We often talk about LLMs as if they are doing something remarkably close to human intelligence. My take is that it is the conversation with the human that creates intelligence and meaning

Agree. LLM looks more like a complex semiotic system, and meaning is attributed to the output by humans during interaction; all intelligence is human, since we interpret the output, but LLM operates only at the level of pure symbols, logical-probabilistic output, which however has nothing to do with meaning. Tho have meaning is to have interaction (with the world), cannot be a closed game of signs.