AI Hegemony Is a 'War of Computation and Data'... "Without an Independent Model, Subordination Awaits"

Jaewook Lee, Director of SNU AI Research Institute, Proposes Solutions for Securing AI Talent

"Investment in Independent Foundation Models Needed Beyond Economic Returns"

"Large Quantities of Tokens Should Be Provided Alongside Salaries"

Lee Jaewook, Director of the AI Research Institute at Seoul National University, is giving a lecture at the Seoul National University AI Executive Program on the 24th. Photo by Paek Jongmin, Tech Specialist

He diagnosed the development of an independent AI foundation model, currently being promoted under the leadership of the Ministry of Science and ICT, as an essential investment for the future of Korean AI, emphasizing that it should not be evaluated solely based on its economic impact. Business executives proposed offering large quantities of tokens as incentives for AI talent to utilize AI.

Lee Jaewook, Director of the AI Research Institute at Seoul National University, made these points during a lecture at the 'Seoul National University AI Executive Program' on the 24th, stressing the need to prepare for potential regulations on China, which currently leads the open-source AI market with models such as Qwen, Deepseek, and Kimi.

Director Lee noted that while the United States and China are establishing a bipolar structure in AI, China is expanding its influence with an open-source strategy, whereas the United States, led by Big Tech, is strengthening export controls on advanced semiconductors and AI models through closed AI systems. He cautioned, "If Korea relies on external technology without its own independent model in this structure, it may lose its options."

He further explained, "The current AI competition is not simply about technology, but rather about computing resources, data, and sovereignty over the models based on these elements." He diagnosed that national competitiveness will be determined by who can secure more computing power and data, and described the core of AI advancement as the 'scaling law.' He elaborated that the principle where performance improves as more data and computation are invested remains valid and predicted that this trend will continue at least until 2030.

Director Lee also presented solutions for business executives seeking to secure AI talent. Citing NVIDIA CEO Jensen Huang's remark, "If a $500,000 engineer isn't using more than $250,000 worth of tokens, that's a problem," he suggested that companies should consider new compensation systems in the race for talent, not only focusing on salary but also on resources for utilizing AI, especially token usage rights. Tokens are the smallest unit by which AI processes and divides sentences and information. As the use of agentic AI rapidly expands, the use of large quantities of tokens is becoming indispensable.

Director Lee also offered advice on the development of domestic AI chips, such as neural processing units (NPUs). He stated, "Hardware without a software ecosystem lacks competitiveness. This is the greatest Achilles' heel of the Korean AI industry," insisting that a software stack and developer ecosystem must be established together.

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