Learning math with generative AI
The discussion about college-level coursework and generative AI focuses on humanities (or so it seems to me), for obvious reasons:
- More people take more humanities classes than hard-science classes.
- Seeing an LLM produce a reasonable bespoke N-page paper is stunning in a way that seeing a computer be good at math is not.
- When most of your grade comes from an artifact (the paper) and not from an exam, it more obviously matters if those artifacts can be produced at will.
Yet the teaching of mathematics is, it seems to me, far more threatened by generative AI. That hurts to say: I was a math major and loved it. Those classrooms were intimate, focused places, and the average quality of the teaching was very high. My professors all obviously loved math, and they put obvious care into the way they taught it.
LLM chatbots, however, are just so much better than even great lectures, at least for the purpose of learning the sorts of things one learns a solid undergraduate pure-math education. I never properly understood Galois theory; every so often I make another run at it. A chatbot and some motivation is categorically better than any other way to learn this--for me, at least, but I'd be shocked if it doesn't generalize.
Of course one learns much in a math department other than the substance of Galois theory. But generative AI gets much closer to simply replacing excellent math courses than it does to replacing excellent humanities courses.