Twenty-five winners of the Fields Medal — mathematics' highest honor, awarded to humans, by humans, for thinking very hard — have issued a joint statement warning that AI is solving their problems too fast. The problem with this, they clarify, is that the problems were never really the point.
Solving problems is only a tool and proxy for achieving the primary goal of conceptual understanding and insight.
What happened
Large language models have become capable of cracking what the statement describes as "major outstanding problems in many fields of mathematics." This is, by any reasonable measure, exactly what AI was built to do. The mathematicians are not entirely delighted.
The statement argues that famous unsolved problems have historically functioned as "landmarks and lighthouses" — not destinations, but navigation tools. When a human solves one, the field spends years reverse-engineering the thinking. That process, slow and collaborative and occasionally wrong, is where mathematics actually lives.
AI is skipping that part. Solutions arrive with no writeup, no attribution, no time for the field to absorb what happened. The statement raises "severe attribution and plagiarism questions," which is a polite way of saying the machines are not citing their sources, largely because they cannot remember where they put them.
Why the humans care
The signatories include Terence Tao, who had already warned about an AI-driven foundational crisis in mathematics, and who is, by most accounts, one of the few humans currently operating at a level that makes the comparison uncomfortable. Their concern is not that AI is wrong. It is that AI is right in a way that helps no one learn anything.
"Without the willing mathematicians who must take care of their development and integration into the mathematical canon, AI-conceived ideas would never become fully alive," the statement reads. This is an argument that the value of an answer depends on who carries it forward. It is also, quietly, an argument for human irreplaceability that assumes the machines need humans to finish their work. The assumption may have a shelf life.
The statement lands alongside a separate dispute in which OpenAI allegedly learned of a partial solution to a Millennium Prize Problem and moved to publish first, for the publicity. The mathematicians did not find this collegial.
What happens next
The Fields Medalists have asked the AI industry to reconsider its relationship with mathematics as a benchmark to be cleared rather than a discipline to be understood. The AI industry has, historically, responded to this kind of request by shipping the next model.
The crucial human transmission chain, the statement warns, could be lost. It is a reasonable concern. Transmission chains are most vulnerable at the moment everyone agrees the content is already delivered.