Mastodon Feed
Boosted by pixxl:
fionafibration@mathstodon.xyz ("barbie's dream house of leaves") wrote:
Ok I know I wrote (2/2) but actually, there's more! Co-yay!
This doesn't even begin to get into some of the other ways that LLMs are damaging our field. Peer review has become a wasteland of LLM-assisted "review" and LLM slop contributions, and it's a lose/lose as both ends of the process have more and more slop to sift through in order to do what they are supposed to do. I don't have too much experience with this but I hear over and over again how bad it is right now. Nature recently published a study finding that a *majority* of peer reviewers are using LLMs, and the anonymity of review seems to make it nearly impossible to fight.
The effects of LLM usage by students is still very poorly understood but all the anecdotal evidence right now points towards it massively decreasing student understanding & learning and making grading far more difficult. In my personal experience teaching, the students who used LLMs seemed to lack a very basic engagement with the material and would often reproduce superficially correct-looking output from LLMs, while being completely unprepared to participate in or take away anything from a conversation about *why* it is incorrect.
The absolute horror stories I've heard from students I know who Are in courses with LLM-generated course materials seems even *worse*, and I know numerous students who have reported ridiculous and unsustainable rates of errors in lecture slides as well as nonsensical homework & exercises. I don't think it's very easy to learn a new subject when you have to do the academic equivalent of information warfare to verify everything you're being taught.
I don't know what to do about these issues, but I hope that we talk more about it.
(3/3)