Levent Alöge, an employee at AI company Anthropic, claims to have used the Claude AI model to find a counterexample to the 87-year-old Jacobian conjecture. This long-standing problem in algebraic geometry concerns the properties of polynomial maps. The scientific community is currently reviewing the claim to determine if the proof holds under rigorous independent verification.
Detailed Coverage
A significant claim has emerged from the field of pure mathematics regarding the Jacobian conjecture, an unsolved problem dating back to 1939. Mathematician Levent Alöge has reported using Anthropic’s AI model, Claude, to identify a potential counterexample to this long-standing theory. If confirmed by the wider mathematical community, this would represent a major milestone in both algebraic geometry and the practical application of artificial intelligence in scientific research.
Understanding the Jacobian Conjecture
First introduced by German mathematician Ott-Heinrich Keller, the Jacobian conjecture is a fundamental question about polynomial maps. It asks whether a polynomial map with a non-zero constant Jacobian determinant—a value derived from partial derivatives—must have a polynomial inverse. Despite decades of effort by mathematicians globally, the conjecture has remained an open question. It is considered one of the most persistent puzzles in the field, as it deals with the core structure of polynomial algebra.
The Role of AI in Mathematical Discovery
Alöge, who is an employee at the AI company Anthropic, shared his findings through social media. He proposed a specific polynomial map involving three variables with a constant Jacobian determinant of negative two. He claims this specific map is not invertible, citing three distinct input points that result in the same output. This finding, if correct, would serve as a direct contradiction to the conjecture’s premise.
While the use of AI in this context is notable, the mathematical community maintains a high standard for such claims. Because the Jacobian conjecture is so deeply embedded in algebraic geometry, any proposed solution or counterexample requires intense scrutiny and peer-reviewed verification. Independent experts must now replicate the findings and ensure that the logic holds without errors.
AI as a Scientific Research Assistant
This event highlights a broader shift in how scientists use technology. Advanced AI models are increasingly being tested for their ability to assist in complex problem-solving, such as pattern recognition and hypothesis testing. While some observers are optimistic about the potential for AI to accelerate mathematical breakthroughs, many academics emphasize that human oversight remains essential for validating complex proofs. Investors and researchers alike are watching to see if this incident marks a reliable expansion of AI’s role in scientific discovery or if it serves as a reminder of the limitations of AI-generated logical arguments in formal mathematics.