If they’re using stories from Am I the Asshole, there may be some issues with recursion.
I strongly suspect a lot of the recent stories on that subreddit are AI generated. As are the responses. So one LLM is prompted to describe (generic AITA scenario, probably about a racial or sexual minority, a fat person, or a feminist behaving badly, because last I checked that sub was crawling with crypto-fascist propaganda), as if they are in the right. Then the second LLM is asked to determine whether or not the first LLM was actually in the right. And the second LLM will probably agree with the first just because they’re both running on similar algorithms.
And if the AITA data set was pulled from pre-LLM posts, the LLM they tested was probably trained on those AITA posts, which could likewise cause different results than stories it hadn’t been trained on.
Wow, I haven’t thought about AITA in years. I didn’t realize how much better my life was without it 😆
Until now, only the rich and powerful were surrounded by walls of sycophants shielding them from negative feedback. AI democratises this.
For just 14.99 per month you too can simulate being a member of the exploiter class! Truly capitalism breeds innovation!
(or I guess it’s free for now (if you don’t value your data), but you get the idea)
Narcissist’s best friend.
AI stands for Asshole Imbecile.
Maga training👍
Not to defend the AI here, but when AITA is the standard you’re using as your bar? The Catholic Church is less judgemental than AITA.
problem LLM’s have always had since forever but it never gets tackled because nobody funding one thinks that’s a problem.
they don’t see any problems with how it acts to humans or vise versa
They’re putting all their efforts into exacerbating it, because the more agreeable it is the more people “like” it, and the more time they spend with it. It gets RLHF’d into being more and more of a sycophant.
