OpenAI has a new plan to push AI into advanced mathematics
OpenAI has formed an independent maths advisory group to review results and guide responsible communication after claiming its AI resolved more than 100 open problems, including Navier-Stokes.
OpenAI has announced an independent Advisory Group on Mathematics and Artificial Intelligence after claiming that a new internal model has resolved the Navier-Stokes problem and more than 100 other long-standing mathematical problems.
The announcement comes just weeks after OpenAI's Navier-Stokes claim triggered a debate over the proof, the role of AI in mathematical research and who gets credit when machines help produce a breakthrough.
The maths claim that started it
On September 8, OpenAI said its internal AI system had solved the Navier-Stokes existence and smoothness problem, one of seven Millennium Prize Problems.
The company says around 10,000 AI agents worked on the problem for 88 hours before producing a result. OpenAI then used Lean, a system that can mechanically check mathematical proofs, to formalise and verify the argument.
But "verified" does not automatically mean the mathematical community has accepted the solution.
However, experts still need to examine whether the argument is correct and whether it actually satisfies the precise problem defined by the Clay Mathematics Institute. Clay's rules require a proposed solution to go through publication and sustained community scrutiny before its $1 million prize can be awarded.
Then came the credit dispute
The Navier-Stokes announcement also brought an unusual academic dispute.
NYU mathematician Tristan Buckmaster and Anthropic researcher Levent Alpöge had been working on related fluid-dynamics problems using AI tools, including Codex and Claude. Buckmaster alleged that OpenAI had learned about their unpublished work and moved quickly on a related result.
ChatGPT-maker disputes that account. The company says its researchers and agents did not see their work before it was publicly released and says no specific user data was accessed to solve the problem.
OpenAI has, however, acknowledged that it could not completely rule out de-identified product data indirectly contributing to model improvement.
The dispute has added another question to the AI-mathematics conversation: when an AI system produces a result, how should researchers establish credit and provenance?
Why OpenAI is bringing mathematicians in
That is where the new advisory group comes in. OpenAI says the group will act as a bridge between the company, mathematicians and the wider public.
It will advise on the significance of new results, how they should be communicated and the academic standards around mathematical research. It will also advise on how AI can support research and learning.
The group will operate independently, with members unpaid by OpenAI and able to offer unsolicited advice or publicly share their views. Its nine initial members include Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ravi Vakil, Edward Witten and Melanie Matchett Wood. It is hosted at the Institute for Advanced Study.
There is one important catch: the group does not have decision-making power over OpenAI and will not advise the company on how quickly it should pursue mathematical progress. That makes the new panel less of a brake and more of a bridge.
OpenAI is moving quickly in mathematical research. Now it is also trying to build a better way for mathematicians to examine, understand and debate what those systems produce.