ClareNow

The Scientific Reason We Can't Pause AI

The most dangerous assumption in AI safety is that we must control an AI that can outsmart us. The second is that we can control how fast we build it.

Andréa Morris, Contributor Forbes 09 Jun 2026, 12:32 2 min read 7/10

Key Takeaways

The Forbes article identifies two core assumptions in AI safety as dangerous: controllability of superintelligent AI and the ability to slow development.
It argues that competitive pressure among nations and companies (e.g., OpenAI, Google, China's AI labs) creates a race dynamic that makes any global pause unenforceable.
Scientific reasons cited include the unpredictability of emergent capabilities and the impossibility of verifying compliance across decentralized open-source projects.
The author suggests that regulatory efforts like the EU AI Act or the US executive order are too slow to match the pace of algorithmic advances.
The piece concludes that safety research must shift from attempting to pause progress to building adaptive governance that scales with AI capability.

A new article on Forbes argues that the conventional wisdom around AI safety is fatally flawed. The most dangerous assumption, according to the piece, is that humans can control an AI smarter than themselves. The second is that we can control the speed of its development—the scientific reality, it contends, makes a pause nearly impossible. The article, titled 'The Scientific Reason We Can't Pause AI' and authored by Andrea Morris, challenges the very premise of pause calls that have echoed from industry leaders and policymakers. It argues that the dynamics of competitive acceleration—driven by corporate incentives, national pride, and the open-source movement—create an unstoppable momentum. The piece points to the fundamental impossibility of verification: even if a moratorium were agreed upon, there’s no way to enforce it across labs and countries. Moreover, the underlying technology is advancing in unpredictable leaps, making any attempt to freeze progress akin to trying to hold back a river. The analysis reframes the AI safety debate not as a question of whether to pause, but how to navigate an inherently uncontrollable trajectory. The outlook suggests that rather than fighting the acceleration, we must design safety systems that can adapt as fast as the technology itself.

"The most dangerous assumption in AI safety is that we must control an AI that can outsmart us. The second is that we can control how fast we build it."

Frequently Asked Questions

Pausing AI development is hard because of competitive pressure between nations and companies, the decentralized nature of open-source AI, and the inability to verify compliance globally. Scientific unpredictability also means capabilities can emerge faster than any moratorium can adapt.

According to the Forbes article, the first dangerous assumption is that we can control an AI that is smarter than us. The second is that we can control the speed at which we build AI.

The article argues that enforcement is nearly impossible because AI progress is driven by multiple actors—including private labs, state-backed programs, and open-source communities—none of which are subject to a single authority. Verification of compliance would require unprecedented surveillance.

The alternative is to design adaptive safety systems that keep pace with AI progress rather than trying to freeze it. This includes iterative governance, real-time monitoring, and safety techniques that scale automatically.

No, the article takes a neutral tone. It does not advocate for speed; rather it argues that the current trajectory is scientifically difficult to pause and that safety approaches must adapt accordingly.

The scientific reason centers on the emergent and unpredictable nature of AI systems, combined with the competitive dynamics that fuel continuous improvement. No reliable mechanism exists to freeze a field that advances through distributed innovation.

Original source

www.forbes.com

Read original

Discussion

Join the discussion

No comments yet. Be the first to share your thoughts!