A pause limited to frontier AI development is a bad deal for the U.S. because China is already out-competing America in most other strategic areas like manufacturing and robotics. A pause would allow China to close the gap on semiconductors while the U.S. freezes its one key advantage, resulting in a geopolitically lopsided outcome.
While a government-mandated AI pause would likely crash the stock market due to regulatory uncertainty, a voluntary pause from the labs themselves could be viewed positively. If framed as a move to improve reliability and reduce political risk, the market might see it as a sign of industry maturity, making it a potentially bullish signal.
Contrary to common belief, authoritarian states like China may be more resilient to AI disruption. They can control AI's diffusion, using it for state stability and deploying it through state-aligned corporations. In contrast, open societies like the U.S. face greater instability from the uncontrollable, widespread proliferation of powerful AI capabilities among the citizenry.
The assumption that open-source models will always catch up to the frontier is flawed. Future AI breakthroughs may depend less on the base model and more on sophisticated, proprietary post-training environments (e.g., for life sciences). This creates a moat that open-source projects, lacking access to these integrated systems, cannot easily cross.
Meaningful AI oversight doesn't have to wait for new laws. The executive branch can act now by pressuring frontier labs to grant approved third-party organizations continuous access for oversight and full access for incident investigations. This creates a powerful, low-friction mechanism for independent evaluation that can be implemented immediately.
A significant but overlooked barrier to AI safety is the threat of antitrust litigation. The US government could view industry coordination on pacing development or safety standards as anti-competitive collusion. This risk of legal action, especially from a hostile administration, disincentivizes labs from pursuing crucial, voluntary safety agreements.
For most of the world, the AI age will bring a strange mix of outcomes. Absolute quality of life will likely improve as economic benefits trickle down. However, these nations will become completely disempowered, effectively functioning as vassals to AI superpowers, with no leverage or say in the global trajectory and unable to protect citizens from AI-driven harm.
Europe can secure its place in the AI future through a 'compute for access' deal. By providing favorable conditions for American hyperscalers to build data centers, Europe can demand guaranteed access to the frontier models running on them. This leverage is backed by Europe's control over key semiconductor manufacturing equipment from companies like ASML.
The primary obstacle to a coherent European AI strategy is not money or political will, but a fundamental misunderstanding among policymakers. Many remain skeptical about the true power of US frontier models and overly optimistic about open-source alternatives. This lack of awareness prevents them from taking the necessary, decisive strategic actions.
The challenge 'If you're a doomer, what are your shorts?' is flawed. Shorting the market is pointless if an existential catastrophe means there's no one to pay you. A more logical financial strategy for someone who anticipates AI-driven chaos is to bet on extreme volatility, near-misses, and non-extinction-level disasters along the way.
The prospect of moving AI compute to space creates a hard deadline for the geopolitical strategies of middle powers. Strategies like Europe's 'compute for access' rely on leverage from controlling terrestrial data center construction. Once compute moves to space, this leverage disappears, forcing these countries to develop new long-term plans.
The danger of mass AI surveillance isn't just about privacy; it's about its interaction with the legal system. The US criminal code is built on the assumption of imperfect enforcement. AI-driven surveillance enables near-perfect enforcement of all laws, including outdated or poorly designed ones, which would create a draconian and unlivable society.