> since Ai doesn't use moving parts

Optional / matter of time, plenty of homebrew projects that link a physical presence and text-to-speech with an LLM.