If you ask an AI a stupid question, it will take your premise at face value and answer.
If you ask a person a stupid question, they will challenge your frame or call you a rude word and walk away.
Until AIs start doing that, regardless of the size of the underlying model, there can be no discussion about whether or not they are sentient.
@PavelASamsonov there is a small but significant fraction of humans that will answer questions as asked no matter how stupid...
because they are on the wrong end of a significant power differential.
Which means that we continue to follow Karel Capek's conception of AI as mechanical servant as a template.
@laprice @PavelASamsonov I think it’s also relevant that AI development is conditioned by product development, in which we traditionally want our products to be inoffensive to users / customers (think how much better behaved ppl are in advertisements than IRL, or Siri’s cute canned answers to stupid user requests).
