작업 소개

MMCA 과천 상설전 «한국근현대미술 Ⅰ»의 전시와 연계하여 박주용 교수, 이미지 감독과 함께한 ‹AI, 잠들어 있는 아트 아카이브를 깨우다-새로운 연결과 해석› 토크 영상을 공유합니다.

이번 토크에서는 AI가 만들어내는 수많은 ‘연결’ 속에서 ‘의미’를 창조하는 인간의 역할에 대해 함께 논의하고, AI로 인해 허물어지는 경계 앞에서 우리가 지켜야 할 예술의 가치에 대해 고민하는 시간을 가졌습니다.

AI와 예술, 그리고 인간의 창의성에 대해 여러분은 어떻게 생각하시나요?
영상에서 그 답을 함께 찾아보세요.

In connection with the exhibition of the MMCA Gwacheon Permanent Exhibition Korean Modern and Contemporary Art I, we share the talk video ‹AI, Awakening the Sleeping Art Archive – New Connections and Interpretations› with Professor Park Juyong and Director Lee MeeJee.

In this talk, we discussed the role of humans in creating “meaning” amid the countless “connections” generated by AI, and took time to reflect on the value of art that we must uphold in the face of boundaries being dismantled by AI.

What are your thoughts on AI, art, and human creativity?
Find the answer together in the video.

Table of Contents

1. Changes Brought by AI
2. Areas AI Cannot Replace and Areas Easily Replaced
3. Human Creativity and AI – An Elegant Connection
4. The Impact on the Cultural and Arts Sector
5. On the Possibilities and Limitations of AI in the Field of Art

1. Changes Brought by AI

Curator Yi Soojung:
As you may already know, I will first briefly introduce our two speakers.

Director Lee MeeJee is the film director of 57STUDIO and the director who produced the introductory video for Korean Modern and Contemporary Art I, currently on view on the third floor. She has continued to work on archive-related projects in collaboration with major art institutions in Korea, including the National Museum of Modern and Contemporary Art, the Seoul Museum of Art, and the Nam June Paik Art Center.

Next, I would like to introduce Professor Park Juyong. He is the author of The Future Is Not Generated and a cultural physicist. After graduating from the Department of Physics at Seoul National University and receiving his Ph.D. in statistical physics from the University of Michigan, Ann Arbor, he served as a researcher at Harvard Medical School and the University of Notre Dame, a visiting professor at the University of Cambridge, and the director of the Post-AI Research Institute at KAIST. He is currently researching the future of science and culture shaped by human creativity at the Graduate School of Culture Technology at KAIST.

I searched for many books related to AI at bookstores. Most of them commonly state that AI will change the world rapidly and in a frightening way, and while reading them I felt a great deal of anxiety and fear. In contrast, The Future Is Not Generated does not focus solely on AI; rather, it traverses science and art as a whole and explores the future of humanity from a humanities perspective. What are we actually expecting from AI, and what are we afraid of? And will AI truly change our world dramatically? I would like to begin with these questions.

In fact, this topic was first proposed to us by Director Lee MeeJee. She looked into various researchers and authors and specifically expressed her wish to speak with Professor Park Juyong. I also understand that she has recently attended many AI-related seminars. Is there anything in particular you reflected on while attending those seminars?

Director Lee MeeJee:
Recently, I have been attending many AI seminars because I wanted to sense the overall direction of the industry. I attended seminars in entertainment, gaming, film and video, traditional culture, and archives. What I felt was that each industry has its own perspective on AI. For example, in the entertainment industry, AI is used with an intense focus on efficiency and cost reduction. Producing effective results relative to opportunity cost seemed to be the key.

In the field of traditional culture and archives, there was significant attention on using AI for tasks that humans cannot accomplish due to physical time limitations—for example, restoration work and the classification, analysis, and statistical processing of vast amounts of archival material. In film and video, interest in AI had already existed long before we began to tangibly feel it. From the creative stage of developing scenarios to post-production and VFX, the scope of AI utilization in film and video is already very broad.

