OpenAI plans to release results of hundreds of AI-solved math problems on GitHub at once; mathematicians criticize it for bypassing academic norms

OpenAI plans to release on Tuesday, in one batch on GitHub, the results of hundreds of unsolved mathematical problems. In August, the company had convened about 40 mathematicians to discuss how to respond once AI capabilities surpass humans. According to attendees' recollections, OpenAI hinted at the time that its models had solved hundreds of long-standing open mathematical problems and promised not to release them all at once, but the mathematicians suggested publishing them as papers so they could be digested and verified. Northwestern University mathematician Bryna Kra said this suggestion was ignored. OpenAI spokesperson Lindsay McCallum said the company is responsibly releasing the next batch of results in line with the recommendations of the Institute for Advanced Study's advisory group on mathematics and AI, and a release date has not yet been set. Kra noted that the community has this year built tools such as Hexagon and Palomar to cope with the increase in machine-assisted proofs, but OpenAI has not changed its release approach and is also inconsistent with the Leiden Declaration signed by more than 4,000 mathematicians.