A Precis of Language Models are not Models of Language

TL;DR

This paper argues that despite their success in NLP tasks, large neural language models do not truly model natural language comprehensively, challenging the notion of AI as a cognitive revolution.

Contribution

It critically examines the limitations of neural language models, highlighting their inability to serve as complete models of natural language understanding.

Findings

Neural language models excel at linguistic tasks but lack comprehensive language understanding.

Current models do not provide a true revolution in understanding cognition.

The paper questions the assumption that AI models reflect human-like language cognition.

Abstract

Natural Language Processing is one of the leading application areas in the current resurgence of Artificial Intelligence, spearheaded by Artificial Neural Networks. We show that despite their many successes at performing linguistic tasks, Large Neural Language Models are ill-suited as comprehensive models of natural language. The wider implication is that, in spite of the often overbearing optimism about AI, modern neural models do not represent a revolution in our understanding of cognition.