Abstract

Contemporary discourse on artificial general intelligence (AGI) increasingly frames advanced intelligence as a singular force capable of resolving large-scale social and political crises through superior prediction, optimization, and control. This article challenges that image. It argues that AGI discourse is often structured by a fantasy of sovereign intelligence: a wish-granting conception of intelligence as a unified authority able to overcome conflict, satisfy collective demands, and relieve political actors of the burdens of judgment—or the fear of an alien mastermind challenging and finally destroying humanity. Against this view, the article develops a critique of sovereign solutionism grounded in crisis theory, political judgment, and recent work on collective and multi-agent AI. First, it argues that crises are not reducible to well-structured problems, because they involve contested ends, legitimacy disputes, and uncertainty about the standards of resolution themselves and require transformations, not solutions. Second, the article argues that recent technical research on collective behavior, multi-agent alignment, and agent societies further weakens the plausibility of AGI as a monolithic crisis-solving entity, while not thereby vindicating a new collective sovereign. Third, it shows that the sovereign image of intelligence is inadequate in six respects: epistemically, no singular standpoint is sufficient to grasp crisis in its entirety; normatively, contested ends cannot be reduced to a single objective function; politically, judgment cannot be legitimately delegated in sovereign form; institutionally, distributed intelligence requires mediation, oversight, and contestation; ontologically, the singular image of intelligence presupposed by AGI discourse is less universal than it appears; and systemically, the decentering of intelligence dissolves the very premise that cognition must culminate in a single sovereign center from which crisis can be surveyed and mastered. The article concludes that a pluralistic philosophy of AI is necessary: one that understands intelligence as situated within plural orders of interpretation, value, responsibility, governance, and technics rather than above them as a final authority. Such a view does not reject AI as such, but rejects the inflation of bounded technical usefulness into a sovereign model of intelligence.