Journalists have uncovered a handful of preprint academic studies with hidden prompts instructing A.I. reviewers to give positive responses
Journalists have uncovered a handful of preprint academic studies with hidden prompts instructing A.I. reviewers to give positive responses
www.smithsonianmag.com
Just a moment...
I'm a bit torn on this one...it's a big f-you for all those low effort "reviews", which I like.
On the other hand, it still is cheating....
In my opinion if someone's using AI to grade the submissions then anyone who games the AI in their submission is 100% blameless.
You can't say that you're going to hire the job candidate with the biggest hat, instead of the qualified one, and then get all bent out of shape when people "cheat" by showing up in big hats. They're literally fulfilling the new criteria that you set. If that puts your criteria in a bad light and gives an advantage to an less qualified person, that's on you.
Yeah, you are probably right...if your reviewer is so lazy, it can be "bribed" with "don't write anything bad about me", it gets what it deserves.
What I'm more concerned is all those LLM-zombies which are listening to these reviews without knowing they are sloppy LLM-reviews. Should they check their sources? Hell yes! Will anyone do it? Unfortunately, no...
What really makes me sad is that this goes against peer review. This entire situation is depressing. We started with poisoning our environment, but now we are poisoning and destroying the very fabric of human ingenuity and creativity in our endeavours. All for chasing some abstract economic valuation.