As a university professor, I have followed the rapid advance of artificial intelligence with a mixture of admiration and caution.
In classrooms, research settings and everyday conversation, AI has moved from a distant technology to an immediate presence. Generative models such as ChatGPT and its peers can now write essays, produce images and respond in ways that often appear thoughtful. Such achievements encourage the belief that we are approaching, or perhaps have already achieved, a form of machine intelligence comparable to our own. Yet this impression becomes less convincing once we examine how these systems actually work. For all their power, AI and the human mind operate according to fundamentally different principles, and that difference has important consequences for education.
Modern AI is built to detect patterns in vast quantities of data and to predict the most likely next output. When asked a question, a chatbot does not reason in the human sense of the term. It does not form concepts, weigh truth claims or pursue understanding for its own sake. Rather, it generates responses by calculating probabilities on the basis of training data and model structure. Its outputs may be coherent, useful and at times impressive, but coherence should not be confused with comprehension.
Human cognition proceeds differently. Faced with uncertainty, we do not rush to infer what is statistically likely. We ask why something is the case, what alternatives might exist and how scattered observations can be organized into a meaningful explanation. We test our assumptions, revise them and often arrive at understanding through trial, error and reflection. Curiosity is central to this process. It is one of the conditions that make knowledge possible.
This distinction was articulated clearly by Noam Chomsky in his New York Times essay, “The False Promise of ChatGPT,” where he described large language models as “lumbering statistical engines for pattern matching.” Chomsky’s point is not so much that these systems are limited as that their limitations are structural. The human mind does not simply register patterns; it seeks explanations.
A simple example in his essay helps clarify the point. If an apple is dropped, AI can predict that it will fall. Human intelligence, however, is not satisfied with prediction alone. It asks why the apple falls, and from that question emerge theories of gravity, force and even the structure of the universe. Understanding requires more than successful forecasting. It requires explanation, counterfactual reasoning and judgment. A language model can generate statements about the Earth being flat or spherical with comparable fluency because it processes both as patterns in data. It does not independently distinguish truth from falsehood in the way a reasoning mind must. Nor does it possess moral understanding. For that reason, it cannot substitute for human thought, however useful it may be as an instrument.
The educational implications of this distinction are substantial. If AI excels at delivering quick and plausible answers, we must ask what happens to the habits of mind that develop through questioning. Here the work of Song Hyun-joo, a professor of psychology at Yonsei University, is especially relevant. Her research team found that Korean toddlers ask, on average, 46 questions per hour. This striking figure suggests that inquiry is not incidental to human development. It is one of the primary means by which young minds make sense of the world.
Yet the spread of generative AI may alter this process. According to Song’s findings, children often ask AI more questions than they ask adults, but they also tend to accept its answers with little scrutiny, even when those answers are plainly mistaken. This is a serious concern. AI systems are optimized to respond quickly and persuasively, not to cultivate disciplined inquiry.
A further concern is that AI increasingly anticipates users’ interests before questions are even formed. Recommendation systems and predictive interfaces offer content in advance, turning many users from active seekers of knowledge into passive recipients of preselected information. If this becomes the dominant mode of intellectual engagement, curiosity itself may wither through disuse. For students in particular, the danger is not simply that they will rely on AI for answers, but that they may lose the habit of formulating serious questions in the first place.
Education, therefore, should not approach AI through either uncritical enthusiasm or blanket rejection. The task is to place AI in its proper role: not as a replacement for human judgment, but as a tool that can support human agency in curiosity and discovery.
In the end, the defining strength of human intelligence is not prediction alone, but the capacity to seek meaning. We observe, question, interpret and revise. We are able to recognize the limits of our knowledge and to treat those limits not as obstacles to hide, but as invitations to think further. That is why curiosity remains indispensable. The most important issue before us is not whether AI will become more powerful. It is whether we will preserve the distinctly human habit of asking questions that no machine can ask on our behalf.
Lim Woong
Lim Woong is a professor at the Graduate School of Education at Yonsei University in Seoul. The views expressed here are the writer’s own. — Ed.
khnews@heraldcorp.com
