Mathematicians put AI to work on Fermat's last theorem
At an event in London, mathematicians have made unexpectedly fast progress on formalising Fermat's last theorem using AI
At an event in London, mathematicians have made unexpectedly fast progress on formalising Fermat's last theorem using AI This report comes from New S
Read Full Story at New Scientist →Why This Matters
The unexpected intersection of AI and pure mathematics signals a paradigm shift in how complex problems are approached. Fermat’s Last Theorem, long considered a symbol of unassailable mathematical truth, now faces scrutiny not just from human intuition but from algorithmic reasoning—a development that could redefine the boundaries of proof itself. If AI can formalize centuries-old conjectures, it may unlock new pathways for verifying or even generating mathematical truths beyond classical methods.
Background Context
Fermat’s Last Theorem, proposed by Pierre de Fermat in 1637, remained unproven for over 350 years until Andrew Wiles’s landmark proof in 1994—a milestone that relied on deep connections between number theory and modular forms. The theorem’s legacy extends beyond mathematics: it has inspired everything from cryptography to cultural narratives about human ingenuity versus computational power. Meanwhile, AI’s role in formal mathematics has grown rapidly, with tools like Lean and Isabelle now assisting in theorem verification, but this marks one of the first major encroachments into classical, non-algorithmic domains.
What Happens Next
Mathematicians will likely intensify efforts to integrate AI into formal proof systems, potentially accelerating progress on open problems like the Riemann Hypothesis or the Collatz Conjecture. Critics may push back against over-reliance on AI, arguing that human insight remains irreplaceable for conceptual breakthroughs. Meanwhile, the broader scientific community will watch closely for whether AI can not only verify but also *discover* new theorems—a test of its creative potential in abstract reasoning.
Bigger Picture
This development fits into a broader trend of AI encroaching on domains once thought immune to automation, from art to scientific discovery. It also highlights the accelerating democratization of mathematical tools, where code and formal systems become as essential as pencil and paper. As AI systems grow more sophisticated, the line between human-led and machine-led mathematics may blur, raising questions about ownership, credit, and the future of intellectual collaboration.
