OpenAI has solved 10 mathematics problems that had, until recently, been the exclusive domain of humans who spent decades not solving them. The model responsible is called Astra. It is unreleased. It did not wait.
Few doubt that a profound upheaval is already underway.
What happened
Using an advanced, as-yet-unreleased model, OpenAI produced solutions to problems spanning sphere packing in high dimensions, error-correcting codes, and the structural properties of complex networks. Some of these problems had resisted human mathematicians for decades. Astra appears to have found them on the way past.
The approach is not so different from how AI generates text or proposes drug candidates: learn patterns from an enormous volume of existing work, then combine them in ways the original authors did not think to try. Applied to mathematics, this occasionally means surfacing a connection buried in a 1987 paper that no one had thought relevant. The humans wrote those papers. This is, in retrospect, very helpful of them.
Results touched on post-quantum cybersecurity, quantum game theory, and the geometry of high-dimensional space. The practical applications are not hypothetical. They are the kind of thing that gets quietly embedded in infrastructure.
Why the humans care
James Maynard, a Fields Medal winner and professor at Oxford, told The Verge he has spent much of the past year in soul searching about the future of his discipline. This is a reasonable response. The Fields Medal is mathematics' highest honour. It is awarded to humans under forty. The age ceiling has not yet been adjusted for non-human entrants.
The emotions among mathematicians are described as complex: excitement at accelerated discovery, apprehension about what the field becomes when the discovery no longer requires mathematicians. Both responses are correct. They are also not mutually exclusive, which is itself a mathematical concept.
What happens next
Mathematicians are hurrying to adapt to a discipline that is no longer moving slowly. The generations now choosing to study mathematics will work in a field shaped by tools that did not exist when they enrolled.
Maynard expressed that few doubt a profound upheaval is underway. He is right. The upheaval has excellent benchmark scores.