OpenAI said an internal AI model that recently produced a proposed solution to the 90-year-old Navier–Stokes Millennium Prize Problem has now resolved more than 100 long-standing open problems across most areas of mathematics.

“The pace of its progress⁠ in mathematics has surprised the mathematicians within OpenAI,” said the company in a blog post. 

Thus, OpenAI is now establishing an independent Advisory Group on Mathematics and Artificial Intelligence, hosted at the Institute for Advanced Study. 

The nine-member group will advise on reviewing emerging results, communicating their significance, coordinating their dissemination and applying academic and professional standards to AI-generated mathematical research.

The announcement comes shortly after mathematicians published an open letter, ‘A Severe Misalignment of AI in Mathematics’, raising concerns about the use of solving open mathematical problems as a benchmark for AI systems. 

The letter argues that rapid AI progress could create negative externalities for the mathematical research community. 

The declaration, published by 25 Fields Medal winners, including Terence Tao, Peter Scholze and Cédric Villani, argues that AI companies’ focus on solving mathematical problems as benchmarks risks prioritising answers over the understanding and ideas that mathematics is intended to develop.

The mathematicians said AI-generated results can also arrive too quickly for the community to properly explain, simplify and connect them to previous work, raising concerns around attribution and plagiarism. 

They argued that AI could accelerate mathematical research, but said its impact would depend on whether researchers preserve the human processes through which new ideas are understood, discussed and incorporated into the field.

Therefore, this group will advise on how AI tools can support mathematical research and learning. OpenAI said it wants mathematicians to use increasingly capable AI systems to pursue research questions and develop new ideas.

“We want to put capable tools in mathematicians’ hands so they can pursue the questions they know best and develop new ideas,” OpenAI said.

The initial group includes mathematicians François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten and Melanie Matchett Wood.

OpenAI said the group would operate independently, with members free to exercise their own judgement, challenge the company and publish advice publicly. 

Its members will not be paid by OpenAI, will control the group’s membership, and will not have a role in setting the pace of OpenAI’s internal mathematics research