From the Linguistic Turn to the Probabilistic Turn: Ontological Conditions of Meaning, Normative Sedimentation, and the Dual Status of Artificial Intelligence

Large language models (LLMs) produce socially intelligible meaning without lived semantics or intentional comprehension. A growing body of philosophical literature interprets this fact as a vindication of the linguistic turn: if meaning is use, systems that master use vindicate Wittgenstein; if conceptual content is inferential, systems that navigate inferences corroborate Brandom. This article argues that the vindication is real but incomplete. LLMs do not merely confirm that meaning is use; they reveal that the efficacy of use depends on a probabilistic infrastructure of stabilization that the linguistic turn never thematized. To articulate this infrastructure, the article introduces the concept of plausibility equilibrium—the mechanism whereby patterns of use acquire sufficient probabilistic weight to function as bearers of meaning and social significance with partial independence from lived experience—and shows that this infrastructure encounters a constitutive limit: it cannot generate genuine normativity. The concept of the normative exterior—the active residue of historical sedimentations that enter into tension with new equilibria—addresses this limit without recourse to transcendent foundations. The article concludes by examining the political consequences of this displacement: the possibility of a plausocracy, a regime in which plausibility progressively supplants justification as the criterion of social legitimacy.