Is Artificial Intelligence a Bicycle for the Mind?
Time: 2025-02-28
Published By: He Liu
Speaker(s): William Vickery(BICMR)
Time: 15:00-16:00 March 3, 2025
Venue: Room 77201, Jingchunyuan 78, BICMR
Renowned mathematicians Terence Tao and Geordie Williamson have both given lectures and written papers on the potential of AI for pure math research. Basic uses include interacting with commercial chatbot software to retrieve math definitions and even receive guidance on problem solving. More advanced uses involve designing neural networks specifically trained on mathematical objects. Davies, Juhasz, Lackenby and Tomasev [2021] used an algorithm called saliency analysis to identify knot invariants correlated to knot signature. Wagner [2021] constructed counter-examples in graph theory using reinforcement learning. Charton used the advanced transformer architecture underlying GPT and Deepseek to train models that could calculate the eigenvalues of a matrix faster. He and Kim in 2019 created an algorithm that could distinguish algebraic features of groups after being trained on the group multiplication table. What do these tools mean for the working mathematician? We will illustrate the discussion with some practical software demonstrations.
