Google DeepMind Launches AI for Math' Initiative to Accelerate Discovery

Google DeepMind, in partnership with Google.org, has introduced the AI for Math Initiative, a global effort aimed at advancing mathematical research through artificial intelligence.

The initiative unites five prestigious institutions—Imperial College London, Institute for Advanced Study (Princeton), Institut des Hautes Études Scientifiques (IHES, France), Simons Institute for the Theory of Computing (UC Berkeley), and Tata Institute of Fundamental Research (TIFR).

It aims to identify the next generation of mathematical problems suited for AI-driven insights, build the infrastructure and tools needed for exploration, and accelerate discovery across science and engineering.

As part of the programme, Google is providing participating institutions with access to advanced systems such as Gemini Deep Think, AlphaEvolve, and AlphaProof.

Google noted that in 2024, its AlphaGeometry and AlphaProof systems reached a silver-medal standard at the International Mathematical Olympiad (IMO). More recently, the company’s Gemini model, enhanced with the Deep Think framework, achieved gold-medal level performance, flawlessly solving five of six problems and earning a total of 35 points.

“Our belief is that AI can serve as a powerful collaborator with mathematicians, augmenting creativity and accelerating discovery,” wrote Pushmeet Kohli, VP of Science and Strategic Initiatives at DeepMind, and Eugénie Rives, Senior Director of GenAI Strategy.

The initiative signals a major pivot: moving beyond AI as a tool for incremental tasks, toward co-developing new knowledge in mathematics and science.

Earlier this year, rival OpenAI announced the launch of OpenAI for Science, a new initiative aimed at creating an AI-powered platform to accelerate scientific discovery.