My argument is that there's no way of distinguishing "behaving like conscious beings" from "being conscious". It does not matter whether it is "genuinely emulat

My argument is that there's no way of distinguishing "behaving like conscious beings" from "being conscious". It does not matter whether it is "genuinely emulating conscious beings" or not. It's a simple extension from "it mimics human text" to "what if it got so much better we can't tell the difference". If we can't tell the difference, definitionally it may as well be conscious. You can disagree with the axiom that it does mimic human text, but generating text with it looks like a pretty good mimicry to me. You may not agree that it will ever get that good - I would also agree with that. I don't think the proposed scenario is at all likely, certainly not in my lifetime, but if it was, I'd be there lobbying for the large language models to have rights.