The Day a Mathematical Giant Fell: AI Solves a 1939 Mystery
While the world was distracted by the spectacle of the World Cup final, a silent revolution occurred in the realm of pure mathematics. An AI model achieved what humans couldn’t for over eight decades: it resolved the Jacobian conjecture, a problem that has tortured the brightest minds since 1939.
The news spread like wildfire through the academic community. By the time Kevin Buzzard of Imperial College London woke up, the result had been verified. By lunchtime, it was the only topic of conversation in the mathematics department. The announcement, posted by Anthropic employee Levant Alpöge, quickly became a viral sensation, garnering over 20 million views on X (formerly Twitter).
What Exactly is the Jacobian Conjecture?
To understand why this is a breakthrough, we have to look at the roots of the problem. Based on the work of 19th-century mathematician Carl Gustav Jacob Jacobi, the conjecture focuses on “maps”—specifically, the conditions under which you can determine a starting input based on a set of outputs.
For 85 years, mathematicians were trapped in a limbo: they couldn’t prove the conjecture true, nor could they find a reason why it was false. That changed on a Sunday afternoon when Alpöge’s AI model provided the answer. By identifying a specific scenario where the determinant remained steady but sent different starting points to the same destination, the AI effectively disproved the conjecture.
Calculation vs. Understanding: The Great Debate
This event marks a pivotal moment in the timeline of AI-driven mathematical breakthroughs. From solving International Mathematical Olympiad problems to disproving Erdős conjectures, AI is moving fast. However, this progress brings a philosophical dilemma: the difference between “how” and “why.”
- The “How” (Calculation): AI can process data and find results with blinding speed, much like a calculator.
- The “Why” (Reasoning): Human mathematicians seek a “story”—a logical chain of reasoning that makes the result intuitive and understandable.
As Akhil Mathew from the University of Chicago noted, while we can verify that the AI’s result is correct, we lack the narrative that traditionally accompanies a mathematical proof. For humans, understanding is about making a concept “fit in your brain,” a feat AI has yet to master.
The Role of Lean and the Future of Proofs
One of the most critical components of this breakthrough is Lean, a specialized computer language used to check proofs. Instead of relying on exhausted PhDs to manually verify hundreds of pages of logic, Lean allows machines to verify accuracy instantly.
Kevin Buzzard suggests that when AI models capable of writing proofs merge with machines capable of checking them, one of the last human advantages in mathematics—the ability to formally certify a discovery—may vanish.
Is the Human Mathematician Obsolete?
With funding for math research declining and PhD admissions dropping, some fear the “death” of the professional mathematician. However, there is a glimmer of hope in the concept of “taste.”
While AI is incredible at finding answers, it is historically abysmal at asking the right questions. The most famous monuments of mathematics are named after the people who posed the problems, not those who solved them. The ability to identify which questions are worth asking—the creative spark of curiosity—remains a uniquely human trait.
Whether we are entering a new age of the “gentleman scientist” or a hybrid era of human-AI collaboration, one thing is certain: the boundary of what is “solvable” has just been pushed further than ever before.
