OpenAI has published full or partial solutions to 372 long-standing mathematical problems, prompting excitement and alarm among mathematicians over how quickly AI is advancing in a field once considered resistant to machines.
OpenAI has published full or partial solutions to 372 outstanding mathematical problems, prompting mathematicians to question how artificial intelligence could transform the field.
The announcement on Tuesday (06.10.26) has been dubbed the “mathpocalypse” and comes just weeks after OpenAI revealed a proposed proof addressing the Navier-Stokes equation, one of seven Millennium Prize Problems considered among the most difficult challenges in mathematics.
The latest results are contained in 722 papers, many of which have yet to be independently verified.
Their sheer number and complexity could itself create a major challenge for mathematicians trying to understand and check the work.
Scott Aaronson, chair of computer science at the University of Texas at Austin, compared the development to a “hunter-gatherer” suddenly finding a huge modern resort built beside them.
The rapid progress has surprised researchers, particularly because AI systems struggled with relatively basic mathematics only a few years ago.
Timothy Gowers, a professor of mathematics at the Collège de France and the University of Cambridge, recently described himself as “shocked” by the pace of progress.
He said: “During the course of this year they’ve gone from mediocre PhD student to very good PhD student, to experienced mathematician level, to top mathematician level.”
Some mathematicians have welcomed OpenAI’s results as a potentially transformative advance.
Others have raised concerns about how the work has been released and what it means for human mathematical research.
Terence Tao, widely regarded as one of the world’s leading mathematicians, shared a statement from the Association for Human Mathematics calling for mathematicians to stop working with OpenAI.
The group said: “Releasing over 700 files at once is not a demonstration of scholarship, but a demonstration of power.
“We urge mathematicians to discontinue their work with OpenAI and to return to a vision of science that centres human understanding.”
OpenAI defended its approach, saying it had consulted extensively with mathematicians about how the results should be published.
The company said: “We want this progress to push the frontier of human knowledge and enable further progress in mathematics.”
The work was produced using an internal AI model whose full capabilities have not been disclosed, meaning its achievements could represent only a snapshot of what increasingly powerful systems may be capable of.
For mathematicians, that creates an unusual dilemma.
AI could help solve problems that have resisted human efforts for decades, but researchers may struggle to verify, understand or build upon solutions generated at unprecedented speed.
As Aaronson’s analogy suggests, the question may no longer be whether AI can enter the mathematical world, but whether humans can keep up once it does.
OpenAI sparks concern among mathematicians as it publishes solutions to 372 outstanding mathematical problems
