Exactly, They’re just probabilistic models. LLMs are just outputting something that statistically could be what comes next. But that statistical process does not capture any real meaning or conceptualization, just vague associations of when words are likely to show up, and what order they’re likely to show up in.