Observing these diverse perspectives, and since I work within the field of visual art, my main question became how AI should be viewed from within that field. I am particularly a practitioner who creates media content based on art archives. From a practical standpoint, AI allows me to approach specialized domains more closely and to try technologies I would not have dared to attempt before. The quality is even continuously improving.

What we learn in art school—since I come from an art school background—includes foundational areas that must be acquired through training. Even in the conceptual domain, there are stages of thinking that require training. AI, however, seems to skip stages that humans must acquire through practice. Of course, the results produced through that shortcut still need to be completed with the trained sensibility we developed in art school in order to be presentable. I personally cannot use AI outputs as they are.

Still, there are areas I would never have dared to attempt before. For example, I do not know how to work in 3D, yet now converting an image into 3D can be done with two clicks. Rendering architectural drawings was something I could not have imagined, but now it is possible. I once tried rendering an exhibition floor plan I had seen in an archive catalogue. There were no surviving photographs, but by processing the drawing through AI, I could imagine how the exhibition might have been installed at the time.

At first, these experiences were simply interesting and fascinating out of curiosity. But when I read Professor Park’s book and encountered the phrase “an elegant connection,” it resonated deeply with me. Unlike many other AI books that focus primarily on technological change, Professor Park’s book centers on the human perspective that observes those changes. For that reason, when this talk was being organized, I expressed my wish to have a conversation with him.

Curator Yi Soojung:
Then I would like to ask Professor Park. The new government speaks of becoming an AI government and staking everything on AI. Many public institutions, including our museum, feel strong pressure to engage with AI and to establish policies around it. To ask directly: will the world truly change that much?

Professor Park Juyong:
It has always been people who do the work—people generate ideas and carry out tasks. When we look at AI on the surface, it speaks like a person and produces images. It is understandable to think that a revolutionary shift is occurring.

However, we need to consider whether it truly operates in the same way as humans, beyond outward resemblance. Can it create something new in the way humans do? To address this, we must consider how AI was made.

When I ask when AI first appeared, many say it was around AlphaGo, about ten years ago. But if we look at history, references to AI-like ideas date back about three thousand years. In Homer’s Iliad and Odyssey, there is a description of doors that open automatically without being pushed—the first recorded automatic door. Homer initially describes such mechanisms as divine rather than human capabilities. He also describes wheeled trays that move on their own—essentially an early vision of autonomous machines. Even then, there was speculation that humans might one day achieve such technology.

That is myth. In the nineteenth century, Joseph Faber created an automaton with an artificial larynx that produced human-like speech. In 1840, people declared that human singers would become obsolete. Yet two hundred years later, that has not occurred. This reminds us to be cautious when predicting total replacement.

Modern AI emerged alongside computers. Initially, it was extremely difficult to make a computer recognize an apple or understand speech. Artificial neural networks were developed, modeled after the brain’s structure. With training on massive datasets, image recognition and speech generation became possible.

However, the path differs from human development. Humans grow through lived experience. AI is trained on enormous quantities of scraped data, repeatedly labeled until it produces correct outputs. Although the result may appear similar—identifying an apple—the path is fundamentally different.

AI performs well in domains with clear answers, such as board games where winning and losing are explicit. But in creative domains—writing, painting—there are no clear criteria of victory. It is generally recognized that AI has not yet reached the depth of lived experience necessary to produce writing that truly moves people.

When humans speak, we embed experience, intention, and judgments about truth. Large language models, by contrast, break sentences into tokens and statistically predict likely combinations. They generate text based on probability, regardless of truth or intention.

Thus, although AI increasingly replicates the outward functions of human activity, its underlying essence remains different. For that reason, we must be cautious about assuming that AI will immediately replace human life or radically transform it. Even now, we remain in a stage that requires careful consideration.

