I feel like this might explain (and validate?) his conservative stance towards generative LLMs leading to AGI/ASI. What do you think?