alt Hacker NewsOk, a publisher stakes their reputation on having reliable problem sets. If you used an LLM, you'd have to proof every single problem to make sure there weren't issues with it that would lead to student's having unintended difficulties with them. Yes it "saves time," until a problem is assigned whose "right answer" takes not only far longer than all the others but is ridiculously complex and impossible to grade or complete in any reasonable amount of time.
Switching the problem sets every couple years is a difficult task in and of itself, and it also keeps answers from circulating amongst students, and saves time for the professor who won't have to do the above every single week, they can just pick however many problems they like out of the book for the relevant section.
> Ok, a publisher stakes their reputation on having reliable problem sets.
Who gives a damn about publishers?
They introduce artificial distribution costs; They are not capable of or refuse to fact-check; the copy-editing they provide is well within the capabilities of ai; at best they introduce asset-providers (eg illustrators) to writers.
> Switching the problem sets every couple years is a difficult task in and of itself
Who cares? The western academic system is already laughably bad at weeding out the incapable. Just let people cheat. It's not like people aren't cheating already.