Language models are better than humans at next-token prediction
This addresses a fundamental question in AI about the core training objective of language models, though it is incremental as it focuses on a specific, narrow task.
The paper tackles the problem of comparing human and language model performance on next-token prediction, finding that humans are consistently worse than even small models like GPT3-Ada in both top-1 accuracy and perplexity experiments.
Current language models are considered to have sub-human capabilities at natural language tasks like question-answering or writing code. However, language models 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 language models 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 language models 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 language models like GPT3-Ada at next-token prediction.