OpenAI posts 700+ AI maths preprints, mathematicians balk

On 6 October, OpenAI uploaded more than 700 mathematical preprints to GitHub, all generated by an unreleased model, according to Nature. The batch did not include solutions to any of the five remaining Millennium Prize Problems, though a separate OpenAI preprint in September had claimed progress on the Navier-Stokes equation.

The reaction from the maths community has been openly hostile. OpenAI was 'tossing the community a messy first draft and expecting us to clean it up for them,' said Ursula Martin, a mathematician at the University of Oxford. One mathematician called the release a 'slopocalypse.'

Michael Harris, a number theorist at Columbia University, has opted out entirely. 'I can't bring myself to spend my time studying [these preprints], so I will not be reading these papers,' he said. Alvaro Lozano-Robledo, a number theorist at the University of Connecticut, put the puzzlement plainly: 'Why is OpenAI trying to solve hundreds of math problems to be released all at once?'

Not everyone was dismissive. Levent Alpöge, a mathematician at Anthropic, called the technology's usefulness to mathematicians 'the most significant moment in mathematical history.' He also acknowledged there would be 'sad stories' for researchers whose in-progress work had been pre-empted by the dump.

Three AI researchers we follow flagged the Nature piece the same day it ran.