> In effect this is how humans respond to prompts no? What's the difference between this and sentience?

An interesting line of question & open research is if we statistically learn similarly - why do we know "what we don't know" & LM cannot. If this isn't working, we probably need better knowledge models