Between Artifice and Enlightenment: Queer Critique, Insecurity, and the AI Large Language Model

Abstract

Over several decades, machine learning systems—often described with the double metaphor “artificial intelligence”—have become indispensable computational tools in the flow and mediation of economic and social capital. These systems can, on the surface, appear to deepen the dialectic of enlightenment identified by Horkheimer and Adorno by enabling ever greater domination over natural systems and human agency. They therefore generate widespread concern that computational advances endanger the security and autonomy of human subjects, especially in the sphere of production and exchange. Contemporary social theory offers conceptual tools that go beyond received categories of domination and autonomy, however, to provide a reanalysis of machine learning systems and their implication in social dynamics and economic relations. This chapter focuses on the ways large language model (LLM) systems demonstrate complex affinities with theories of mimesis and simulacrum, and therefore how such systems and require a rethinking of performativity and artifice. Recent queer critique—in which reflection on computational systems and simulacra is proliferating—and contemporary social theory, when read together, thus reveal how rhetoric about LLMs demonstrates slippages between agency, performance, reproduction, representation, and meaning that must be unpacked historically and conceptually before metaphors of the “artificial” and “intelligence” can carry analytical weight in addressing today’s socioeconomic relations. Ultimately, this theoretical-conceptual reading highlights the stakes of the problem of truth in systems where artifice manifests both socially and computationally.