Comment by jebarker

It seems like a poor fit to me precisely because correctness is boolean, difficult to measure and getting it wrong is very bad. I do think there's a place for AI here but it's probably not LLMs in their current form.