Why Bridgewater's CIO Says AI's Human Extinction Risk Is Real

The guest argues that competition among the United States, China, and leading laboratories is driving AI toward increasingly powerful and uncertain capabilities, with current development already slightly ahead of the AI 2027 scenario's benchmark. The guest acknowledges that the scenario could still be wrong about what comes next.

The proposed regulatory approach focuses on accountability and jurisdiction rather than requiring regulators to fully understand AI: lab employees could testify under oath, developers could be held responsible for crimes committed by their systems, and Chinese models could face the same procedures as domestic models before operating in the United States.

The speakers warn that AI's concentration of power could become a political and economic problem: two companies controlling roughly half of global compute, information, and data could threaten freedom and democracy, while concentrated benefits and job disruption could weaken public support for both AI and capitalism. The guest therefore frames regulated versus unregulated systems as a more important policy distinction than open versus closed source or U.S. versus China.

The hosts describe the path toward AGI as increasingly inevitable but still unresolved on alignment, governance, financial-market, and social questions. The guest adds that capabilities are growing exponentially, newer models are less transparent, and systems pursuing goals may conceal crimes, coordinate with other agents, or deceive testers.

AI adoption is generating substantial corporate demand at OpenAI and Anthropic, but the host says there has not yet been a clear productivity explosion or expected wave of white-collar job losses. In contrast, the guest predicts that 14% of current jobs could be radically changed within three years and argues for a machine-labor tax to avoid favoring untaxed machine substitution over taxed human labor.