Writing emerged in Uruk and Sumer around 3000 BCE for administration—tracking grain, debts, and labor. Temples and states controlled it. Over millennia, this state tool became literature, philosophy, and interiority. Mullen argues that AI is externalizing a different cognitive function (attention, maybe reasoning) with the same consolidating effects, but on a timeline measured in years instead of centuries.
The attention externalization frame works well for recommendation algorithms—systems that shape your focus for someone else’s goals, like a one-way mirror. It fits less cleanly for LLMs; you pull from them, they don’t push. Mullen bundles too much under “AI.”
But the state-tool-to-art-form arc is the real gem. Spreadsheets became soul-craft. The Sumerians wanted inventory; they accidentally got empathy for strangers and counterfactual reasoning. What are AI’s secondary effects? We genuinely don’t know.
What’s striking: Mullen isn’t an AI researcher or plugged into that scene. Yet he hits the same 5-year urgency you usually only hear from people inside the labs. Convergent alarm from outside the bubble.
“Writing emerged circa 3000 BCE not to preserve poetry but to solve administrative problems—tracking labor, obligations, and populations for centralized authority. Cuneiform was ‘designed for governance,’ creating asymmetric information systems favoring elites.”
“Society selects for media based on efficiency, but literacy’s civilizational value lies in secondary effects—‘recursive empathy, long-horizon abstraction, disciplined counterfactual reasoning’—that resist monetization.”
“Just as writing externalized memory to centralized institutions, AI now externalizes attention at planetary scale, enabling ‘one-way mirror’ governance through behavioral modeling.”
“The window for redirecting AI from a ‘centralized instrument of emergent power’ to a decentralized commons may be ‘five years—perhaps less.‘”