Or "not intelligent enough parrot". Even humans can be unintelligent.
Perhaps we need multiple layers? LLM does for us language modeling, may we need some deduction model? Maybe an empirical validator model? Humans after all have such difficulty with truth, how can we blame our technology?
Literal parrots are perfectly viable species, no humans required.
Vs. AI's (good and bad) are no more independent of humanity than Band-Aids and Pop-Tarts. If a flaming Pop-Tart causes a house fire, nobody wastes time blaming the Pop-Tart.
Perhaps we need multiple layers? LLM does for us language modeling, may we need some deduction model? Maybe an empirical validator model? Humans after all have such difficulty with truth, how can we blame our technology?