Marking the 10th anniversary of the Go match between AlphaGo and Lee Sedol 9-dan, a meaningful announcement was made recently. Google DeepMind plans to establish an "AI Campus" in Seoul. Models such as AlphaFold and AlphaGenome will meet researchers from the Korea Advanced Institute of Science and Technology (KAIST) and Seoul National University and be utilized in life sciences and climate research. This reflects a direction that views AI not as a tool to replace humans but as a partner that produces results together with human researchers. The reason Google chose Korea is clear. Ranking second in the world in the number of paid ChatGPT subscribers and first in "AI innovation density," Korea is an attractive stage where global AI companies want to collaborate.
At this juncture, however, one question needs to be raised. Are we cultivating enough talent to leap forward as an AI powerhouse over the next decade using this stage as a springboard? The rapid increase in the number of AI users is encouraging. But using AI does not automatically nurture AI talent. On the contrary, as research findings continue to suggest that AI use may weaken thinking capacity, potential risks to the cultivation of future AI talent are becoming apparent.
Late last year, a research team at the Massachusetts Institute of Technology (MIT) divided participants into three groups — those who used AI, those who used search engines, and those who relied solely on human thinking — had them write essays, and published the results of measuring their brain waves. Brain connectivity was strongest in the group that used neither AI nor search engines, and weakest in the AI-using group. In particular, when participants who had been using AI wrote without it at the end, their brain activity was weaker than that of participants who had not used AI from the start. The researchers described this outcome as "cognitive debt." A study by Microsoft Research and Carnegie Mellon University issued a similar warning. The researchers found that the higher the trust in generative AI, the less critical thinking takes place. The moment AI does the thinking for us, productivity may appear high, but once we lose the power to judge for ourselves, that productivity does not last long.
Accordingly, the talent we must nurture should not stop at using AI extensively but should become people who think together with AI. They are people who can scrutinize the answers AI produces, pose deeper questions, and catch the context AI misses. Such capabilities, paradoxically, come from the experience of thinking without AI. Only those who have grappled with problems on their own, hit walls, and tried again can doubt the answers AI offers. That is why education in the AI era must design two stages of training together. First, cultivate the ability to think deeply without AI, and build on that the ability to use AI critically. This principle poses different challenges for K-12 education and university education. In K-12 education, when to place AI in students' hands matters. AI should enter as a tool only after students have had enough time to choose words as they write, get stuck in front of problems, and shape their own thoughts while reading books. AI tutors (teaching assistants) and digital textbooks must not become shortcuts that skip over that time.
University education also stands before weighty concerns. Entry-level hiring in several white-collar job categories is already declining. In the past, a natural division of labor existed, with universities teaching general competencies and companies handling on-the-job training at the workplace. Now, however, as AI rapidly replaces the work of junior employees, this ladder is shaking. Therefore, universities must evolve beyond general competency education toward cultivating talent that companies can immediately put to work upon graduation. At the curriculum level, universities must develop the ability to evaluate results produced by AI, redefine problems anew, and make judgments based on expertise in one's own field.
Ten years ago, AlphaGo came to Korea as the shock that "AI beats humans." Ten years later, the agreement the same company has made with Korea is a promise that "AI makes discoveries together with human researchers." What is needed to turn that promise into real results is not only more AI users but people who can think more deeply together with AI. Nurturing such people is precisely what will determine the true competitiveness of Korean AI over the next decade.
