I think at "the writer has no understanding of the language, when the backend clearly does?" you circle back to the argument's main problem - is it possible for the non-human backend to "understand" the language, even though it behaves like it does.