kidkyan ⭐️

A Fields Medal is basically the Nobel Prize of math. One of the guys who won it cried so hard Tuesday night that he woke up his daughter. The reason? OpenAI dumped 719 AI-written proofs onto the internet in one night, claiming to solve math problems that have stumped humans for decades. 16 of them were problems Hugo Duminil-Copin spent his career working on. He told Le Monde they annihilated his field of research. Here's the part the headlines are missing though. Nobody actually knows if the proofs are right. Not one of the 719 has been peer reviewed. Three already got pulled within a week because of a single sign error, and 14 more got quietly revised. The model spent about 3 hours of compute per result, but a human checking the 3,500 pages covering just Hugo's 16 problems would need years. So every working mathematician wakes up to the same lousy choice. Spend the next decade grading a machine's homework, or ignore the dump and risk burning years on a problem that's secretly already solved. And the reaction might do more damage than the papers. Some mathematicians are already saying they'll share fewer of the problems they work on. Duminil-Copin says the whole culture of circulating ideas is shattering. That culture is the engine of the field. Hilbert posted 23 open problems in 1900 and they steered a century of research. Posting an open problem is how a young mathematician finds their life's work. If posting one now means handing a lab a target, people stop posting. Which means OpenAI could be wrong about every single proof and still break mathematics. A field where people stop sharing problems stops being a field.