AI for maths research
Contribution to article IMProofBench: Benchmarking AI on Research-Level Mathematical Proof Generation
Extracts of interview with Google DeepMind, after I obtained early access to Gemini Deep Think. Announcement on X by DeepMind, by Demis Hassabis.
Case studies on how to use AI to discover (known yet highly non-trivial) connections between different mathematical structures:
- Slides of Talk at ICMS AI × Mathematics 2026
- Talk at Google Developer Group AI for Science – Australia
- Summary slides
- Calabi-Yaus and the tardis:
- Animation:
For education
Use LLMs to support learning in mathematics: guide.
AI-supported learning: example from game theory.
Reinforcement learning student projects
A group of students I supervised taught a Reinforcement Learning Model to play the snake game:
A group of students I supervised taught a Reinforcement Learning Model to play PegSolitaire (standard board). The model achieved a perfect solution (last peg remaining is in the center):
Michel van Garrel 