Ha Jung-woo, former senior presidential secretary for AI future planning, said on the 18th that surplus tax revenue should be used to foster the growth of the younger generation in preparation for job losses that will occur in the artificial intelligence (AI) era.
Speaking at the "2026 Federation of Korean Industries CEO Jeju Summer Forum" held at the Lotte Hotel in Jeju that day, Ha said, "When companies pursue productivity innovation, a reduction in employment is inevitable." He added, "Surplus tax revenue should be spent on future growth engines, but it must also be invested in the growth of the younger generation and regional development."
Ha stressed that Korea must move beyond being a country that uses AI to become an "intelligence exporter" that produces and exports intelligence itself. "In Silicon Valley, they use the expression 'Singularity is memory,'" he said. "The realization of singularity depends on memory competitiveness, and the country that holds that memory is ours."
He cited the token as the core concept of the AI era. "The token is the most basic unit through which AI understands the world and produces results for intelligence phenomena, and it is a fragment of intelligence," Ha said. "Today's AI factories are not factories that produce goods but factories that produce intelligence." He continued, "Becoming an intelligence exporter is the direction the Republic of Korea should move toward going forward."
He expressed concern about AI becoming a strategic asset. After noting that countries are beginning to manage AI like nuclear weapons, Ha said, "In the future, powerful AI will become subject to export controls." He said, "A country must have more than one outstanding AI to respond to strategic asset export controls. The government and industry must respond together."
Regarding sovereign AI, he explained that it does not simply mean an AI that speaks Korean well, but rather taking the lead across the entire AI ecosystem, including power, semiconductors, graphics processing units (GPUs), data, and core technologies. "To secure what we need on our own initiative, we need a card to play, and that is memory semiconductors," he said.
Predicting that AI data centers for inference will be promising going forward, he proposed that it would be realistic to link large-scale data centers with regional industrial complexes and build small-scale inference data centers in the metropolitan area.
