‍

Christiane Paul on Defining AI Art

Christiane Paul curated the Whitney Museum of American Art's Marina Zurkow: Parting Worlds / Hyundai Terrace Commission: Marina Zurkow. Peter Bauman (Monk Antony) spoke with Paul about the exhibition and how Zurkow’s ecologically attuned practice engages with site specificity. Paul also gives a definition of AI art before covering broader questions of art’s pace, technological criticality and the cyclical relationship between institutions and the art market.
‍
Peter Bauman: Your exhibition with Marina Zurkow recently opened at the Whitney. Why did you choose her work as a way to speak to our current moment?‍

‍Christiane Paul: Marina's exhibition, Parting Worlds, was constructed around the annual Hyundai Terrace Commission—the second in the series. When you're working outside on a terrace, you're ultimately also working on a public art piece because an outside location is part of a larger environment. So what was important to us in terms of the selection of the artist was site specificity.

As a first step, I compiled a list of artists. Then every exhibition has to go through a proposal and selection process. The chief curator and director have to greenlight it. We landed on Marina because her work had high potential to do something really interesting with the location of the meatpacking district. She had already created two earlier pieces called Hudson Follies (2022) and Does the River Flow Both Ways? (2022) in collaboration with James Schmitz that showed a view of the Hudson River split into a world above and below the water. It is this software system that Marina built on to create a site-specific, software-driven, hand-drawn animation about the meatpacking district for the Whitney terrace.
‍

‍

Then we decided to extend the exhibition into the Kaufman Gallery, which is adjacent to the terrace and has windows with a view of it. We're showing two more software-based animations of Marina’s in that gallery. One is a collection work called Mesocosm (Wink, TX) (2012).

The other is a new piece called The Earth Eaters (2025), which is a modern fairy tale about the history of mining with visuals based on the aesthetics of woodcuts. They are actually drawn from two books—one  by the founder of geology, Georgius Agricola, and the other by Conrad Gessner, the founder of zoology.
‍

‍

Marina trained an AI model on those woodcuts and generated an array of islands and a menagerie of animals. In the work you’re seeing the Earth being continuously wounded and harmed by an army of ant-like miners. From the holes in the Earth, war machinery appears, clad in the metals for which the Earth is being mined. Particularly in this day and age, I think this work is incredibly resonant.

What attracted me to Marina's work, aside from the potential for site specificity, was how it could speak about our present moment. Marina's work has always been focused on ecologies and ecological systems.
‍

As we are struggling currently with redefining environmental protection in the face of a climate crisis and economic turmoil, awareness of ecologies has become ever more important. 

‍
‍The River is a Circle (2025) on the terrace takes you through six time periods of the meatpacking district’s history, starting with the Lenape. The Whitney is actually built on a former Lenape tobacco field and there was a Lenape trading post nearby.

Then we move through colonial times and the 19th century—through the 1970s and the decay of the piers, when all of the art and nightlife happened there—on to a period of gentrification. Finally, we see two possible futures. It is not a chronological work. The assets representing the different periods appear as what Marina calls “gangs of time,” intersecting with each other in ever new configurations.

The work emphasizes the occurrence of these cyclical social systems, the extinction of species and the effect of trade and economics. I think it's incredibly important to look at these structures today and evaluate the world we're living in through these larger ecological and social systems.

‍Peter Bauman: It shows how software, generative and AI art can engage with the issues of our time rather than just producing novel visuals. Another artist sensitively engaging with issues of our time is Maya Man, whose work, A Realistic Day In My Life Living In New York City, you also recently commissioned for Whitney's On the Hour series. Can you talk about the selection of Man and how the work rethinks net art today?‍

‍Christiane Paul: On the Hour is a relatively new series but it is a continuation of Sunrise/Sunset, which has been running since 2009. Sunrise/Sunset was a 30-second intervention into whitney.org that would play at sunrise and sunset times in New York City. While we loved the series, it gave limited exposure to the artwork, which is why we switched it to On the Hour. Now you can see these works hourly for thirty seconds.
‍

The choice of Maya aligned with the mission for artport on a larger level—same for Sunrise/Sunset and On the Hour. I'm always trying to show the diversity of approaches within the networked environment.

I would say that any work that goes on view on artport redefines what web-based art is at any moment in time, whether it's created by an artist who has been working in that field for a few decades or for a few years. This process of redefinition is ongoing.

Maya's work is particularly interesting to me because her practice is so deeply web-based, as opposed to many artists right now who are broadening their work across different media. Maya is really focused on the web and the exploration of social media platforms.
‍

‍

‍Peter Bauman: You mentioned that in Zurkow’s work, The Earth Eaters, the artist trained an AI model on old images. I’m fascinated by Joanna Zylinska's critique of generative AI in her book AI Art (2020). She likens much of it to spectacle-based art and compares it to a glorified version of Candy Crush. Comparing it to Candy Crush implies superficiality on both the aesthetic and conceptual level—all slop, in other words, to her.‍

‍Her viewpoint spotlights a seeming rift emerging in digital art around AI. As a curator deeply engaged with digital practices, what do you think of Zylinska’s critique? When you're curating work that is telling our digital story, how do you navigate between technological novelty and critical inquiry? And how do you even define AI art?‍

‍Christiane Paul: I very much appreciate Joanna's book and her fantastic work in this area. One thing the book does not do is provide a deeper definition of AI art, as opposed to Martin Zeilinger's book Tactical Entanglements (2021), which includes what I think is an excellent one. By my definition of AI art, I would not even call a lot of what Joanna refers to “AI art.”