2. Areas AI Cannot Replace and Areas Easily Replaced

Curator Yi Soojung:
You have briefly mentioned areas that AI cannot replace. Could you distinguish more clearly between areas AI may replace in the future and those it may not, and elaborate further?

Professor Park Juyong:
This is a question many people are concerned about. If AI does the work, how will we make a living? But as I mentioned earlier with the example of singers 180 years ago, we must be very cautious about concluding that the world will change entirely because of a single development.

When discussing whether AI will replace something, we must first clarify the question. When a new technology emerges, certain professions may indeed disappear—just as coachmen largely vanished when automobiles became widespread. However, if you compare news articles from three or four years ago with those of today, you will notice that the professions confidently predicted to disappear then are not the same as those predicted now. Even journalists who made confident claims did not truly know.

Let me give an example. Three or four years ago, there was real fear in Seocho-dong, Korea’s legal district. Newspapers claimed that judges, prosecutors, and lawyers would soon disappear because computers would handle everything. Some even argued that since the legal system seemed to favor the wealthy and powerful, AI might make it more equitable. Yet within a year, that fear subsided. People realized that legal work is not simply about fluently interpreting text. It involves understanding circumstances and discerning what a person truly seeks. It is not enough to be trained and perform well on exams; understanding people is essential.

Likewise, since we do not yet fully understand the nature of current AI, confidently declaring replacement is highly likely to be mistaken.

Curator Yi Soojung:
Even children around us joke, half seriously and half anxiously, that they do not need to study English because AI will handle it. We ourselves have joked that if curating is largely about data, perhaps AI could do it better.

In your book, you discuss creativity extensively. You provide scientific explanations and many examples. Could you elaborate further on creativity as presented in The Future Is Not Generated?

Professor Park Juyong:
Before discussing creativity directly, let me return to the idea that AI learns by processing data. Those who visit museums likely enjoy seeing new paintings and reading new texts. But from the perspective of a creator, creation is not about shaking together everything one has learned and producing an average result. It is about making something others have not seen.

There is a significant difference between what AI produces and what humans create.

First, creativity requires novelty. Here, there remains a clear difference between computers and humans. If a computer simply learns from existing data and averages it, that is different from creating something no one has previously seen. For humans, producing what has not yet been seen is central to creativity.

Second, novelty alone is not sufficient. If I, without knowledge of a field, arbitrarily combine elements that no one has seen before, people will not consider it creative. They will simply ask, “What is this?” Creativity requires something new that others can still relate to. It must be new yet resonant.

Where does that come from? Steve Jobs once said that creativity comes from human experience. At first glance, one might say that experience is not entirely new—someone has lived it before. But the point is that each individual’s experience is unique. Although billions of people have lived, no two have had identical experiences. What I experience is something possible for humans, yet new to others.

Creative individuals are often those who remember their experiences well and have accumulated many of them. Only then can they produce something that is both new and understandable—something others can perceive as plausible, useful, or meaningful.

This is precisely what computers lack. They collect data, but that is not personal experience. It is merely the aggregation of others’ experiences. Moreover, since they do not live through experience themselves, they cannot discern what is more meaningful or more human. They calculate averages.

That is why, when we encounter AI-generated creative outputs, we often feel that they are plausible—because they resemble what others have done—but lack something genuinely new or deeply human. That is where the difference emerges.

3. Human Creativity and AI – An Elegant Connection

Curator Yi Soojung:
That was a very clear explanation. Continuing on creativity, Director Lee, as a creator working in the field, you must be experiencing many tangible effects. From your perspective, what aspects of applying AI feel positive, and what concerns do you have as a creator?

Director Lee MeeJee:
As I mentioned earlier, what I feel most strongly is that AI allows me to “sample” specific technical fields. This may be because I work in media. The range of programs I handle tends to be broader than in other genres, so my experience may be somewhat limited in scope. Because of that, rather than judging AI as positive or negative, I am more interested in the perspective from which I choose to view it.

