Marcus' points routinely fail to pass scrutiny, nobody in the field takes him seriously. If you seek real scientifically interesting LLM criticism, read François Chollet and his Arc AGI series of evals.