Research

The Civilizational AI

From Objective Optimization to Political Development in AI

By David Mark

Contemporary AI alignment treats safety as a constraint problem: train a capable system, then steer it toward approved behaviour through RLHF. This paper argues that this paradigm is structurally inverted. Drawing on political theory, we reframe the challenge: models are not merely trained on data; they are socialized by it. The dominant pretraining corpus constitutes a Hobbesian state of nature, a normatively incoherent environment where truth and falsehood compete solely on frequency, and no sovereign hierarchy arbitrates value.