Superintelligence and the Triple Risk of an AI Race

URL SCAN: Superintelligence and the Triple Risk of an AI Race
FIRST LINE: # Computer Science > Computers and Society

The Dissection

The paper frames superintelligence as a three-headed strategic hazard: rival defeat, war, and loss of control to superior agents. Its proposed solution is restraint, safety regulation, and cooperation between states. It identifies the accelerator—the AI race—but treats that accelerator as the central disease.

The Core Fallacy

Under the Discontinuity Thesis, the race is not the root cause. It is the competitive expression of P1: once AI becomes cheaper and more capable than human cognition, deployment pressure becomes structural. P2 makes durable human-only economic zones impossible. P3 then destroys the mass employment–wage–consumption circuit.

The paper correctly sees that safety competition can create catastrophic control failures. But it omits the economic discontinuity entirely. Even a perfectly controlled superintelligence would still concentrate productive power in whoever owns compute, energy, models, and deployment infrastructure. Safety may prevent extinction. It does not restore human economic necessity.

Hidden Assumptions

• States can coordinate effectively enough to bind strategic actors.
• Regulation can slow deployment without creating stronger incentives to defect.
• Human institutions remain capable of governing systems vastly more capable than themselves.
• The principal stake is national dominance rather than ownership and control of AI capital.
• Avoiding an arms race preserves the existing economic order.
• Human survival is the decisive threshold; mass loss of productive participation is left unexamined.
• The claim that China is not pursuing a race is asserted in the abstract, not demonstrated there.

Social Function

This is a partial-truth transition-management memo with an element of elite self-exoneration. It elevates the problem into an arms-control dispute between governments, where regulation and diplomacy appear capable of containing it. That framing is useful for reducing catastrophic risk, but it also displaces the harder question: who owns the machine economy after human labor stops being economically necessary?

The paper is not mere copium. Its control-risk warning is real within its own premises. Its failure is narrower and more consequential: it treats safety governance as if it could preserve human agency and the postwar economic circuit.

The Verdict

The paper sees the detonator but not the corpse. It identifies war and loss of control, yet misses the structural death of wage-based capitalism that follows cognitive automation. Cooperation may delay catastrophe; it cannot reverse the Discontinuity. The decisive conflict will not merely be over who reaches superintelligence first, but who owns it—and what becomes of everyone who no longer produces anything the system requires.