Industry Odisha Bureau, Oct 08: OpenAI revealed that an unreleased AI model has produced results for 372 challenging mathematics problems, covering areas including algebra, number theory, topology, mathematical logic, and theoretical computer science.
OpenAI CEO Sam Altman shared the announcement on X. He described the development as the beginning of a new era of scientific discovery. The company said the results were generated using the same model associated with its recent work on the Navier-Stokes problem.
According to the details from an official source, the model was tested on around 4,000 mathematical problems, with the vast majority of the reported results produced through the same process. OpenAI said each result used an average of about three hours of ChatGPT Pro thinking computation.
However, OpenAI stressed that the work is still part of an ongoing process, with some results yet to undergo peer review. The company said it was publishing additional technical details to improve transparency and allow researchers to examine the findings.
The announcement follows OpenAI’s recent consultations with an independent advisory group on mathematics and AI at the Institute for Advanced Study. The group described the release as the beginning of the process of incorporating AI-generated work into established mathematical knowledge.
OpenAI said it plans to support workshops, conferences and other programmes focused on understanding major mathematical results produced by AI.
