Linguistics Seminar Talk - Roni Katzir

Title:  Large language models and human linguistic cognition

Abstract 

Several recent publications in cognitive science have made the suggestion that Large Language Models (LLMs) have mastered human linguistic competence and that their doing so challenges arguments that linguists use to support their theories (in particular, the so-called argument from the poverty of the stimulus). Some of this work goes so far as to suggest that LLMs constitute better theories of human linguistic cognition than anything coming out of generative linguistics. I will review this line of work and discuss the role that LLMs might have in linguistics. I will start by noting that the architectures behind current LLMs lack the distinction between competence and performance and between correctness and probability, two fundamental distinctions of human cognition. Moreover, these architectures fail to acquire key aspects of human linguistic knowledge — in fact, they make inductive leaps that are not just non-human-like but would be surprising in any kind of rational learner. Given that LLMs cannot reach or even adequately approximate human linguistic competence they of course cannot serve to explain this competence. These conclusions could have been (and in fact were) predicted on the basis of discoveries in linguistics and broader cognitive science over half a century ago, but the exercise of revisiting these conclusions with current models is constructive: it highlights ways in which artificial neural networks can inform cognitive science by serving as proxies for general-purpose learners. The exercise also suggests ways in which insights from cognitive science might lead to artificial neural networks that learn better and are closer to human linguistic cognition.