Google DeepMind and Fields Medalist Terence Tao Detail AI-Powered Mathematical Discoveries with AlphaEvolve

Summary

Google DeepMind’s AlphaEvolve AI is accelerating mathematical research at an unprecedented scale, according to a new paper published this week by collaborators including renowned mathematician Terence Tao. The research shows how the AI agent tackled 67 challenging problems, rediscovering top solutions and finding novel constructions for several long-standing challenges. The work demonstrates a powerful new method for human-AI collaboration in pure mathematics. It uses AI’s ability to search vast problem spaces to generate insights that complement human intuition, potentially speeding up the path to solving famously difficult conjectures. An Evolutionary Engine for Mathematical Discovery Operating differently from general-purpose chatbots that often struggle with logical rigor, AlphaEvolve uses a structured evolutionary framework. It functions as a “generic evolutionary coding agent,” employing large language models like Gemini to propose, test, and iteratively refine algorithmic solutions. This work builds on DeepMind’s initial unveiling of the tool in May 2025. According to the research paper, “…AlphaEvolve is a powerful new tool for mathematical discovery, capable of exploring vast search spaces to solve complex optimization problems at scale.” In a detailed blog post, Tao explained the AI’s core method involves evolving Python code that searches for a solution, rather than creating the mathematical object directly. This “search mode” allows a single, slow LLM call to trigger a massive, cheap computation as the generated search heuristic explores millions of possibilities on its own. A contrasting “generalizer mode” tasks the AI with finding formulas that work for any given number, aiming for broader applicability. Getting started on new lines of inquiry is remarkably efficient with this process. The researchers highlight that for many of the problems they explored, “…on average the usual preparation time for the setup of a problem using AlphaEvolve took only up to a few hours.” Such rapid setup allows mathematicians to systematically investigate large classes of problems that would otherwise require extensive, bespoke computational work. From Moving Sofas to Kakeya Sets: AI Tackles Open Problems While the system successfully rediscovered known solutions for most of the 67 problems, its most significant contributions came from finding novel approaches. Th…