I learned linear algebra and would discuss it with GPT4. Sometimes I made mistakes, sometimes GPT4 made mistakes. Once GPT4 wouldn't agree with something the

I learned linear algebra and would discuss it with GPT4. Sometimes I made mistakes, sometimes GPT4 made mistakes. Once GPT4 wouldn't agree with something the textbook said. I said "it's in the text book", it said "then the text book must be wrong", "no, you are wrong", "I'm sorry, but that statement is not generally true", "here's the proof"--only after I gave GPT4 the proof did it finally accept the textbook was correct. It was also able to detect subtle mistakes in the proof, and could not be persuaded by a faulty proof. [0] I think the biggest help was just participating in conversations about math, anytime I needed. It made me more engaged, more focused on what the textbook was saying and whether or not the textbook was matching what GPT4 and I had discussed. You know the saying, "the easiest way to get an answer online is to post the wrong answer and start an argument", something like that. Well, that's similar to what GPT4 was doing for me, it would have an engaged discussion, maybe an argument, maybe leave me wondering about something, and that was very motivating when reading the textbook. The textbook still played a central role in my learning. (GPT4 did catch a mistake in the textbook once though.) [0] Here's a previous comment of mine about learning linear algebra from GPT4: https://news.ycombinator.com/item?id=36244561 https://news.ycombinator.com/item?id=36244561