The AI Revolution in Mathematics: A New Era of Discovery

In the summer of 2025, AI models astonished the world by solving five out of six problems at the International Mathematical Olympiad, a feat that signaled the dawn of a new era in mathematics. Initially, these models were seen as tools for solving known puzzles, but their potential soon became evident as they began to assist mathematicians in discovering and proving new results, a process that once took weeks or months now accomplished in mere days. Prominent mathematicians like Terence Tao and Akshay Venkatesh acknowledge the transformative power of AI, though they caution against losing the essence of mathematical understanding.

AI's impact is not limited to academia; mathematicians are migrating to tech giants and startups, recognizing the synergy between machine learning insights and mathematical precision. The First Proof challenge in early 2026 further showcased AI's prowess, with models solving over half of the research-level questions posed to them, marking a significant milestone in AI's journey through mathematical academia.

The evolution of AI in mathematics didn't happen overnight. Since 2018, efforts by entities like Google DeepMind have been underway, with breakthroughs like AlphaEvolve, which uses genetic algorithms to solve complex problems. This model, developed by Tao and others, demonstrated AI's ability to improve on existing solutions and tackle new challenges, albeit with some limitations and failures.

Despite AI's impressive capabilities, it often makes bizarre errors, leading some to dismiss its usefulness. However, others, like Ernest Ryu, find value in AI's iterative process, using tools like ChatGPT to refine proofs and discover new solutions. Ryu's work on a longstanding optimization problem exemplifies AI's potential to accelerate mathematical discovery.

AI's role in mathematics extends to abstract areas like algebraic combinatorics and algebraic geometry, where it uncovers hidden structures and generates novel proofs. The discovery of hypercubes in permutation groups by AlphaEvolve is a testament to AI's ability to reveal unseen mathematical beauty.

Yet, the rise of AI in mathematics brings challenges, particularly in education and the integrity of mathematical research. The risk of AI-generated nonsense polluting academic journals is real, and educators face the dilemma of how to teach mathematics in an era where AI can solve problems instantly.

Despite these challenges, the consensus is that AI will not replace mathematicians but rather augment their capabilities. The analogy of AI as a jumping robot, capable of scaling walls but not planning long-term strategies, highlights the complementary role AI plays in mathematical exploration.

Looking ahead, the rapid advancement of AI suggests that it will soon generate mathematics surpassing human capabilities. However, the human element, the artistic side of mathematics, remains crucial. As AI continues to evolve, it is hoped that it will help mathematicians uncover mysteries that would otherwise remain unsolved, preserving the beauty and essence of the discipline.