‘Depressing that machine solved 90-year-old math problem’: Indian scientist on OpenAI’s claim

On September 8, OpenAI made a startling claim: their unreleased internal model had solved a 90-year-old math conundrum known as the Navier-Stokes Millennium Prize Problem. This solution is worth a $1 million prize from the Clay Mathematics Institute. OpenAI's 166-page paper only cites a 2017 proof by Indian scientists Nishant Singh and S Sridhar, who were initially reluctant to publish their findings.

Singh, now an associate professor of astrophysics, explained that he and Sridhar were against the trend of publishing many papers, believing it harms the progress of science. Their proof, which focused on exact solutions to the Navier-Stokes equations, was initially rejected but later published in 2017. OpenAI's claim, if verified, would make their achievement the second Millennium Prize Problem solved, following Grigoriy Perelman's resolution of the Poincaré Conjecture in 2003.

However, Singh expressed disappointment that a machine had solved this problem while many human mathematicians had struggled for decades. He also questioned the usefulness of a dry solution without the human insight that comes with trial and error. Despite this, Singh acknowledges potential benefits of AI-generated solutions, such as enhancing weather prediction tools.

Renowned mathematician Terrance Tao also voiced concern over the misuse of powerful AI tools, fearing it might temporarily advance problem-solving but harm future progress and understanding of existing solutions. OpenAI's decision came after rumors that rival AI firm Anthropic had supposedly solved the problem, with mathematicians Levent Alpöge and Tristan Buckmaster also working on a related part of the problem using OpenAI's Codex, potentially exposing the company to controversy.

OpenAI maintains they cannot rule out de-identified data derived from Anthropic's usage contributed to their model's improvement.