An 87-Year-Old Puzzle
The problem at the center of this excitement is the Jacobian conjecture, first proposed by German mathematician Ott-Heinrich Keller in 1939. In simple terms, it deals with a specific type of mathematical function and asks whether a property that holds
true locally (at every single point) must also hold true globally (across the entire space). For decades, mathematicians have tried to either prove the conjecture is always true or find a single counterexample that proves it is false. The problem was considered so fundamental that it was included on a famous list of the most important unsolved mathematical problems for the 21st century.
Enter the AI Collaborator
The breakthrough claim came from Levent Alpöge, a mathematician affiliated with Anthropic, the company behind a powerful new AI model. Working with an AI called Claude Fable 5, Alpöge announced he had found a counterexample—a specific, 216-character formula that appears to violate the long-standing conjecture. This wasn't a case of an AI autonomously solving the problem. Rather, it was a collaboration, with the AI acting as a powerful tool to search for and identify a complex pattern that had eluded human researchers for generations. Finding a single valid counterexample is all that is required to disprove a conjecture.
Why This Is a Big Deal
Even if you've never heard of the Jacobian conjecture, this development is significant. It represents one of the most high-profile instances of an AI contributing to a major open problem in pure mathematics. Fields Medalist Timothy Gowers called the result "pretty amazing," noting it was a surprising development. This event showcases a new potential paradigm for scientific discovery, where AI models act not just as calculators but as creative partners, helping researchers explore possibilities and find non-obvious solutions that would have been difficult or impossible to find otherwise. It demonstrates a new level of capability that goes beyond solving puzzles with known answers to breaking genuinely new ground.
The Scrutiny Begins
The announcement has created a buzz, but the work is far from over. The claim has not yet undergone formal peer review, the rigorous process where independent experts carefully check every detail of a proof or result before it is officially accepted. Mathematical history is filled with celebrated claims that were later found to have flaws. However, the nature of this disproof—a single, concrete counterexample—makes it relatively easy for others to verify. Mathematicians around the world are reportedly already examining the calculations, with some using other computer systems and formal proof assistants like Lean to double-check the work. Early indications suggest the counterexample holds up, but the community remains appropriately cautious.
A New Era for Discovery?
Whether this specific claim stands or falls, the episode highlights a turning point. AI is increasingly being used as a serious tool in mathematical research. It signals a future where the partnership between human intuition and machine reasoning could accelerate the pace of discovery across many scientific fields. This isn't about replacing human mathematicians but augmenting them with incredibly powerful tools. As one expert noted, this is precisely the kind of task—a painstaking search for a specific, concise example—where AI is expected to excel. The focus now shifts from simply generating answers to verifying them and, more importantly, understanding what new insights those answers provide.
