Large language models have learned to use language

TL;DR

This paper discusses how large language models have acquired language use capabilities, suggesting that new evaluation methods are needed to understand their language knowledge and implications for language science.

Contribution

It highlights the importance of reevaluating language knowledge assessment in large language models and explores their potential to advance language science.

Findings

Large language models have learned to use language effectively.

Traditional evaluation methods may no longer be sufficient.

A new paradigm for understanding language models is needed.

Abstract

Acknowledging that large language models have learned to use language can open doors to breakthrough language science. Achieving these breakthroughs may require abandoning some long-held ideas about how language knowledge is evaluated and reckoning with the difficult fact that we have entered a post-Turing test era.