Ugh , I agree. I would imagine this all comes from fine tuning, or RLHF, whatever is used. I’d bet LLMs trained on the internet without the final “tweaking” s

Ugh , I agree. I would imagine this all comes from fine tuning, or RLHF, whatever is used. I’d bet LLMs trained on the internet without the final “tweaking” steps would roast most of my questions … which is exactly what I want when I’m wrong without realizing it.