The key thing to understand here is that LARGE LANGUAGE MODELS ARE WRONG ABOUT SHIT ALL THE FUCKING TIME.
We used to sarcastically say, "Every time I read a news article about my area of expertise, I'm appalled by everything it gets wrong, but I'm sure the news articles about things that aren't in my area of expertise are all perfect!"
It's no longer news articles people blindly trust, it's LLMs, which can produce bullshit at the speed of light.
…