One of the moments when I found AI particularly fascinating was when I felt its structure was very similar to Nam June Paik’s Paik-Abe Synthesizer. In 2023, when I participated as an artist in the exhibition 《다다익선: 즐거운 협연》 at the Gwacheon branch, I had the opportunity to study Nam June Paik’s archive in depth.

The year 1969 is especially interesting. Two video synthesizers were developed that year: Nam June Paik’s Paik-Abe Synthesizer and Stephen Beck’s Video Synthesizer. As an aside, 1969 was also the year of the moon landing, and a time marked by strong anti–Vietnam War sentiment, hippie culture, and anti-war movements. The upheaval and confusion we feel in 2025 likely had parallels then. I believe art reflects its 시대. In that sense, our current discussions about AI may be part of a larger cycle within art history.

Let me explain why I felt AI resembled the Paik-Abe Synthesizer. In the AI tools I have used, one inputs a prompt, passes through a generative stage, and then receives an output. In early AI systems, once the prompt was entered, the user could not intervene during the generation process and simply had to wait for the result. For example, if I input “apples on a plate” and specify a 16:9 ratio, it generates four images of apples. The generation stage functioned as a single step.

Now, however, the input stage has become highly segmented and detailed. Users can intervene in various ways and apply more specific expressions. For instance, if I ask to create an image of a woman holding a phone in a particular space, the system generates it accordingly. There are many tools, and while I do not intend to promote any, some recent systems have been remarkably innovative.

When I considered this, I thought about how Stephen Beck’s video synthesizer was composed purely of electronic signals without a camera. In contrast, the Paik-Abe Synthesizer required input from an external camera, capturing real-world images and transforming them. Stephen Beck’s synthesizer generated images from electronic data alone, while the Paik-Abe Synthesizer distorted and recombined existing reality to produce new meanings.

In that sense, early AI feels closer to Stephen Beck’s system, while current AI feels more like Nam June Paik’s. Historically, we tend to reference the Paik-Abe Synthesizer more often as a creative tool. From that perspective, I find this correspondence interesting, and it is also one reason I value archives.

Nam June Paik left many statements that feel almost prophetic. One of them is: “What is important in art and technology is to humanize technology.” I find this insight profoundly perceptive, especially considering that he said it in the 1960s.

If asked whether AI can replace human creation, I would answer that humans will not allow it to. I do not think humans would permit that. Rather, I believe AI will develop tremendously as a partner in creation and as a powerful creative tool alongside humans.

Curator Yi Soojung:
Not long ago, we held a seminar in Canada on how AI influences video production. The directors who participated described AI as a skilled assistant who answers even at dawn, is kind, and never gets angry—a very friendly companion. Professor, would you like to add anything to what Director Lee has just said?

Professor Park Juyong:
Listening to the explanation of the two synthesizers is quite interesting. One remains purely mechanical, while the other requires human intervention to create something new. You mentioned that technology alone does not become art without human involvement, and we accept that as obvious. Art expresses human emotion and incorporates technique.

Yet in the AI industry, we often hear the opposite—that humans will become unnecessary. When we think about art, what seems self-evident to us—that human presence is essential—is treated as optional or even dispensable in technological discourse. Some engineers genuinely wonder why, if a machine can generate an image from a prompt, humans would still need to intervene.

We have experienced a similar dynamic before. During the pandemic, people were desperate to go outside once restrictions were lifted. Meanwhile, developers of technologies such as the metaverse insisted that we would not need to go out—that we could remain within screens and receive everything digitally. People longed for physical presence, yet technologists assumed what people truly wanted was to stay inside. I feel that this mindset continues.

Humans want to use their hands and pursue what they themselves wish to do. Yet the narrative often suggests that if we remain still, computers will do everything for us. Recently, discussions about AGI—artificial general intelligence—have gone further, claiming superiority over humans in all domains. In the arts especially, humans wish to create what they want to create. But some narratives suggest not merely assistance but a future in which humans have nothing left to contribute compared to machines.

