OpenAI posts 700 maths preprints online: mathematicians are up in arms

In August, a gathering of roughly 40 renowned mathematicians convened at OpenAI's request to assess the progress of artificial intelligence in the mathematical domain. During these conversations, OpenAI asserted that it had resolved numerous intricate mathematical puzzles that remained unsolved by humans. By September, a storm of controversy erupted concerning OpenAI's claims of attribution, data privacy, academic plagiarism, and transparency.

Two mathematicians, Tristan Buckmaster and Levent Alpöge, had been leveraging OpenAI's tools to aid their research on the Navier-Stokes existence and smoothness problem, a challenge associated with the prestigious Millennium Prize Problem, which promises a $1 million award. Before they could share their results, OpenAI revealed that its internal model had already resolved the same aspects of the problem.

In response, OpenAI announced on September 8 that it had been training a superior internal model since August 28, with performance continuously enhancing. The company disclosed a solution to the Navier-Stokes existence and smoothness issue, a formalization of the proof in Lean, and a write-up of the discovery. This revelation threatened to overshadow Buckmaster and Alpöge's considerable efforts.

Furthermore, they accused OpenAI of exploiting their private conversations to expedite its own research. Shortly after, Buckmaster sent OpenAI a stern ultimatum: either co-author the paper with the company or allow OpenAI to publish first.