First, I agree with your assessment that AI is not "alive" in the way we commonly understand the concept. For me a big part of that concept is agency or the ability to self-determine future behavior. AI only exists to the extent humans allow it to do so. Thus we can conclude it has little to no agency. Second, AI is at base a set of probability structures that as Mr. Smith writes (correctly I think) that reflect human culture. That image in the mirror is me, but it also isn't me. I'd argue it's more not me than it is me. At this point, I'd argue that AI for the most part walks like a duck, talks like a duck, and maybe even replicates other features of "duckness" but fails the duck equivalent of the Turing test. All that said, when I was doing research for my dissertation I worked with agent based models. Many times agents in the models would behave individually and collectively in ways for which there were no instructions. This was also a manifestation of a probability structure but it certainly looked like they were learning and adjusting based on the new information, and in that way capable of a primitive Bayesian action - but still not alive.