Getting Korea's Narrative Right: AGI Is a Productivity Shock, Not a Justification for Public Compute

A recent set of commentary in Korea has raised concerns that artificial general intelligence (AGI) will eliminate most forms of labor income, deepen inequality, and justify the creation of public compute infrastructure. One example is a short “World Research Trend” brief published by the Taejae Future Consensus Institute, which summarized and interpreted Pascual Restrepo’s new National Bureau of Economic Research paper and then drew sweeping policy conclusions from it.

This is just one research note from a single institution, not a broad policy consensus. But interpretations like this can still influence policymakers if left unexamined, so it is worth clarifying what Restrepo’s model does and does not imply.

Restrepo’s paper is an interesting theoretical construction, yet the way it is being read risks pushing Korean policy discussions toward overly pessimistic conclusions. This is not entirely new for Restrepo. He and his frequent co-author Daron Acemoglu have long advanced a distinctly anti-automation narrative. But much of the Korean commentary goes further by using the model as evidence that labor income will collapse, economic power will concentrate in the hands of compute owners, and Korea should consider public compute infrastructure or even a universal basic income.

Getting the narrative right matters. Starting from the wrong premise leads to bad policy, and AGI is too important an issue for policy to be built on pessimistic or overly literal interpretations of theoretical boundary cases. The problem is that these conclusions rest on a fundamental misreading of what the model shows.

Restrepo’s model rests on an extreme and implausible assumption: AGI can fully automate all tasks across the economy. This is not a realistic trajectory but a theoretical boundary case. In the real world, technology adoption is gradual, uneven, and constrained by organizational capacity, regulation, labor-market frictions, and, of course, limitations in technology itself. Historically, automation has never eliminated net jobs. Jobs evolve, tasks shift, and productivity gains create new labor demand.

Yet some commentary in Korea takes Restrepo’s boundary case seriously enough to generate sweeping conclusions: that labor income will rapidly collapse, that compute owners will dominate the economy, and that the state must intervene by building or nationalizing compute infrastructure. These interpretations go far beyond what the model supports. Leaning into public-ownership narratives would weaken private investment, reduce competition, and slow innovation—the opposite of what Korea needs.

Korea’s vulnerabilities also have little to do with AGI. The country’s long-standing challenges—low SME productivity, slow digital diffusion outside large conglomerates, and rigid labor institutions—existed well before AGI. These structural issues, not theoretical automation scenarios, should be the focus of policy.

A more grounded approach views AGI not as a force that eliminates work but as a productivity shock that can expand economic output, especially in the face of Korea’s aging population and retirement crisis. Policy should emphasize wider AI diffusion, increased private investment in compute, workforce mobility, and more competitive market structures. This direction aligns better with how technological change has historically played out and with Korea’s actual economic needs.

Korean policymakers should also be mindful of how AGI narratives can migrate toward market restrictions. Framing AGI as a threat that requires public compute infrastructure, digital-sovereignty mandates, or state-led AI development creates momentum for policies that limit market access for global technology providers.

This pattern is familiar. Similar narratives in Europe have fueled ex-ante platform rules, data-localization mandates, and public-cloud requirements that ultimately weakened innovation and hurt consumers. These policies did not make Europe more competitive; they made European firms more dependent on fragmented, less capable alternatives. Korea risks drifting in the same direction if AGI is framed as a justification for expanding state ownership rather than strengthening private sector capacity.

AGI pessimism also risks becoming a tool for broader anti–Big Tech efforts. If AGI is described as a threat to work, democracy, or social cohesion, it becomes easier to justify policies that target U.S. cloud and platform companies through excessive obligations or structural restrictions.

Korea benefits from deep alignment with the U.S. technology ecosystem—across cloud, semiconductors, and research partnerships—and maintaining that interoperability is strategically important amid intensifying U.S.–China technology competition. Policies that weaken this alignment would not strengthen Korea’s innovation; they would isolate it and leave Korean firms further behind in the AI era.

Narrative shapes strategy. If Korea adopts an overly pessimistic interpretation of AGI, it is more likely to pursue policies that reduce innovation, isolate its tech economy, and undermine long-term competitiveness. A realistic understanding of AGI—as a productivity accelerator rather than a trigger for labor extinction, and as a supporter of broad-based growth rather than an inequality accelerator—provides a better foundation for policy. The priority should be strengthening Korea’s capacity to adopt and diffuse AI across industries, not preparing for a scenario in which human work disappears and profits soar.