this claim around creativity is too strong, IMHO. truly novel ideas from LLMs are surely rare but i don’t think that any math proves they are impossible. note that the objective function and the outputs are not the same.
Gary Marcus argues math doesn’t disprove LLM novelty
By
–