All definitions are arbitrary if you're unwilling to couch them in human experience, because humans are the ones defining. And my main difference is right there in my initial response: an LLM is a stateless function. At best, it is a snapshot of a human brain simulated on a computer, but at no point could it learn something new once deployed. This is the MOST CHARITABLE interpretation of which I don't even concede in reality, it is not even a snapshot of a brain.