AI for Math talks by Haocheng Ju and Guoxiong Gao at Stanford

Shurui Liu

Over the past two weeks, Haocheng Ju and Guoxiong Gao gave talks on AI for Math at the Stanford Student Learning Seminar on Mathematics and Computation.

Haocheng and Guoxiong are math PhD students at Peking University, advised by Professor Bin Dong. They are among the pioneers of AI for Math research, working both in natural language (e.g. Matlas, Rethlas, Danus, Pharos) and in formal Lean 4 (e.g. LeanSearch, Archon).

In their talks, they generously shared ideas and lessons from their exploration of AI for Math.

  • Haocheng spoke mostly about semantic representations of mathematics. As a corollary, he showed how the top 4 journals became the top 4 (Annals, JAMS, Acta, Inventiones, IHÉS — and yes, the top 4 consists of five journals), and introduced Rethlas. Slides (PPTX)
  • Guoxiong spoke about why the harness still matters, and introduced Danus and Pharos. Slides (PDF)

Comments