Determining trustworthiness of LLM responses is like determining who's the most trustworthy person in a room full of sociopaths.

I'd rather play "2 truths and a lie" with a human rather than a LLM any day of the week. So many more cues to look for with humans.