Right - humans encode knowledge in language all the time, but it's certainly not the only way we keep it in our heads. Supposedly Sora is trained to have a bui

Right - humans encode knowledge in language all the time, but it's certainly not the only way we keep it in our heads. Supposedly Sora is trained to have a built-in physical world model that gives it a huge advantage in its video generation abilities. It will be interesting to see what the same approach would give us with something like GPT-4.