I see this as a significant crisis—not a crisis of humanity, but a crisis revealing how little technology understands humanity. And that narrative is gaining strength.

Because I am a scientist rather than a technologist, I feel we must not ignore these more fundamental aspects. What encourages me, however, is that even amid claims of total replacement, artists continue to seek what remains uniquely human. In that sense, the position of artists becomes even more important. They represent a group that can firmly assert that humans are beings who must do what they themselves wish to do. For that reason, I actually have greater expectations.

Director Lee MeeJee:
Yes, I feel similarly.

Curator Yi Soojung:
Returning to creativity, earlier you introduced the concept of “an elegant connection.” Could we explore that idea further?

Professor Park Juyong:
Earlier I described creativity as something new rooted in human experience. The concept of “an elegant connection” elaborates on that definition.

The question is: where does the new emerge from? Let me give a simple example. We have five fingers on one hand. Suppose that any two fingers can connect. How many possible hand shapes could result? There are ten possible pairs among five fingers. Each pair can either connect or not, which yields 1,024 possible configurations. Even with just five elements, the number of possible connections becomes large.

If we consider both hands—ten fingers—the number of possible configurations rises to 38 trillion. The point is that with a small number of elements, the number of combinations becomes astronomically large.

This suggests that when we create something new, we do not necessarily need to invent something entirely unprecedented. Novelty can arise from recombining what already exists. That realization itself offers hope.

However, not every connection is meaningful. If I arbitrarily combine elements without coherence, no one will call it creative. Among countless combinations, some feel elegant—useful, beautiful, resonant. Creativity lies in identifying those elegant connections that allow others to perceive value or beauty. Occasionally, such connections even transform the world.

Let me offer an example from art history. In Impressionism—often regarded as the beginning of modern art—Claude Monet is well known for his water lilies and Japanese garden. In the medieval period, artists were largely supported by the Catholic Church, and painters depicted biblical scenes, imagining events no one had actually witnessed.

With the arrival of railways in Europe, people began to encounter landscapes previously unseen. Monet began to consider that painting these scenes had value. Rather than depicting subjects commissioned by the Church, he connected what he directly observed with what he painted. He noticed that light and wind changed what the eye perceived—that leaves were not simply green but contained yellows and reds depending on the moment.

What he connected was the visible world and the painted surface. In hindsight, it seems obvious. But once seen, it cannot be unseen. That is why Monet is recognized as a creative figure who initiated something new in art history—because he discovered and articulated an elegant connection.

Curator Yi Soojung:
The Monet works you mentioned will soon open on our first floor gallery. It might be meaningful to see them in person and reflect on what you’ve just described.

Director, you said the idea of “elegant connection” really struck you while reading the book. You also mentioned that it reminded you of a specific artist—Lim Dong-sik. Could you share that connection with us?

Director Lee MeeJee:
When I read the section on “elegant connection,” the first artist who came to mind was Lim Dong-sik. In 2020, I participated as a commissioned artist in the exhibition 《Rise Up: Lim Dong-sik》 at the Seoul Museum of Art. While producing a documentary film about him, I spent time in his studio, observing his working process and closely examining his archive.

He once said during his YATOO activities, “When I throw something into the field, the field throws something back to me.” That single sentence, to me, encapsulates his philosophy.

Lim Dong-sik is almost an archivist. His studio was overflowing with materials—performances, photographs, writings, drawings from his youth. All of these were eventually donated to the archive museum in Pyeongchang. His works today are based on that accumulated archive. He currently focuses on painting.

Was he originally a painter with exceptional technique? No. He began painting because a friend casually suggested, “Why don’t you try drawing?” From that simple suggestion, he started.

If you look at his paintings, they are rendered with extremely fine brushes. He meticulously transcribes what he sees. A single painting takes an enormous amount of time. But for him, that time itself is part of the work.

His paintings contain layers of time. Archive materials reappear within the composition—drawings, lines of a palm, earlier sketches. Sometimes he adds snow in winter, then later removes it when spring comes. The surface builds up in thick layers.

