The Unbreakable Riddle
Since 1939, a question known as the Jacobian Conjecture has haunted the world of mathematics. Proposed by German mathematician Otto Keller, the problem is deceptively simple to describe but has proven nearly impossible to solve. Imagine a machine that
takes in a set of numbers and, using a specific type of polynomial equation, spits out a new set. The conjecture proposed that if the machine met a certain simple test, you should always be able to reverse the process perfectly—taking the output and figuring out the exact, unique input you started with. For 87 years, mathematicians believed this was likely true, but no one could prove it for every possible case. It was a fundamental question about the nature of functions and reversibility that resisted every attempt at a final answer.
A Modern-Day Breakthrough
The breakthrough didn't happen in a hushed university hall, but with an almost casual post on social media. Levent Alpöge, a Turkish-American mathematician at Harvard University, announced he had found a counterexample that proves the conjecture is false. The twist? He did it in collaboration with an advanced AI model named Claude Fable 5, developed by the company Anthropic. After a friend brought up the problem, Alpöge tasked the AI with exploring it while he was watching the FIFA Club World Cup final. The AI didn't prove the conjecture true; instead, it produced a specific case that violated the long-held assumption, doing what generations of human mathematicians could not.
The Game-Changing Formula
In mathematics, you don't always need a long, complicated proof to make a big statement. Sometimes, all you need is one perfect counterexample. The AI model generated a formula, just 216 characters long, that acts as this definitive counterpunch. This formula describes a process that passes the initial test of the Jacobian Conjecture, suggesting it should be reversible. However, it leads to a situation where three different starting points all lead to the exact same output. If multiple inputs create the same output, you can't work backwards to a single, unique origin. With that single, verifiable example, the 87-year-old conjecture was shown to be false. The finding was quickly reviewed and confirmed by other leading mathematicians, including the celebrated Terence Tao.
A New Era of Discovery
This event is about more than just one problem. It signals a major shift in how scientific and mathematical research might be conducted. The AI wasn't just used as a super-calculator to crunch numbers; it was used as a creative partner to explore complex ideas and find a path that had been missed for decades. This success is seen by many as a powerful demonstration that AI can accelerate discovery by generating novel insights and challenging long-standing assumptions. While some remain cautious, the collaboration between Alpöge and the Fable model is being hailed as a landmark moment—the hardest mathematical problem yet solved with the help of artificial intelligence, opening a new chapter in our quest for knowledge.
