OpenAI released 722 manuscripts on GitHub, detailing progress on 372 major math problems, including solutions. These findings, from an unreleased ChatGPT model, may intensify AI debates in the math community. The company shared the model’s reasoning and compute estimates, though some advisory group recommendations were not fully followed.
The results were expected, as OpenAI said last month that its models had resolved “more than 100 long-standing open problems across most areas of mathematics.” OpenAI included the model’s reasoning, compute estimates, and information about the number of problems attempted.
OpenAI said it followed guidelines for release of the papers as recommended by its independent Advisory Group on Mathematics and Artificial Intelligence (AGMAI). They are publishing the results in a GitHub repository, with protocols for paper revisions and citations.
Read Also: Dual Ethernet ports boost reliability for servers and networks
Some of the key results claimed by OpenAI are a solution to the four-dimensional Kakeya conjecture, along with improvements to key computer algorithms and progress toward the Riemann hypothesis. Nearly every paper was produced in response to a single prompt given to a single AI agent.
The results still need to be assessed by mathematicians before their impact can be understood. MIT mathematician Andrew Sutherland told SciAm, “Until and unless they release the model and people can replicate their results, I think you should treat any claims about one-shotting problems with a single agent as unverified.”