Q&A With Ho Tzu Nyen on AI and Art

Editor’s Note: This story is part of Newsmakers, an ARTnews series where we interview the movers and shakers who are making change in the art world.

When Hong Kong‘s M+ opened in 2021 after over a decade in development, the contemporary art museum instantly became central to the city’s art ecosystem, putting on ambitious exhibitions of both Asian and Western art.

It also became central to the cityscape, thanks to an enormous LED facade that stands at over 200 feet tall and over 350 feet high. M+ commissions works to fill the facade throughout the year, with the most high-profile of these being co-commissioned with Art Basel and launching during the company’s Hong Kong fair.

Related Articles

The latest artist to try their hand at the M+ Facade is Singaporean artist Ho Tzu Nyen, whose work Night Charades will be on view through June 29. (Concurrently, Ho is the subject of a solo exhibition of recent films and works, “Three Stories: Monsters, Opium, Time,” at Kiang Malingue gallery through May 13.) A tribute to Hong Kong cinema from the 1980s and ’90s—often considered the industry’s “golden age”—Night Charades uses artificial intelligence video generation to create characters who perform actions and gestures that recall the iconic films of directors like Wong Kar-wai, John Woo, Stephen Chow, and others in infinitely generating combinations.

Ho, who represented Singapore at the 2011 Venice Biennale, is known for experimental films, sculptures, and multimedia installations that challenge established notions of history, identity, and culture. As always with the artist, Night Charades is far from the simple exercise in nostalgia or whiz-bang AI prowess it might appear as. As he told ARTnews, he used AI for the first time in his practice in an effort to “decontextualize” this beloved culture. What’s left, he said, is “a strange and maybe slightly monstrous combination of different types of cultural references” that blur the boundaries between past and future.

ARTnews spoke with Ho about Night Charades, the ethics of AI, resisting nostalgia, and creating a work fit to the M+ Facade’s gargantuan proportions.

This conversation has been edited and condensed for concision and clarity.

ARTNews: Research is obviously very important to your practice, and I imagine it was very important to this work in particularly, given its many references to Hong Kong cinema. Could you walk me through the research and conceptualization process for Night Charades?

Ho Tzu Nyen: For this project, it was a little bit different [than past projects] in the sense that Hong Kong cinema, particularly of the ’80s and the ’90s, were things that I grew up with. They were very much part of the fabric of my childhood. Hong Kong pop culture, from the films to the television to the pop music, were things that I was completely immersed in. In a sense, this work is special just because it drew so much on my own experience of growing up with this very special cinema. But research is definitely part of my process. This time, I was researching the social and historical context from which this cinema emerged. I know the films so well but, of course, when I was growing up in Singapore, I didn’t have that much knowledge about the context. Now, almost 30 or 40 years later, I’m researching the context of these things that I’m so familiar with.

Were there particular things that you learned during the process that surprised you?

A lot of my surprises came from learning about the specific ways Hong Kong cinema developed—for example, in relation to Western cinema. As a kid, you have a sense that Hong Kong cinema, on some level, duplicated Western genres while at the same time creating something different. As I was doing the research, I was very intrigued to know how much of this duplication was deliberate. It was a very conscious strategy of using Western cinema genres, while also working against them. Also, I was always very confused as a kid, because in Hong Kong films, there would often be a film that became very popular and then immediately after there would be a string of other films that resemble it. You often didn’t know if it was a real sequel or not. Sometimes, even the same actors would be in these [derivative] films. And now I learned that it was, in fact, a deliberate strategy that other producers would make similar films, even with very similar titles, just to ride on the popularity of the first film. It was an industry-wide practice.

Right, how many versions of Ip Man are there?

Ip Man was kind of the last great example of this phenomenon. Earlier, in the ’90s, there was a very popular Hong Kong film called Young and Dangerous. It was about street hooligans. After that became a hit, there were so many other similar films, many of them starring the same actors. This [strategy] is one thing that really amused me to learn. As part of the research, I was also watching documentaries. There’s a more recent film about the stunt industry in Hong Kong cinema [Kung Fu Stuntmen: Never Say No! (2020)]. A lot of superstars like Sammo Hung and Jackie Chan emerged out of this stunt and action choreography industry. When you watch the documentary, you can see how they did their stunts, and it was crazy. There were basically no special effects. When there was an explosion in a building and people flew off from the seventh floor, they really did it. They jumped out of the building from the seventh floor onto a mattress and cardboard on the ground floor. When I was watching the movies as a kid, I always assumed there were some tricks, but, in many ways, it was all real. This was mind blowing for me.

