Language models are better than humans at next-token prediction (2212.11281v2)

Abstract: Current LLMs are considered to have sub-human capabilities at natural language tasks like question-answering or writing code. However, LLMs are not trained to perform well at these tasks, they are trained to accurately predict the next token given previous tokes in tokenized text. It is not clear whether LLMs are better or worse than humans at next token prediction. To try to answer this question, we performed two distinct experiments to directly compare humans and LLMs on this front: one measuring top-1 accuracy and the other measuring perplexity. In both experiments, we find humans to be consistently \emph{worse} than even relatively small LLMs like GPT3-Ada at next-token prediction.