Policymakers and tech executives frequently draw on atomic-age frameworks to propose nonproliferation regimes, verification bodies, and arms control agreements for artificial intelligence. However, the historical record of nuclear governance is far more muddled than this analogy suggests, having relied heavily on superpower coercion, domestic priority shifts, and improbably good luck during near-miss crises like the Cuban missile crisis.
Furthermore, the nuclear order is currently fraying as bilateral treaties collapse and states expand their arsenals. Crucially, nuclear technology differs fundamentally from AI because fissile materials are physically constrained and traceable, whereas AI components are built by private companies with profit motives across nearly every sector. This technological divergence, combined with the absence of a catastrophic catalyst to generate sustained political pressure, means the atomic blueprint cannot provide a ready-made governance framework for regulating fast-advancing artificial intelligence.
The reliance on atomic-age precedents obscures the foundational difficulty of regulating a dual-use capability developed primarily within commercial sectors rather than state-directed defense laboratories. Unlike the International Atomic Energy Agency's inspections of physical fissile material enrichment facilities, monitoring frontier artificial intelligence requires auditing decentralized data centres and proprietary software updates that move faster than traditional regulatory cycles.
No comments:
Post a Comment