Mathematicians get a say in how OpenAI reports its math

OpenAI said on September 21 that an independent group of mathematicians will advise it on how it assesses and communicates new mathematical results, in a post on X. The body is called the Advisory Group on Mathematics and Artificial Intelligence. It is hosted at the Institute for Advanced Study in Princeton and starts with nine members: François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten and Melanie Matchett Wood.

OpenAI listed two more jobs for the group: upholding academic and professional standards, and building tools that support mathematical research and learning. The announcement came after the abrupt publication of a solution to the Navier-Stokes Millennium Prize problem. OpenAI also claims the same internal model has resolved more than 100 other open problems across most areas of mathematics, TechCrunch reported.

Mathematicians had already pushed back. On September 9, Caltech mathematicians wrote that the school’s mathathon was “likely to have destructive impacts”, and OpenAI stopped sponsoring the event a day later. Also in September, 25 Fields Medal winners signed an open letter arguing that AI labs threaten their intellectual work as they try to one-up each other on famous problems. Camillo De Lellis of IAS is the only one of the nine group members who signed it.

The members are unpaid. They can give advice OpenAI did not ask for, take their views public and decide who joins. According to the group’s website, OpenAI first approached some of its members about setting up an external advisory board. In agreement with OpenAI, they formed an independent group instead and invited the others to join. The group says it will publish its recommendations on the site and will advise any AI company whose models are likely to have a significant impact on mathematics.

Hairer said the group receives no financial, technical or other support from OpenAI, and that no member has signed agreements limiting what they can say or do beyond the standard confidentiality required for advance access to research, The Verge reported. The outlet adds that it is still unclear which of the nine OpenAI approached first.

The group’s role also has limits. OpenAI’s announcement says “the group will not be responsible for advising us on how to pace our internal progress on mathematics.” The IAS put it this way: “Although we will give advice, we do not have decision making power at any AI company, and the responsibility for the decisions made by any company will rest with that company.”