Self-Correcting Mechanisms

Every information network in history has faced the same vulnerability: inability to recognize its own errors before they become catastrophic. Roman roads transmitted military commands efficiently across three continents—and transmitted plagues with equal efficiency, because the network optimized for speed, not content quality. The medieval Catholic Church built pre-modern Europe's most sophisticated information network but suppressed the self-correcting mechanisms (dissent, questioning, empirical verification) that might have prevented institutional corruption. Twentieth-century totalitarian propaganda achieved information saturation at unprecedented scale, but saturation made them brittle—systems that cannot hear criticism cannot detect failures, and undetected failures accumulate until collapse. Harari places this pattern at the center of his AI analysis: the danger is not that AI is evil but that it is powerful, and powerful information networks lacking self-correction eventually destroy societies depending on them. The solution is not dismantling the network—that option vanished around 2024—but building into its architecture the capacity to detect and respond to failures.