Ask HN: Stochastic mirrors rather than stochastic parrots?

Would it be more helpful to people if we described generative AI as a “stochastic mirror” - an approximate reflection of reality - rather than as a “stochastic parrot”?

This term perhaps encompasses a broader range of Generative AI models (image, video, etc).

It might also help people to understand why well-generated text may still not be an accurate description of reality. The generated text may be linguistically excellent (syntactically correct) and yet still not be an accurate reflection of anyone’s reality (semantically wrong).

Has anyone else come up with this alternative term before? I could only find some algorithm using the term which isn’t the same as what I mean.