Language models are better than humans at next-token prediction

Open in new window