OpenAI released 722 mathematical papers, which prove and disprove a range of problems, according to reports. The release included 30 papers pertaining to number theory, of which only one was formally verified in Lean.
Furthermore, the AI system reportedly solved a problem connected to the Navier-Stokes equations of fluid dynamics. Francis Johnson noted that he previously spent twenty-five years working on the problem, producing two books on it, and that he personally gave up.
Regarding the speed of the achievement, Johnson stated, “I'm surprised that [AI] has done it quite so quickly, but I'm not surprised that it's done it.” Terence Tao cautioned that the manner in which discoveries are being generated and “dumped” onto the mathematical community might actually damage the field.
The Advisory Group on Mathematics and Artificial Intelligence, which was established, suggested that companies should disclose the model's name, the prompts used, and the estimated computation cost. The Advisory Group also recommended that any proof released by an AI laboratory should be formalized as much as possible.
Moreover, the Group suggested that companies should reveal how many other problems of similar difficulty their models attempted and failed to solve, alongside details on how those problems were selected. Lindsay McCallum Rémy stated that they are continuing to explore other community-hosted alternatives for this release that meet the committee’s guidelines.
The Advisory Group further advised that the mathematical community should be given time and space to assess this work, while also committing to improving future papers via citations, mathematical exposition, and result presentation.
