“The Large Language Model framework lacks a mechanism for feedback on whether its output is correct because there is no objective definition of the 'right thing to say'.”