York University, and Levent Alpoge, who works at OpenAI rival Anthropic, released AI-assisted work on three related equations. Buckmaster said OpenAI took up the problem only after word of his research spread and pursued an unusual approach he and Alpoge had spent months developing.

OpenAI researchers denied having access to the rival team's work. Bubeck said OpenAI did not use their prompts or proofs to direct its models or agents. OpenAI later said that, while unlikely, it could not rule out that de-identified data derived from the rival researchers' use of its products helped improve its models.