Mathematician Challenges AI Math Breakthrough Claims

OpenAI announced this month that it had produced solutions to 10 long-standing mathematics problems, some of which had gone unsolved for decades. The claim landed hard in a field that prides itself on moving slowly, checking work, and building on centuries of accumulated proof. James Maynard, a Fields Medal winner and professor at the University of Oxford, told The Verge he's spent much of the past year in what he calls "soul searching" over what AI means for his discipline.

The unease isn't about whether the technology works. Generative AI trained on vast troves of mathematical literature can already recombine known results and methods in ways that surface connections human researchers missed, sometimes pulling together tools from entirely different branches of math. What's less settled is what that capability does to the people who spend their careers on this work, and to the slow, deliberate culture that has defined mathematics for generations.

Maynard isn't alone in feeling pulled between two reactions at once. Other mathematicians describe a similar split: genuine excitement about faster discovery, paired with something closer to dread about what gets left behind.