Pausing AI at human level harder than pausing now, researcher argues

The article presents a debate within AI safety thinking: some argue for pausing AI development immediately, while others believe we should wait until AI reaches human-level capability before restricting it. The researcher responding to this disagreement acknowledges that immediate pauses face real obstacles—international coordination is difficult, and implementation would be challenging even with agreement. However, the core argument is that pausing later would be even harder. The reason centers on economics: once AI systems reach human level, their economic value would be so high that the financial incentive to continue development would intensify dramatically. Beyond the direct incentive, industry opposition through lobbying would likely create additional political and practical barriers to halting progress. This framing suggests that if safety concerns are legitimate, the window for intervention narrows over time rather than widens, because the sunk costs, competitive pressures, and profit motive all compound as capability increases.