OpenAI has created an external advisory group to evaluate AI’s mathematics research results and discuss how they are disclosed. The step follows concerns in the mathematics community about verification and recognition of researchers’ contributions after the company announced its AI was rapidly solving unsolved math problems.

On Sept. 21, IT media outlet TechCrunch reported that OpenAI said it had set up a mathematics and AI advisory group at the Institute for Advanced Study (IAS) in Princeton, New Jersey. The initial members are 9 mathematicians. They will assess the importance of AI research results and coordinate the timing and method of disclosures.

On Sept. 8, OpenAI announced a solution to the Navier-Stokes equations problem, one of mathematics’ Millennium Prize Problems. It then claimed the same internal AI model had solved more than 100 additional unsolved problems across several areas of mathematics. However, this was OpenAI’s announcement and does not mean each solution has completed independent verification by the mathematics community.

The advisory group will not be paid by the company. It can offer opinions or publicly state its position even without a request from OpenAI. It will also manage member appointments on its own. However, it has no authority to slow OpenAI’s internal research pace or change its research direction. The Institute for Advanced Study also said it provides advice, but decision-making authority at AI companies and the resulting responsibility lie with the companies.

Earlier, 25 Fields Medal-winning mathematicians said in an open letter they were concerned that mathematicians’ intellectual work could be threatened as AI labs competitively announce solutions to famous unsolved problems. Among the new advisory group’s members, the only person who signed the letter is Camillo De Lellis (카밀로 데 렐리스) of the Institute for Advanced Study.