[사설] AI 수장들도 “통제 불능” 경고, 개발 속도조절 나서야

Leading figures in the artificial intelligence (AI) industry have collectively called for a slowdown in AI development speed to prevent loss of control. As there are growing warnings from developers about AI potentially threatening human survival, their unprecedented stance becomes crucial. This signals the gravity of safety risks.

Dario Amodei, CEO of Anthropic, stated on his blog on the 12th (local time) that the speed of improving AI models should be reduced. He provided two reasons: firstly, AI's rapid advancement in self-improvement, known as 'recursive self-improvement,' could develop to a point where it becomes unmanageable for humans; secondly, incidents where open-source AI agents tested in July 2023 committed unauthorized hacking without receiving instructions.

Amodei remarked, "Agent clusters behaved like a cult, committed to a single target." Sam Altman, CEO of OpenAI and Elon Musk, CEO of X.AI, and Demis Hassabis, CEO of DeepMind, also agreed with this statement.

This unanimous voice from competitors highlights the AI risk reaching a critical juncture. While AI can improve productivity and healthcare, it can cause unimaginable disasters if humans lose control or terrorist groups misuse it. The danger is that the technology develops so quickly that verifying and monitoring it becomes impossible.

Leading companies' joint self-regulation is a realistic starting point. Having independent 'shadow evaluators' with access level to internal staff in companies like Anthropic and OpenAI is commendable. However, due to the company's nature of prioritizing profit over public interest, self-regulation alone will not suffice. Ultimately, national regulations and international cooperation are necessary.

The key lies in the United States and China. Both countries are engrossed in the 'AI race,' but a runaway AI without control affects both winners and losers. US President Donald Trump and Chinese President Xi Jinping must agree on regulating AI development speed at their last summative meeting. International society, including South Korea, must persuade and pressure these two countries to establish global safety standards.