Watching him work, I sensed something almost ascetic. Past and present accumulate within a single canvas.

He lived in Wongol Village and maintained close relationships with the local community. Many paintings emerged from daily exchanges—someone suggesting he draw a tree, and he responding, “Shall I?” Over time, he documented shared moments, seasonal changes, friendships. Small drawings accumulate, then are gathered into larger compositions. Spring, summer, autumn, winter—then those cycles are layered again.

For him, recording daily life is both process and artwork.

Standing beside him, I felt that nature, life, and community are inseparable from his practice. When you see his paintings in person, there is a quiet emotional resonance. They are not technical displays. They are carefully built through time and attention.

This is why I believe that while AI can generate countless combinations, transforming those possibilities into elegance depends on the creator’s lived experience and attitude. The way one approaches the work—how one inhabits time—becomes visible in the result.

When I read about “elegant connection,” he was the first artist I thought of. And just to clarify—he is very much alive and well.

Curator Yi Soojung:
We often hear that AI processes things quickly. But when you look at the production dates of some artworks, they span years—even over a decade. Sometimes a piece begun ten years ago is completed later, and the official date reflects that long duration. That slowness seems unrelated to efficiency, which is often described as AI’s greatest strength. Perhaps artists operate along a completely different trajectory.

Professor Park Juyong:
There are certainly areas in life where efficiency matters. But the philosophy driving much technological development today prioritizes efficiency above all else.

In that framework, one might ask: Why spend years painting when a machine can generate an image in seconds?

But consider this carefully. When Lim Dong-sik paints falling snow, he is not simply filling white space efficiently. He likely experiences the falling snow in the act of painting it. When we interpret the work, we sense that lived moment. AI does not inhabit that space. It can produce one hundred snowflakes instantly and ask, “Doesn’t this look convincing?”

Yet in art, we are not seeking mere convincing resemblance.

There is an example often cited in discussions of AI-generated art: a version of Leonardo da Vinci’s Mona Lisa rendered in the style of Vincent van Gogh’s The Starry Night. Technically impressive, perhaps—but what is missing?

If one assumes that combining two famous works must produce something superior, that logic reveals a misunderstanding of art. Van Gogh did not paint like Leonardo because his artistic intention was entirely different. His brushwork reflected his desire to express movement, turbulence, emotional intensity. Leonardo, by contrast, pursued refinement and serenity.

Art is inseparable from intention. It is rooted in what the artist seeks to express. Simply merging styles because they are popular misunderstands that purpose.

If one believes humans will eventually be surpassed and rendered unnecessary, then faster synthesis appears superior. But what Lim Dong-sik demonstrates is what it means for a human being to make art.

Can an AI replicate not only the visual result but the layered process of time, intention, and lived experience that produces emotional resonance? I remain skeptical. Not because of technical limitations alone, but because the frameworks guiding AI development often do not account for those dimensions at all.

4. Impact on the Cultural and Arts Sector

Curator Yi Soojung:
Perhaps art holds more future and hope than we assume. I believe that if you see the works we mentioned earlier in person, everything we discussed will become much clearer. If you have the opportunity, I encourage you to visit the exhibition—currently, I understand it is on view at the Daejeon Museum of Art storage space.

Let me move to the next question.

As I mentioned earlier, in this era often described as the “AI age” or even the “AI government,” many institutions—particularly cultural and arts institutions—feel increasing pressure to engage with AI in some capacity.

Last year, the head of the National Film Board of Canada, which holds decades of vast video archives, gave a lecture. He explained that they had decided to commit fully to AI. The reason was simple: they possess an enormous volume of video data, and even writing brief descriptions for each item would take archivists an overwhelming amount of time. Even if every staff member worked on it, the task would not be completed. Without that organizational groundwork, the material could not be effectively used.

Their solution was to develop an AI system that analyzes video content automatically, enabling the archive to become accessible and usable.