Putting a text prompt into an image generator and creating beautiful—or not—visuals does not automatically result in AI art.
‍

I define AI art as art that conceptually, practically and critically engages with AI technologies in a sophisticated way. 

In the media and public discourse, we often conflate any visual output generated by AI with AI art. I do not think that corporate AI models generate art. That is the critical distinction I would make.

I would also not dismiss formal explorations. Generative software art from Casey Reas to LIA has created formally beautiful work, as have many artists in other media. I do not think that this work should be dismissed because of its more formal qualities. There's a huge difference between the work of these artists and what users, in a more casual way, could get out of any of the text-to-image generators.
‍

As a curator, I'm not looking for any specific markers of quality. I evaluate digital art, including AI art, always on the same basis—that is, how does an artist engage with the medium conceptually and practically in a sophisticated way? 

That is the main criterion. I find technical novelty deeply uninteresting. Of course, I do look at what artists are doing with the newest technologies. But I'm not coming to the work thinking about how “it's new. It's the first.” That’s a completely overused yardstick. It's really all about how an artist engages with the technology.

‍Peter Bauman: What I found disappointing about Zylinska’s critique is that she largely collapses a diverse field of deep generative practices into a single critical frame—of Candy Crush. She does express some fascination with the artists she discusses—Gene Kogan, Mike Tyka, Memo Akten and Mario Klingemann—but ultimately portrays their work as complicit in a spectacle-driven, corporately aligned vision of AI art.‍

‍Christiane Paul: Joanna focuses more on examples and not on a close reading of the artists’ complete body of work. Memo Akten’s Learning to See is not Candy Crush. His PhD thesis, in which it is included, critically investigates deep learning models as an artistic medium for new modes of performative, creative expression.

You can always single out a work by an artist to dismiss it. You could even disregard work by the most revered artists on that basis. I could easily find works that are “only” visually beautiful by artists who have done a lot for the digital medium and are deeply critically engaged with it.

Joanna’s book is not an art-historical deep dive into AI art, which wasn't the goal. It was inspired by the 2017 edition of the Ars Electronica Festival in Linz, based on AI. Her critique focuses on the political underpinnings of the current AI debate and the way it feeds into art.
‍

‍

‍Peter Bauman: Your perspective on AI is particularly relevant because it’s been about a year since your show, Harold Cohen: AARON, at the Whitney. It was quite prescient in that since then, the talk of AI has only gotten more intense—louder and more frequent. Where do you stand on these controversies, like copyright, that have simmered for nearly a decade but are coming to a boil in 2025?‍

‍Christiane Paul: With the Cohen show, I wanted to contrast his approach based in symbolic AI with current practices. When it comes to issues of ethics and bias, I'm, of course, deeply concerned—as is every artist working in the field. Honestly, I have not talked to any artist engaging with AI technologies who isn't. It's something that you need to keep in mind and engage with on a level of criticality, highlighting how ethics filter in when it comes to training data in general and the process of training models on contemporary artists' work without permission or compensation.

It is one of the biggest issues we’re facing today and there are many discussions about how to best approach it. I hope that copyright law will more deeply consider all of these issues. Maybe even labor law can address the problems on some level. These are some of the frontiers of AI technologies. And as curators and artists working in the field, I think we're continuously engaging with that.

‍Peter Bauman: You mentioned contrasting Cohen’s symbolic, rules-based AI with current deep learning practices. I was recently speaking with Pindar Van Arman, who knew Cohen, and he mentioned how Cohen strongly distrusted deep learning. There has been this intense, historical rivalry between symbolic AI and neural network-based approaches—going back to the late ‘60s and Minsky.‍

‍Today, computational philosopher AA Cavia argues that the nature of computation itself is changing in the age of deep learning. He notes a shift from the hard-coded symbolic logic of Cohen’s expert systems to deep learning’s subsymbolic computation—where knowledge is no longer binary or human-readable but instead operates through latent spaces.‍

‍Then I spoke recently to Gene Kogan and Tom White, who seemed to imply something similar—that the personality of programming is changing.‍

‍Do you agree that we’re seeing a fundamental shift in the nature of computation or is this another marginal step in a long process?‍

‍Christiane Paul: I do not see it as a shift. That may be because I'm taking more of a bird's-eye perspective on the historical trajectory. Of course, there’s no comparison between today’s neural networks and Cohen’s expert system that gave him much more control over the art program he was creating. Today’s neural networks represent a new era that would not be possible without big data. It’s driven by a statistical approach—just look at how ChatGPT has grown and the trillions of assets that AI systems are trained on today.

