Mission: Impossible Language Models

TL;DR

This study tests GPT-2's ability to learn synthetic impossible languages, revealing limitations and challenging claims that LLMs can learn languages impossible for humans, thus informing linguistic and cognitive research.

Contribution

The paper introduces a systematic set of synthetic impossible languages and evaluates GPT-2's learning capacity across these, providing empirical evidence on LLMs' limitations.

Findings

GPT-2 struggles with impossible languages compared to English.

Learning impossible languages varies with language complexity.

Results challenge the claim that LLMs can learn all language types.

Abstract

Chomsky and others have very directly claimed that large language models (LLMs) are equally capable of learning languages that are possible and impossible for humans to learn. However, there is very little published experimental evidence to support such a claim. Here, we develop a set of synthetic impossible languages of differing complexity, each designed by systematically altering English data with unnatural word orders and grammar rules. These languages lie on an impossibility continuum: at one end are languages that are inherently impossible, such as random and irreversible shuffles of English words, and on the other, languages that may not be intuitively impossible but are often considered so in linguistics, particularly those with rules based on counting word positions. We report on a wide range of evaluations to assess the capacity of GPT-2 small models to learn these uncontroversially impossible languages, and crucially, we perform these assessments at various stages throughout training to compare the learning process for each language. Our core finding is that GPT-2 struggles to learn impossible languages when compared to English as a control, challenging the core claim. More importantly, we hope our approach opens up a productive line of inquiry in which different LLM architectures are tested on a variety of impossible languages in an effort to learn more about how LLMs can be used as tools for these cognitive and typological investigations.