Overcoming AI ethics, towards AI realism

Discussions about artificial intelligence invariably include a nod to ethics. AI ethics has permeated the growing discourse surrounding AI, leading to numerous frameworks and principles intended to guide ethical design. This widespread surge in AI discussions, both academic and public, underscores a significant gap in normative political theory, a gap that urgently needs addressing. Although AI ethics as applied moral philosophy has been criticised as decontextualising AI or as outright useless, there remains a profound lack of understanding the proper political normativity of AI. The critique of AI ethics typically focuses only on feasibility concerns or moral harms, approaches that fail to capture the normative sources from which AI as a political phenomenon draws. The result is a depoliticisation of AI, risking further mystification and giving AI providers the means to justify illegitimate power relations. By leveraging the recent realism-moralism debate in normative political theory, I aim to show that the realist tradition can be the unexpected corner from which we can study these consequences and suggest a substantively different approach to AI in future, moving from AI ethics to AI realism.