I'm not sure how Cavia defines them, but symbolic AI, statistical AI and subsymbolic AI have been main categories of AI for quite some time. For me, there has always been a continuum of these technologies. One thing that was very clear to me early on is that statistical AI alone would be a dead end because it relies on training with existing data sets and is largely driven by ideal data points, merging optimization, et cetera. That can only be a dead end because there is no consideration of context.

If we look at today’s AI, we see—not quite good old-fashioned expert systems but—the idea of knowledge modules returning on a larger scale. To me, this is the more successful approach: combining rule-based reasoning models with statistical ones—a real fusion.

I would say that there have been shifts in programming—but not necessarily computation. When it comes to computation, you can build a trajectory leading to quantum computing and that will be very different. That will be a major shift—a completely different paradigm.

Programming approaches are continuously changing; languages are changing. But we're still operating within the same parameters. We're constantly fusing and redefining boundaries. Today's deep learning is very different from what we have seen before and requires different approaches. But I would say that within the overall parameters of programming, we still see a natural continuation of its foundational principles—an evolutionary process.

‍Peter Bauman: These approaches, even going back to Cohen’s expert systems from the early ‘70s, engage with our most rapidly emerging technologies. Curator and critic Helen Molesworth has challenged digital art by saying it doesn't align with the “speed of art.” She argues art is something intentionally slow—unfolding, hibernating and smuggling meaning across generations.‍

‍She contrasts art’s deliberate pace with the chaos of modern, networked life’s endless notifications—and she contrasts it with digital art.‍

‍Should art be slowing us down? Or should it reflect the pace of life right now? What role does digital expression have if limited to the slowness Molesworth prescribes?‍

‍Christiane Paul: Helen’s premise doesn't quite work for me, although I'm a big fan of hers. She's wonderful and has done amazing work.
‍

I would say the flaw of her argument is that it sets a pace for art. I believe art is setting the pace. The respective medium is setting the pace. 

Certain art media have operated at faster paces for decades—video art of the '60s for example. Artists were going onto the street with their porta-packs and engaging with ideas of instantaneous broadcasting. That's slow by today's standards but it was a fast technology back then.
‍

Digital art is moving incredibly fast and artists working with it have to engage with the technology at that pace. 

That does not eradicate another part of Helen's argument—that art from a bird's-eye view certainly moves slowly and hibernates.
‍

We're definitely engaging with the same topics over centuries—but from different perspectives.

You see ideas dying down or rising up again after decades. On that level, it's true that art history moves at a slow speed. There is also a third layer I would introduce: artists themselves. No matter how fast the medium they're engaging with, artists often hibernate, take their time and return to ideas.

I would make a distinction between a large art-historical view and the speed at which certain movements and ideas arise. The speed of the medium is set by the real world—and artists have to work in that world.
‍

Art is always about engaging with the current moment in time. Digital art is ideally positioned to reflect on the world we live in today, which is more and more algorithmically driven. 

So digital art is incredibly important. Artists using photography or video ultimately did the same: they were engaging with the technologies of their time—how they were changing societies as well as representation.

‍Peter Bauman: Speaking of the current moment, it seems like we're seeing renewed institutional interest at a time when art markets are broadly slowing. How do you reconcile that contrast? Last time we spoke for a piece, you said that institutional interest can be cyclical, too. Is the art market distress more of a short-term indicator (noise), while the institutional interest is what's more enduring (signal)—or might it even be the opposite?‍

‍Christiane Paul: Institutions definitely intersect with the market in many ways but they also, in terms of their mission, are operating very independently of it. Institutional collecting is not driven by market interests.
‍

Museums do not buy and sell continuously. If they're collecting, they are focused on collection building and telling a story. 

We're looking at two very different operations. The fact that the art market is not doing well is unavoidable because markets are not doing well anywhere on this planet right now. That will also change again at some point. The current low point is completely understandable.

As you mentioned—and as we discussed in our previous conversations—there are always these cyclical movements of institutional interest. We had the same gap—or disconnect between market lows and institutional highs—during the dot-com bust. The big digital art shows at institutions around the 2000s were technically happening after the dot-com bubble burst.
‍

I am not sure if the current moment of institutional interest is sustainable—if it will remain at the current level. But overall, there definitely has been progress in that digital art is becoming more integrated into institutions—being exhibited, being collected, et cetera. 

Invariably, the cycles are continuously propelling us forward a little bit—in slow motion. We're always making a little progress.
‍

‍
----
‍


Christiane Paul is Curator of Digital Art at the Whitney Museum of American Art and Professor Emerita in the School of Media Studies at The New School. She is the recipient of 2023 MediaArtHistories International Award and the Thoma Foundation's 2016 Arts Writing Award in Digital Art. Her books include Digital Art (Thames and Hudson, 4^th ed., 2023) and A Companion to Digital Art (Blackwell-Wiley, 2016).
‍
‍‍Peter Bauman (Monk Antony) is Le Random's editor-in-chief.