Complex statistical models can be conscious. The reason AIs aren’t conscious is they’re not embodied beings and don’t have any need to develop it to stay alive.
So how being embodied helps consciousness emerge? If it’s just the statistical model that underpins the consciousness, then it doesn’t matter if it’s embodied or not. It doesn’t matter to the algorithm what hardware it runs on.
This seems irreverent to me. If you created an evolutionary algorithm, where these very complex statistical models competed for survival, I don’t see how that would help one of them become conscious.
I agree with that. By "AIs" I mean currently existing ML models we're calling AIs. That's not how we create them.
If we did make them that way, they wouldn't do large-language-model stuff though - what's that got to do with survival? Humans went millions of years without bothering to read that much too.