Towards a Posthumanist Critique of Large Language Models

Abstract

This article develops a critique of large language models (LLMs) from a posthumanist perspective. The first part focuses on Emily Bender’s critique of LLMs in order to highlight how its conceptual and political axioms have informed recent critiques of ChatGPT. We make a case that this anthropocentric perspective remains insufficient for adequately grasping its conceptual and political consequences. In the second part of the article, we address these shortcomings by proposing a posthumanist critique of LLMs. To formulate this critique, we begin by drawing on Eric Hörl’s contention that the age of digitalization (what he calls “cybernetization”) demands a radical redefinition of the concept of “critique” (Hörl et al., 2021, 7). Relying on Hörl’s intervention, we then gradually develop a posthumanist framework by grounding it in four interlinked concepts: general ecology, machinic agency, machinic surplus value, and cosmotechnics. After advancing the said theoretical framework, our conclusion mobilises it to outline a posthumanist critique of LLMs.