OpenAI posts 722 AI-generated math preprints, pulls three

On 6 October, OpenAI uploaded 722 AI-generated mathematical manuscripts to a GitHub repository, covering purported progress on 372 open problems in geometry, algebra and computer science, Nature reports. A day later, OpenAI withdrew three of them after a sign error invalidated the argument in one paper and the constructions used by two dependent manuscripts.

"The sheer number of papers posted at the same time is 'staggering'," wrote Alvaro Lozano-Robledo, a number theorist at the University of Connecticut in Storrs, on X. "Why is OpenAI trying to solve hundreds of math problems to be released all at once?"

One physicist called the dump a "slopocalypse". A group of mathematicians went further, writing that "Mathematicians did not ask for this work to be done" and that "Releasing over 700 files at once is not a demonstration of scholarship, but a demonstration of power," urging colleagues to discontinue their work with OpenAI.

Not everyone agreed. Levent Alpöge, a mathematician at Anthropic, called the release "the most significant moment in mathematical history" and singled out quasi-Riemann and no-Siegel-zeros results. Two researchers in our Who's Who tracker shared the Nature piece.

The manuscripts were produced by an unreleased internal frontier model pointed at roughly 4,000 open problems. The batch does not contain complete solutions to any of the five remaining Millennium Prize Problems, though some papers claim partial progress on the Riemann hypothesis and the Hodge conjecture. OpenAI's stated goal, per a company spokesperson, is to "provide tools for mathematicians and scientists".