Chiang also claimed that LLMs are text-based DeepFakes, I thought that was the more interesting point (maybe harder to argue correctly about).