It's pretty easy to make a structural/functionalist argument that LLMs, as they currently exist, are very unlikely to maintain a continuous conscious state. They basically wake up, say one word, then die, hundreds of times a second. The flow of information through them has no feedback paths.