Comment by case

I think the issue is that LLMs can be so confidently and convincingly wrong about anything.

And if we (LLM tool operators) don’t know the subject matter in question, we can’t easily distinguish what they are right or wrong about.