Talking to AI Like It Matters: Reflecting on Human-AI Interaction — LessWrong