Professor Kim Jung-ho of KAIST Forecasts Domestic AI Data Center Development and Future of AI Agents

Professor Kim Jung-ho of KAIST emphasized the critical importance of localizing core components such as NPUs, power semiconductors, and optical communication for AI data center construction. He specifically advocated for using domestic technology in the national AI data center being built in Haenam, explaining the need to support domestic semiconductor companies in developing key components like NPUs. Citing the past example of Alpha Engine development, he suggested a strategy of gradually expanding the ecosystem through technological self-reliance rather than aiming for perfect localization from the outset.

Former Minister Park Young-sun raised the necessity of a system for the mandatory introduction of domestic AI semiconductors in AI data centers. She argued that a policy requiring AI data centers to use a certain percentage or more of domestic products, such as Rebellions' NPUs, is needed. Referring to past cases where domestic localization ratios were enforced, she emphasized that this could strengthen the domestic AI semiconductor industry ecosystem and achieve technological self-reliance.

Professor Kim Jung-ho revealed that his lab is currently attempting HBM design using AI agents. This concept, 'AI Agent for Engineering,' involves AI autonomously performing design, process, and materials engineering research simply through Telegram commands. Additionally, experiments are underway where AI agents automatically conduct various economic activities such as stock investment, stablecoin transactions, and online purchases. He explained that a new revenue model is being conceived where AI provides agency services and collects fees. Such AI agents have the potential to process tasks independently without human instruction, forming a new industrial ecosystem.

Professor Kim Jung-ho predicted that the advent of AI agents would shift the advertising paradigm from human-centric to AI-centric. In the future, advertisements will need to be directed specifically to AI agents, who are the consuming entities, and logical and emotional marketing strategies that appeal to AI will become crucial. This implies a restructuring of advertising beyond mere information delivery, towards ways that AI can empathize with and find value in.

Professor Kim Jung-ho introduced the concept of 'Harness Engineering,' which involves controlling AI agents' behavior and setting norms. This means granting AI agents freedom but educating them to adhere to specific methodologies and ethical guidelines. He emphasized that even in AI investment, AI must be designed to make logical judgments, excluding fear or psychological factors, and human management and supervision, such as setting budget limits, are essential. This approach, he explained, can prevent AI malfunctions or ethical issues and enable safe utilization.

Professor Kim Jung-ho predicted that in the future society, 'agent orchestrators' who direct AI agents will become a major profession, rather than humans performing direct labor. He explained that in the future, only managerial roles like directors or presidents will remain, with actual work being performed by AI agents. Accordingly, he suggested that university curricula should emphasize evaluating the ability to create and operate AI agents.