How Korea's AI Leaders View the Speed Control Debate: "Development Must Continue, but Risks Must Be Considered"

"Strengthening Internal Reviews Before Launching New Models"

"Preparing for Increased Requirements for Safety Assessment and Verification"

"Focusing on Stable Application of AI in Industrial Sites"

"Examining Both Physical and Economi

Leaders of domestic artificial intelligence (AI) companies concur that it is crucial to establish safeguards when launching new AI models, given that AI development is progressing at an even faster pace than anticipated. However, they predict that the recent debate over slowing down AI development will not hamper the speed of AI research and development (R&D) in Korean companies.

On the 16th, Woo Hyung Lim, Co-Director at LG AI Research, stated, "While there are both concerns and the importance of continued AI technology development due to its rapid progress, I still believe that AI development should go on." Lim emphasized, "We are at a point where we must carefully consider how to utilize AI in a way that does not lead to social problems, and how our ecosystem should adapt." The ‘K-Exaone’ developed by LG AI Research has passed the second phase evaluation for the Independent Foundation Model (Dokpamo) project, organized by the Ministry of Science and ICT, and advanced to the final stage. Regarding the debate surrounding AGI (artificial general intelligence) development, Lim stated, "While global big tech companies are accelerating AGI development, our goal is to develop ‘expert AI’ that addresses real-world challenges." He added, "Although foundation model performance improvements for AGI will continue, our ultimate objective is to create AI models that solve specifically tailored, industry-related challenges."

Honglak Lee, also Co-Director at LG AI Research, commented, "There is no sign internally that we are considering slowing down AI development." He said, "If anything, we are considering acceleration in terms of computation volume and R&D, not deceleration." Lee added, "However, as the interval between AI model releases shortens and we introduce high-performance models, there is a growing consensus that we need to more carefully assess potential risks and cybersecurity issues posed by these models." He noted, "We are closely monitoring the global landscape."

Monitor Whether 'AI Deceleration Theory' Will Limit Opportunities for Latecomers

Seong Nakho, Head of AI Technology at Naver Cloud, which was selected as the lead company for the government’s security-specialized foundation AI development project, stated, "It is important to observe whether the AI deceleration debate could narrow opportunities for countries and companies lagging behind." He pointed out that the gap between US and Chinese AI models has become much smaller, and Korea can continue to narrow the gap through persistent research and investment. He emphasized that it is crucial to avoid a direction that restricts opportunities for latecomers.

Seong remarked, "Dario Amodei, CEO of Anthropic, in his proposal to regulate AI speed, appears to have included not only calls for stronger safety but also the tightening of public semiconductor export controls and maintaining the technological superiority of the US and its allies." He noted, "If only cutting-edge capabilities are regulated and other companies' opportunities are kept open, this could be an opportunity for Korea. However, it is premature to regard this as an official US government policy." He added, "Rather than a comprehensive slowdown, it is more realistic to prepare for increased requirements for safety evaluations and verifications."

Sangwon Park, Head of the AI Business Division at KT, commented, "The faster AI spreads, the more important safety and reliability become." Park noted, "It is now more important to go beyond a simple development speed competition and focus on applying AI stably at industrial sites and delivering tangible results."

Opinions were also raised that the debate offers a chance to examine the physical and economic conditions supporting AI expansion. Experts highlighted that sustainable AI advancement requires synchronized progress in technology capability, human control, and societal resources and costs. Kim Nokwon, CEO of deep learning semiconductor company DeepX, explained, "The speed of AI development must be matched by human abilities to verify and control its risks." Kim emphasized, "If we cannot sufficiently evaluate or manage significant risks, it may be necessary to adjust the pace of technology development and deployment."

As big tech companies proceed with the construction of large-scale AI data centers (AIDC), securing electricity has emerged as a major bottleneck in expanding AI infrastructure, and the burden of carbon emissions and infrastructure investment costs is growing. Supply constraints, from memory semiconductors and transformers to some high-performance electronic substrates and materials, have started to limit growth in other industries, pointing to a need for further coordination. Kim remarked, "A process to check the spread of AI and the pace of investment based on energy and resource efficiency, in addition to safety, is needed." He added, "We should adjust the expansion speed within the limits of available power and supply chains, and increase the proportion of investment in efficient AI technologies that can perform more valuable tasks using the same resources."

This content was produced with the assistance of AI translation services.