OpenAI's Astra AI cracks 10 longstanding maths problems
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI revealed solutions to 10 longstanding mathematical problems found by its prototype AI Astra, spanning topics from sphere packing in higher dimensions to quantum game theory
- The standout result: a single counterexample disproving the conjecture that all countable groups are "sofic" — a property first described in 1999 that every previously known group had possessed
- Francesco Fournier-Facio at the University of Cambridge, who has worked on soficity since starting his PhD in 2020, says the solution builds heavily on 2016 and 2019 papers by Andreas Thom and Gábor Kun, arguing AI "couldn't have [done it] without prior human work"
- OpenAI initially claimed the 10 problems had seen "no progress... for at least a decade," but Fournier-Facio called this inaccurate and the company revised its statement
- Fournier-Facio argues current AI math wins favor easily verified counterexamples over developing new theory, where "there's no tick at the end" to confirm correctness
- Abhishek Saha at Queen Mary University of London called the release "the most impressive display of AI mathematical prowess to date," noting some solutions are actual proofs, not just counterexamples
- Elon Musk tweeted the discovery was evidence the singularity has been reached
Why it matters: Fournier-Facio warns the 'AI breakthrough' narrative obscures human work — the soficity counterexample relied on Thom and Kun's 2016/2019 papers — and OpenAI had to revise its claim that the 10 problems had seen no progress in a decade, a transparency miss that complicates how mathematicians credit AI-assisted discoveries.
