OpenAI Releases Solutions to Hundreds of Math Problems
OpenAI published solutions to hundreds of long-standing mathematical problems on GitHub, sparking backlash over the bypass of traditional peer-review processes.
OpenAI released 722 manuscripts on GitHub containing solutions to hundreds of long-standing mathematical problems, including the Navier-Stokes Millennium Prize problem. The company stated that a new internal model trained since August 28 resolved more than 100 open problems across most areas of mathematics.
The release drew sharp criticism from the mathematical community and academics at New York University and Northwestern University. Critics argue that OpenAI and its rival Anthropic are bypassing traditional scientific peer-review processes, instead using blog posts and social media to promote their models ahead of potential initial public offerings.
Controversy intensified following allegations from New York University professor Tristan Buckmaster, who accused OpenAI of front-running his work on a Millennium Prize problem. Internal meeting notes suggest OpenAI researcher Sébastien Bubeck pressured Buckmaster to exclude a collaborator from Anthropic and allegedly warned that Buckmaster would ruin his career if he went to the press.
OpenAI spokesperson Lindsay McCallum denied claims of "mobster behavior" and stated the company is working with the Advisory Group on Mathematics and Artificial Intelligence to inform its release process. However, the Advisory Group has urged AI labs to use established academic channels rather than using scientific results as marketing vehicles.
