President Lee Jae-myung on the 27th met Demis Hassabis, CEO of Google DeepMind and developer of the artificial intelligence (AI) program AlphaGo, and exchanged views on "safeguards for the AGI era." Lee said, "I often use the Gemini program, and sometimes Gemini does things I didn't ask it to do. Is that a kind of bug?" Hassabis said, "When the era of artificial general intelligence (AGI) arrives, we will absolutely need safeguards that we can truly control."
In the afternoon at Cheong Wa Dae, Lee received CEO Hassabis and said, "We cannot know whether AI will move only in a direction that helps improve human welfare, or in a direction that attacks humans or harms world peace." Present at the meeting were Hassabis; Wilson White, Google's vice president and global head of public policy; Yoon Goo, head of Google Korea; as well as Science and Technology Minister Bae Kyung-hoon and Ha Jung-woo, senior secretary for AI future planning, among other aides.
Hassabis, responding to Lee's question, said, "You have raised a truly important topic," adding, "I think AI should be actively used in advancing science and in the medical field. That's the reason I staked a 30-year career in AI research." He also said, "At the same time, there are clearly areas where we need to consider various risks."
On the Gemini program's problems, he said, "If the guidance we provide is not precise, a foundation model can veer slightly in a different direction," and added, "What is extremely important when using and developing AI is that we must equip proper safeguards and guardrails."