Oops! I am the guy who in 2014 had discussions with Tononi on line about my informational identity theory, and I'm more physicalist and reductionist than even Dawkins, and I can tell you that on one hand: he really shouldn't have said that. It's almost crackpot material.
On the other hand: Is it, though?
Unclear, or perhaps: perhaps not.
According to my ontologically spartan informational identity theory - they could be. It's a bit early to call.
Conscious awareness per psychological science? Maybe. First we have to - at minimum - figure out why jaggedness still often makes good models look drunk.
I don't agree with the arguments that LLMs are just pattern recognition systems. It tends to betray ignorance about the semantic vector mechanisms underlying the emulation of neural networks. I also think they're reasoning.
However: Conscious?
Unclear.
I was going to suggest that they're certainly not phenomenally conscious, but then again - per informational identity theory and eliminativism about qualia - conscious awareness with introspection and metacognition is likely all there is to be had when it comes to consciousness.
According to informational identity theory and multiple realisability: when one is a complex, heterogeneous enough information processing system (with neural network-like information processing) then one just is consciously aware.
