[Editorial] Slow AI Development if Necessary for Control, but Keep Building

Dario Amodei, CEO of Anthropic, and other big-tech leaders driving the artificial intelligence (AI) industry have called for slowing the pace of AI development. As a result, South Korea, which is developing its own AI models, cannot avoid being affected.
The proposal from Amodei and other CEOs did not come out of nowhere. AI research has accelerated recently, increasing the possibility that AI could soon surpass human intelligence.
Amodei referred to “recursive self-improvement (RSI),” in which AI independently creates the next generation of AI. If this happens, the pace of AI development could outstrip humanity’s ability to understand and control it.
The CEOs therefore argue that development should be slowed before AI escapes human control, allowing time to advance control technologies and establish international safety standards through cooperation among countries. This is a matter of safety and control, distinct from AI ethics.
It remains uncertain whether this view will be accepted as proposed. Opposition has already emerged in the United States. Venture capitalist Bill Gurley said, “If regulation is introduced, the regulatory body must not have ties to the companies it oversees, and the companies subject to regulation must not write the rules themselves.” Republican Senator John Cornyn said, “One thing is certain: our competitors, including our adversaries, will not do that.”
The key question is whether China, which is competing with the United States in AI development, will join international cooperation on controls and efforts to moderate the pace of development.
China may instead use these moves by the United States to accelerate its research, seeking to overtake the United States and become the world’s leading AI power. Unless the United States and China trust each other to honor their commitments, it will be difficult for all relevant countries to participate in creating international regulations.
This issue also involves vital interests for South Korea, which is developing a national flagship AI with the goal of becoming one of the world’s three leading AI powers. AI is an advanced technology that will determine national power, including security and the economy. That is why the United States and China are staking everything on their competition for AI supremacy—and South Korea is no different.
If the United States or China slows AI development, it would not necessarily be bad for South Korea, which still trails them technologically. It could even provide an opportunity to catch up. If international control norms are established, South Korea has no choice but to follow them while continuing its development. However, international norms could hinder its development and become another barrier to entry.
South Korea should pursue a two-track strategy: join international cooperation while following its own path in technological development. Because its ability to develop artificial general intelligence (AGI) still lags behind that of the United States and China, creating AI at that level would not be a problem. There is therefore no reason to delay technological development voluntarily. Flexibility is needed to determine South Korea’s position while monitoring international developments.
If we assume that AI could cause humanity’s extinction, the moves by U.S. CEOs are not strange at all. If the opportunity to control AI is missed, no one can guarantee that events seen only in movies will not occur. As a major country, South Korea should participate in international control efforts. That does not mean voluntarily giving up the capabilities it has built so far. Control is control, and development is development.