This situation is not unique. Many institutions deal with archives. Even on a personal level, each of us carries vast amounts of data—photos and videos stored on our phones, accumulated quietly over time.

In this sense, AI appears to offer the possibility of saving time and resources, or making previously unmanageable tasks feasible. From your perspective, Professor, how might AI reshape arts institutions? And particularly from an archival standpoint, what new capacities could it provide?

Professor Park Juyong:
That is a very important point. As I mentioned earlier, AI has strengths and weaknesses. One of its clear strengths is rapid data processing. Tasks that were previously too time-consuming for humans may become manageable.

However, there is something we must be cautious about. Even when AI processes the same dataset, it cannot independently determine what is valuable. It does not inherently understand significance. Humans must teach it what counts as meaningful.

So, while AI may shorten the time needed to search or retrieve information, considerable effort must be invested beforehand in labeling, tagging, and defining categories. For example, if you ask AI to identify works that express “dark human emotions,” the system first needs to be trained on what “dark emotion” means. That requires extensive tagging and conceptual framing by humans. Only under that condition can AI retrieve results effectively.

Another issue is that AI classification systems may inadvertently limit future interpretation. If we delegate classification entirely to AI, we risk losing alternative ways of organizing or understanding the material—ways we have not yet imagined. Once data is structured in a particular way, other forms of access may become more difficult.

There is also a practical consideration: is AI always more efficient? Recent studies suggest that in some workplaces, teams that heavily relied on AI sometimes spent more time completing tasks than those who did not. Although AI produces fluent results quickly, users must carefully verify accuracy. That verification process can negate initial time savings.

Therefore, while AI holds significant potential in archival contexts—particularly for handling overwhelming quantities of data—we must remain attentive to how we train it, how we define value, and what kinds of interpretive frameworks we embed in the system. The speed of retrieval is only meaningful if guided by thoughtful human judgment.

5. On the Possibilities and Limits of AI in the Arts

Curator Yi Soojung:
We often find ourselves planning or proposing projects that involve AI, so I think we need to keep your earlier remarks carefully in mind.

Although the intention may not have been to “use AI” from the outset, the video we created together drew considerable public attention. Director, this project seems to have been a turning point for you in thinking more deeply about AI. Could you share that process as a concrete example?

Director Lee MeeJee:
This project is, in a way, why I am sitting here today.

When curators commission a video, they usually provide the exhibition overview. I am always grateful for that, especially because I often have the opportunity to sit at the same table and discuss ideas during the conceptual phase. Curator Yi also shared the exhibition overview with me, and certain lines immediately stood out.

For reference, I produced the media content shown at the entrance of the 3rd floor exhibition hall for Korean Modern and Contemporary Art I at the MMCA Gwacheon. Among many pages of text, these lines felt highlighted to me:

“To reinterpret modern artists’ works through the perspectives of contemporary creators.”

“To reflect on the lives of modern artists and discover questions that remain valid for creators today.”

“To seek artistic resonance that transcends time through appreciation.”

“To propose diverse modes of viewing.”

Reading this, I felt strongly that the exhibition sought to connect creators and viewers. Large-scale exhibitions of modern or classical works are not so common, and this one runs for nearly two years. That extended duration made the curatorial intent feel even more meaningful.

My assigned site was a long horizontal media panel at the entrance—this is not something I design myself; it is a mission given to me. Standing before that long wall, I felt it resembled a window. It is the threshold viewers pass before encountering the artworks.

So I interpreted it as “a window where gazes linger.” I imagined figures within modern artworks gazing across time toward the viewer. I wanted to create a sense of waiting and mutual observation—a narrowing of emotional and temporal distance. That became the conceptual direction.

The most important principle was to preserve the original aura and historical context of the works. My mission was not to impose my interpretation, but to strengthen the connection between artwork and viewer. Therefore, I approached AI as a tool for subtle mediation rather than transformation.

