OpenAI to Dump Hundreds of AI Math Solutions on GitHub — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI plans to release hundreds of AI-generated math solutions on GitHub on Tuesday, people familiar with the plans told WIRED, despite having consulted ~40 mathematicians in August who urged the company to publish proper papers rather than blog posts or tweets.
- Lindsay McCallum, OpenAI's spokesperson, said the company began training a new model on August 28 that has resolved the Navier-Stokes Millennium Prize problem and more than 100 long-standing open problems across most areas of mathematics.
- Tristan Buckmaster, an NYU mathematician, accused OpenAI of front-running his work on a Millennium Prize problem in September after the company deployed thousands of agents upon hearing "rumors" others were closing in; he claims OpenAI researcher Sébastien Bubeck implied his collaborator Levent Alpöge should be excluded from any paper and threatened him during credit negotiations.
- Mathematicians including Bryna Kra of Northwestern and Nestor Guillen of NYU told WIRED there is a "perception of mobster behavior" from AI companies, and that traditional scientific processes for releasing and attributing results have been cast aside as OpenAI and Anthropic race toward their respective IPOs.
- OpenAI assembled an advisory group in mid-September to help set standards for communicating new results, but mathematicians say the company has made little progress and has not adopted community-built verification tools like Hexagon and Palomar.
- Some OpenAI employees reportedly believe their technology has rendered mathematics obsolete, with one person telling WIRED that mathematicians were briefed about breakthroughs the way "police might want to notify family before reporting a death."
Why it matters: Mathematicians who build careers on properly attributed, peer-reviewed work now face hundreds of AI-generated results being dumped on GitHub without traditional scientific papers, and the community-built verification tools (Hexagon, Palomar) that could help triage the deluge are being ignored by OpenAI. The result: a growing credibility crisis for AI-generated mathematics just as both AI labs race toward IPOs.
Ask SkimNews
