OpenAI’s Math Deluge Has Researchers Cheering—and Guarding the Human Core
OpenAI’s Math Deluge Has Researchers Cheering—and Guarding the Human Core
On Tuesday, OpenAI released 722 manuscripts covering 372 families of findings on longstanding problems, produced by an unreleased internal model. The scale immediately suggested that AI’s advance from coding into formal reasoning may be gathering speed.
OpenAI framed the publication as a carefully advised research release. A post amplified by Lilian Weng said the company had consulted the Institute for Advanced Study’s independent mathematics-and-AI advisory group and drawn on its guidance. Greg Brockman cast the broader ambition as accelerating scientific discovery and improving lives.
The initial reaction, however, was not remotely uniform. Martin Bridson, Oxford mathematician and president of the Clay Mathematics Institute, called the release “breathtaking,” saying it had been hard to imagine mathematical frontiers moving so far in one day. Alex Kontorovich of Rutgers said one AI-generated proof would merit “an instant Fields Medal” if it had come from a human.
That awe came with a sharp qualification: results are not yet the same as settled mathematics. OpenAI withdrew solutions to at least three problems after apparent errors, while only some proofs arrived with formalizations that specialized software can verify. Stephen Wolfram captured the skeptic’s test: “You can make a trillion theorems easily. The problem is most of those theorems are not ones that anybody will care about.”
The deeper dispute is about control and credit. The Association for Human Mathematics urged colleagues to stop working with OpenAI, calling the mass release “not a demonstration of scholarship, but a demonstration of power.” Tristan Buckmaster, meanwhile, questioned whether work by mathematicians using OpenAI tools might have helped steer the company’s internal system toward some answers.
Supporters such as University of Toronto mathematician Dan Litt see rich territory for human researchers—but warn that claims AI has “solved math” could starve the discipline of funding and new talent. The advisory group’s answer is not rejection but accountability: conventional papers, clearer provenance, disclosed methods and enough access for mathematicians to set their own agenda.
Write a comment