Why transformers are obviously good models of language

TL;DR

Transformers demonstrate superior performance in modeling language, and their architecture aligns with certain linguistic theories, suggesting they may be the most effective current models for understanding language.

Contribution

This paper highlights the theoretical connections between transformer architecture and linguistic theories, advocating for increased scrutiny of these approaches in linguistics.

Findings

Transformers outperform alternative models in language tasks.

Empirical success supports the relevance of linguistic theories embodied by transformers.

Encourages linguistics community to evaluate transformer-based theories more critically.

Abstract

Nobody knows how language works, but many theories abound. Transformers are a class of neural networks that process language automatically with more success than alternatives, both those based on neural computations and those that rely on other (e.g. more symbolic) mechanisms. Here, I highlight direct connections between the transformer architecture and certain theoretical perspectives on language. The empirical success of transformers relative to alternative models provides circumstantial evidence that the linguistic approaches that transformers embody should be, at least, evaluated with greater scrutiny by the linguistics community and, at best, considered to be the currently best available theories.