An Enduring Mathematical Mystery
Imagine drawing a map on a sheet of rubber, then stretching and twisting it without ever tearing or folding it back on itself. The Jacobian Conjecture, first proposed by German mathematician Ott-Heinrich Keller in 1939, asked a question that was similar
in spirit. It dealt with polynomial functions, which are fundamental building blocks in algebra. In simple terms, the conjecture proposed that if a mathematical transformation using these functions was locally reversible at every single point, it must also be globally reversible. In other words, if you could undo the process in any small area, you should be able to undo the entire transformation perfectly, returning every point to its original position. For decades, this seemed intuitively true, yet no one could construct a definitive proof. The problem became legendary, attracting and defeating generations of brilliant mathematicians and earning a spot on a prestigious list of mathematical problems for the 21st century.
A Counterintuitive Answer
The long-standing puzzle was not solved with a complex, multi-page proof. Instead, it was disproven with a counterexample—a single, concrete case that violates the proposed rule. This breakthrough was the work of Levent Alpöge, a mathematician affiliated with Harvard University and the AI company Anthropic. Working with Anthropic’s frontier AI model, Claude Fable 5, he discovered a specific function that met all the starting conditions of the conjecture but was not, in fact, globally reversible. Alpöge announced the discovery not in a formal journal but in a casual post on the social media platform X. The counterexample itself was astonishingly brief, consisting of a 216-character polynomial map. In mathematics, a single valid counterexample is all it takes to prove a universal conjecture false. The dream that had stood for 87 years was broken in an instant.
The AI-Human Collaboration
The role of the AI was not merely to check calculations. Instead, it acted as a true research partner. While many mathematicians over the decades had tried to prove the conjecture was true, the AI was reportedly tasked with searching for evidence that it might be false. This involved navigating a near-infinite space of mathematical possibilities to find a 'needle in a haystack'—the one specific function that didn't play by the rules. The discovery highlights a new paradigm in scientific research. Finding the counterexample was not a simple matter of a single prompt, but likely involved a sophisticated dialogue between the mathematician and the machine. While the exact process has not been fully detailed, it represents a powerful fusion of human intuition and artificial intelligence's brute-force search and pattern-recognition capabilities.
Verification and Reaction
Given the history of failed proofs surrounding the conjecture, the mathematical community was initially skeptical but moved with incredible speed. Because the counterexample was so explicit and short, it could be tested and verified by others almost immediately. Mathematicians around the world used computer algebra systems and formal proof assistants to confirm that Alpöge's function did indeed disprove the conjecture for general cases in three dimensions and higher. The two-dimensional version of the problem, however, remains unsolved. The reaction was a mix of shock and excitement, acknowledging that a major open problem had fallen while also heralding a new era. Some noted that while finding a counterexample is a huge achievement, it is a type of search problem where AI excels, and it may not offer the same deep structural insight a traditional proof would have provided.
A New Frontier for Science
The fall of the Jacobian Conjecture is not an isolated incident. It is the latest and perhaps most high-profile example of AI contributing to frontier science. In recent years, AI models have helped solve other long-standing problems in physics and combinatorics, and have even started achieving perfect scores in the International Mathematical Olympiad. What was science fiction just a few years ago is now becoming a reality: AI is evolving from a tool that analyzes data to a partner that can generate novel hypotheses and discover solutions to problems that have long been considered out of reach. This shift promises to accelerate the pace of discovery across all fields, from pure mathematics and theoretical physics to drug discovery and materials science, changing not just the answers we find but the very way we ask questions.
