The Political-Economic Risks of AI

The political and economic risks of artificial intelligence have been over shadowed by fears of malicious superintelligence and killer robots. Due to AI’s distinctive features—automation of cognitive tasks, global scalability, general-purpose technology, and importance to national security—its impact could be unlike earlier rounds of automation. It is possible that AI creates a superabundant world with unprecedented human freedom. In this essay, however, I will explore a tail risk in which human-level artificial general intelligence (AGI) radically concentrates the global economy, breaks democratic and egalitarian institutions, and tears the social fabric, collapsing human productivity. The closest precedent would be the cultural devastation of indigenous societies by colonialism. I will describe how this process might unfold and propose measures to ensure AI has widespread benefits. Competition policy emerges as a critical tool, as do adaptive changes to political institutions. Without appropriate measures, there may be no AI-driven growth take-off and the inequality that emerges would dwarf anything experienced to date.