For years, the mathematical community regarded large language models with a healthy dose of skepticism. In a field defined by absolute rigor, the "hallucinations" and logical lapses of generative AI were seen as disqualifying. But that dismissive stance is rapidly evolving into a collaborative one. What began as an experiment with mathematical puzzles has transitioned into a legitimate tool for breaking new ground, turning months of manual labor into days of algorithmic processing.
Terence Tao, a Fields Medalist at UCLA, has identified 2025 as the year AI crossed the threshold into genuine utility for the discipline. The shift isn't characterized by a single, world-shaking breakthrough, but rather by a steady stream of professional-grade results. Mathematicians are now finding that these models can help navigate complex proofs and suggest novel strategies that human intuition might overlook.
The integration of AI into the field takes several forms. In some instances, mathematicians engage in iterative dialogues with models like ChatGPT or Gemini to refine proof strategies. In more autonomous cases, specialized algorithms are now capable of formulating conjectures, executing proofs, and verifying their own logic with minimal human oversight. While the human element remains central to defining the questions worth asking, the machinery of the answer is increasingly digital.
With reporting from 3 Quarks Daily.
Source · 3 Quarks Daily