For example, if you input an image of a mother and child listening to a gramophone into an AI system, it may generate exaggerated movement—the mother suddenly standing, the child spinning energetically. But I repeatedly refined my prompts to restrain such excess.

I wanted only what felt plausible within the painting’s world: the gramophone turning gently, the mother fanning herself slightly, the child’s legs lightly swinging. I aimed for motion that remained within the imaginative horizon already suggested by the original work.

For me, it was important that viewers see the painting first—not think, “AI was used cleverly.” The artwork should come first; the technical mediation should be secondary.

Interestingly, comments on social media reflected this. One viewer wrote that the advantage of AI lies in using it “in moderation”—without causing discomfort. That response made me realize that audiences are perceptive; they sense the creator’s intention. Public literacy regarding AI is higher than we may assume.

However, there are limitations.

For example, Lee Jung-seop’s abstracted human forms were often misinterpreted by AI. It sometimes rendered them as wooden barrels in motion. The system failed to recognize the human abstraction. This reveals a fundamental limitation: AI processes image, form, and composition statistically, but lacks deep contextual knowledge of Korea’s historical and cultural background.

When we speak of developing “our own AI” that learns Korean cultural context, I deeply sympathize. Even in my work, I noticed that if I ask three major generative AI platforms to “draw a Korean woman,” each produces a different archetype. Developers’ cultural assumptions are embedded in the system.

AI is created by humans. Even if it expands autonomously through data, the cultural bias of its training corpus inevitably shapes its outputs. If AI does not understand the historical and cultural context embedded in artworks, it remains limited to surface-level visual manipulation.

Curator Yi Soojung:
In another seminar, a Canadian director shared something similar. She asked AI to “draw a middle-aged woman.” The result resembled an elderly grandmother—contrary to her lived perception of contemporary middle age. When she requested “a woman lying on an operating table,” the AI generated a lightly dressed young Asian woman. She realized that the developers’ implicit assumptions shaped the output. Bias is unavoidable.

Professor Park Juyong:
That example is very telling. Earlier, we discussed intention and context. Without them, AI produces averages.

Imagine a population of fifty million contributing data. There is no reason for the statistical average to resemble any particular individual. The more data you add, paradoxically, the more distant the output may become from personal intention.

So even if AI identifies something “typically Korean,” that typicality may resemble no real person. It becomes abstracted generalization.

In your case, Director, you wrestled with the system because it did not fully grasp your intention. You wanted restraint; it defaulted to spectacle. That tension will likely persist. Even if AI improves, it will never perfectly understand every Korean individual—because individuals are irreducibly different.

The question, then, is not whether AI can approximate cultural norms, but how we, as creators, continue to assert intention, context, and lived specificity within and against those averages.

Curator Yi Soojung:
Today we spoke about what Professor Park called an “elegant connection.” Although we come from different fields and backgrounds, I hope this gathering itself has become a small example of such a connection—one that leaves you with new and perhaps unexpected thoughts.

Please travel home safely. If you have time, I encourage you to visit the third floor to experience the exhibition and the media work together. May I invite our speakers to share a final word?

Director Lee MeeJee:
Thank you for coming all the way here and for listening so attentively throughout this long conversation. I hope we can cross paths again, here or elsewhere.

Professor Park Juyong:
I will leave you with one thought.

Many of you in the arts may already know this, but when you ask people outside the field, they often imagine Picasso as belonging to a distant past. In fact, Pablo Picasso lived long enough to use a computer. And when asked about it, he once remarked, “Computers are useless. They can only give you answers.”

What he meant, I think, is that whether one paints or uses any tool, what matters is the question: What do I want to express, and how should I express it? The tool itself does not determine that.

In that sense, the idea that computers might replace humans is not new. Even in the 1980s, Picasso already understood that the essence of creation lies elsewhere.

As conversations about AI continue to grow louder, I hope we remember that we already know the answer. If we remain attentive to our humanity, there will always be things only humans can do.

Thank you.