Hope we all recognize that no matter how many parameters the current generation of ML models have, they are fundamentally a bunch of dump memories. Nothing new.