Off-Centre AI: On Alignment, Antihumanism and AI Ethics

This article examines the legacies of antihumanism within the context of artificial intelligence (AI) research. In response to the ethical risks inherent to the proliferation of quasi-autonomous computational agents, the AI industry, legislative bodies and academia often advocate for a “human-centred” approach that would align AI with human values, desires and goals, ostensibly to make these technologies more transparent and trustworthy. The rhetoric of human-centred AI reveals many of the conceptual limitations that antihumanism had already identified in the foundationalist and subjectivist assumptions of Western philosophy. This article considers some of these challenging limitations while also problematizing antihumanism itself. Through Derrida’s notion of “centre” and an analysis of alignment methodologies, the article demonstrates how attempts to decentre the human often reconstitute new forms of authority. It then examines the possibility of surpassing standard anthropocentric approaches in AI while maintaining a critical philosophical engagement with the structurally necessary yet precarious character of organizing principles.