Hacker News Reader

Comment by tptacek | Hacker News Reader

I think you'd be crazy to say LLMs are blockchain-style hype when it comes to software development but I don't begrudge anybody who believes they're not currently workable for the kinds of problems they work on; I think reasonable people can disagree about how ready for prime time they are for production software development.

But for math tutoring? If you claim LLM math tutoring is demoware, you're very clearly telling on yourself.

View on HN

I wouldn't trust the LLM's raw output to be correct, but math is provable and if there was a filter between the LLM's output (which would be in some more rigid/structured format, not free form text) and whatever the user sees that tries to prove the LLM's output is correct (and try again if it goes wrong[0]), then i can see LLMs being perfectly fine for that.

In fact i'd say in general anything that LLMs produce that can be "statically checked" in some way, can be fine to rely on. You most likely need more than a chat interface though, but i think in general it is plausible for such solutions to exit.

[0] hopefully it wont end up always failing, ending up in an infinite loop :-P

(OP) In my post, I actually ask the question of whether a student would _want_ to interact with the tutor, not if the tutor is capable of providing good instruction. These are drastically different critiques.

I have seen LLMs fabricate bogus calculations; I personally would be hesitant to use an LLM as my one and only source of math learning, but I suppose using it in conjunction with something like Math Academy mitigates that issue? You've clearly had good success here, but any problem areas with the LLM to watch out for?

Reply on news.ycombinator.com