Extreme Power Concentration: A Map and Research Directions

A LessWrong researcher is warning of a risk that might seem counterintuitive: even if the alignment problem gets solved and AI reliably does what its operators want, we face another challenge. AI could concentrate power so extremely that governments become unaccountable in ways no state has managed before. Per Pepijn Cobben, two mechanisms drive this. First, AI automates labor. If machines replace human workers across sectors, people lose bargaining power. Capital becomes a substitute for labor, and capital is already highly concentrated. Second, runaway AI research gives whoever achieves a breakthrough decisive economic, military, and cultural advantages. Cobben uses the IEMP model—which breaks power into ideological, economic, military, and political dimensions—to map where concentration happens. Economically, the labor share of GDP, now around fifty-five percent, would fall, shifting wealth and control to capital holders. Militarily, drone or robot armies would be more reliable than soldiers who might refuse orders. Ideologically, AI systems already outperform humans at persuasion and dominate attention through recommender systems. These all feed into political power, which depends on bargaining between capitalists and labor. If labor's leverage erodes, nations could become like oil-rich states: wealthy but undemocratic and unaccountable. Cobben argues this is an important but neglected risk, with far too few researchers working on it.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton