I think I've now experienced the full quadfecta of AI slop in my academic life. I've now encountered slop in all of the below situations, in order of occurrence:
- AI slop submissions from students
- AI slop report, peer reviewing a paper I submitted
- AI slop paper submission, that I was asked to peer review
- AI slop report, peer reviewing a paper that I was also asked to peer review. I submitted a handwritten review; I did use AI to help me read and process the paper, but I wrote the review myself. But apparently my co-reviewer decided to submit what appears to be an AI generated report.
Someone asked me: "Does it really matter if it's slop if the content is accurate?"
I replied, "The bigger problem is the slop overload. The referee reports are extremely long, much longer than a typical human written one, and they usually nitpick to death everything, even things that most reviewers think are minor, unsubstantive issues. Moreover, AI writing is often needlessly verbose and jargony, making it hard to read and understand."
I basically explained that the bigger problem isn't that the slop is wrong (even though it sometimes is), it's that the slop is so overwhelming in quantity that we're running out of bandwidth to even evaluate what's right, what's wrong, and what's important and what's not.
@delete in 48 hours
Just have your clanker fix all the trivial stuff to satisfy their clanker. Almost all the readers will be clankers anyway and it’s always best to write with your audience in mind.
Actually, one thing I've found is that if you ask a clanker for feedback it will nitpick endlessly. On judgment calls over trivial matters, it will start going in circles (take this out, put it in), all of which really doesn't affect the paper at all.
You're right about the audience. I think traditional publishing is totally broken, and I don't know if it's worth my time to continue (especially given my lack of social status within the profession).
My wife thinks I should do just enough to get full professorship before deciding to stop, but the benefits of that are pretty minimal. So I'm still weighing.
Also, when I’m reading something that I know is AI-generated, I feel uncomfortable continuing—it’s as if I’m being disrespected, as if I’m being treated like a fool.
I say this sincerely, because I once did the same thing myself—I was basically acting like a bot, posting AI-generated content just to gain some credibility. Thanks to a few people here who brought me back down to earth, I realized the hole I was falling into. Now, I’d rather write something just once a month than churn out a ton of content that shows a lack of respect for the people reading it. After all, I wouldn’t want anyone to do that to me.
I'm glad you realized it! I never knew you as a slop poster. What was your thought process when you decided to post AI generated content? Did you not trust how people would respond to your own voice?
I’ve been seeing a lot of the same as a BIP Editor. Slop submissions, slop replies, now also slop reviews by other people, and tons of pull requests to fix nitty issues in reference implementations that never bothered anyone, because reference implementations are not living code.
The worst is when authors reply to a multi-hour human review with auto generated replies and then instead of addressing just the feedback to the original text, change every single line of the document or outright push a completely new document. How am I now supposed to efficiently evaluate whether the feedback was processed? I would basically have to start from scratch, both to see whether the raised issues were addressed and what new issues got introduced by the rewrite.