OpenAI's dump of more than 300 maths proofs - solutions or partial solutions - to some of mathematics' most enduring problems has understandably shaken the international maths community to its core, if not destroying its foundations completely. However, anyone following the edge of AI closely enough over the past year or two knew that this was imminent.
On the night of 6 October, OpenAI released a dump of 722 AI-generated mathematics papers, representing somewhere north of 372 proofs (or partial proofs) of unsolved and open maths problems (about 300 of them have been formally verified). The proofs wiped out decades (and more) of ongoing human enquiry in several mathematical subfields. And each proof took a mere three hours of compute time.
To say this went off like a thermonuclear weapon in the maths community would be a pallid understatement. It is much worse than that. Thousands of careers will lose their fuel and their joy, maths PhD programmes will be decimated, university students will flee to other subjects, and some maths departments might disappear in the wake of student exits. One commentator was more brutal. He said, without much emotion - maths is cooked.
AI achievement vs human tragedy
Follow us on WhatsApp | LinkedIn for the latest headlines
It is very difficult to celebrate this considerable AI achievement without due discomfort at the human tragedy it now represents and what it portends. We'll get to that, but first let's attend to some of the grief and loss.
On the night of the dump, Hugo Duminil-Copin, winner of the 2022 Fields Medal - the closest...