Hinton (Nobel Prize winner, "Godfather of AI", key contributor to training functions) recently gave a talk at the UK Royal Institute.

Possibly the most controversial and interesting assertion he makes is that, by his analysis, current multimodel LLMs do in fact have subjective experience -- and this may be a separable idea from the idea of "consciousness", ie, LLMs have a subjective experience, but not consciousness.

(hopefully the link goes to 36:33, where he talks about this)