On 21 September OpenAI announced an independent advisory group for mathematical research. Called the Advisory Group on Mathematics and Artificial Intelligence, it is hosted at the Institute for Advanced Study (IAS) in Princeton. The bigger news is the claim that came with it: OpenAI says the same internal model has resolved more than 100 additional open problems across most areas of mathematics.

The claim follows the company's abrupt publication of a solution to the Navier-Stokes Millennium Prize problem. Part of the mathematical community objects to the pace. Earlier in September, 25 Fields Medal winners signed an open letter arguing that AI labs threaten mathematicians' intellectual work as they race to one-up each other on famous problems.

What the group will do

According to OpenAI's post, as quoted by TechCrunch, the group "will serve as a bridge to the mathematical community and broader public, giving mathematicians a voice in how we move forward." The current task, as the group describes it on its own site, is very concrete: advising OpenAI on how to coordinate the release of a large number of significant results its internal model is reported to have produced.

The group's terms:

  • Members are unpaid, and the group operates independently of any AI company.
  • Recommendations will be published on the group's website.
  • Members can offer unsolicited advice, go public with their views and control their own membership.
  • The group is willing to advise any AI company whose models are likely to have a significant impact on mathematics.
  • It has no decision-making power at any company.

The last point matters. OpenAI's post states plainly that "the group will not be responsible for advising us on how to pace our internal progress on mathematics." IAS makes the same point: "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."

The members

The first nine members are François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten and Melanie Matchett Wood. The list includes some of the field's best-known names, among them Fields medallists Gowers, Hairer and Witten. According to TechCrunch, only De Lellis also signed the letter from the 25 Fields medallists.

The group's site says OpenAI first approached some of its members about setting up an external advisory board tied to the company. They chose instead to form an independent group and invite the others, and OpenAI agreed.

Why it matters

In mathematics, a claim to have solved an open problem only means something once others have read and checked the proof. Releasing more than a hundred results at once pushes that verification work onto volunteer referees and makes it harder to tell which results are truly new and which are small steps. That is exactly the group's job: assessing the significance of results and coordinating their release.

The group is also asking the mathematical community for input through an online form. Responses will inform its recommendations and will not be made public without permission. No date has been given for the first recommendations.