How AI Changes, and Does Not Change, How I Do Mathematics

Summary

In a guest post by Rachel Webb, initially converted from another file format using AI, Terence Tao presents an argument about how large language models may affect mathematical research. The author says AI does not substantially change the standard for mathematical interest: research still involves understanding phenomena and crafting clear, beautiful, and interesting narratives, often connected to applications or major open problems. LLMs do change execution by providing access to many standard lemmas and accelerating parts of the research process, potentially making previously impractical areas worth exploring, although faster routes have tradeoffs. A solution to a famous open problem would not eliminate the value of related questions or partial understanding, the essay argues. The post gives two human reasons to continue doing mathematics even if machines eventually produce more interesting mathematics: people can find the activity intrinsically enjoyable, and mathematics creates shared communities. It warns that AI may tempt researchers to prefer answers over understanding and could weaken collaboration by increasing fear of being scooped or replacing conversations with machines. The author therefore favors using AI in ways that preserve enjoyment, understanding, and contact with other mathematicians.