Did your discovery of all the repetition and remaking inherent to the Hong Kong film industry influence how you approached Night Charades? The work is in many ways a deconstruction and then remaking of all these films.

My process for this work was a series of remaking and reconfiguring. I was using what came before while, also trying to put my own spin on it. Repetition and difference. At the same time, this method of using what already exists and reconfiguring has always been very much part of my practice. I often work with found footage or existing narratives and then I try to subvert them. When I started reading about the history of Hong Kong cinema, I began to wonder if I got this tendency from my own experience of watching so many Hong Kong films as a kid. This notion of taking something and redoing it in your own way. Maybe the roots of my practice are connected to growing up simultaneously with Hollywood cinema, Hong Kong cinema, Japanese films and anime. There’s this practice of picking from both the East and the West and fusing them together and remaking them into something that’s your own. That’s the tendency of my practice, but it’s also pretty much the way Hong Kong cinema evolved too.

You’ve used archival footage in many past projects of yours. For example, in Hotel Aporia (2019), a six-channel video work staged in a Japanese inn, you used archival footage and excepts from Japanese films to explore Japanese imperialism during World War II. But was there a different way of thinking when using AI to create an entirely new version of past footage for Night Charades? What is your responsibility to the archive you are pulling from?

It’s different with each specific project, because the way that I use existing footage is technically and conceptually different each time. At times, I have used found footage in a very direct way without much transformation or modulation of the original materials. I have also altered found footage, for example, through roto-scoping. This was the first time I used AI to reprocess the existing materials. It’s difficult to say if there is one general rule or principle, but, in general—and I hope this doesn’t sound lame—intention matters. I try as much as possible to embark on all these projects with well-meaning intention. I don’t usually cloak or disguise the facts or the sources that I draw from.

Night Charades grew out of an homage to the original materials. That’s the starting point. But this is the first time I’m using AI processes and I must say that there is not one single AI process for image generation. Each one has its own specificities. Night Charades was a process of discovery for me as to what these tools are and how they work. I do believe that AI profoundly transforms the original materials.

For example, think about cinema in relation to photography. Photography was the writing of light—that’s literally what it means. In the past, cinematic images were inscribed by light onto the negatives. This process changed with the digitization of the filmmaking process. The relationship between the light and the inscription became much more complicated and less direct, as we involved pixels and digital information. When it comes to AI, there is a total and absolute removal from natural light. In an almost ontological sense, AI-generated images are images produced without light. They are born out of … I don’t quite know how to describe it. They are produced from a closed world that is not open to outside light.

Some have argued that the usage of AI in art results in a total removal of historical context from the source imagery and produces a homogenized result. (This was a major criticism of Refik Anadol’s interpretation of the Museum of Modern Art archive with Unsupervised.) Given historical specificity of Hong Kong cinema, how does that context or its removal operate in Night Charades?

It’s hard for me to reduce the work to a single intention, but maybe there are multiple ones. I think it’s also fine if the different intentions of a work contradict each other, because contradictions are a part of life.

The process of generating these images was a deliberate de-contextualization. We removed the background of all of the images. The bulk of what you see in Night Charades is a figure with no background. This removal is a very literal kind of decontextualization but, at the same time, of course, the actions, gestures, and poses of the figures recall the original films. For someone like you or me, these gestures trigger our memories of the originals. But maybe half of the audience, or more, has no idea what these films are, or what these gestures or actions refer to. Night Charades is a public work. It’s projected on this giant screen. But it divides the audience. I find that very interesting. The next step for me is thinking about what these two sets of audience get from the work. For us who know the films, it’s recollection and memory. But for those without this context, the work becomes a kind of riddle or rebus. The actions become enigmatic, and you have to guess what it is.

Another aspect that interested me a lot was the specificities of the tradition of animation and its relationship to AI. Traditionally, when creating 3D animation, air dynamics—like how wind changes, the movement of hair or cloth—is the most complex and expensive thing to do. It’s why Pixar characters are usually very smooth or bald. But with AI, you can do wind dynamics and movement quite easily. For Night Charades, a lot of the characters are dressed in clothing with a lot of pleats. For every one of their movements, you can see how the cloth moves in a strange and slightly artificial way, which I find quite beautiful. This is a different aspect to the work—using the original actions to create new kinds of feelings and sensations.

Is there a hope embedded in the work that, for the second audience, which sees the work without context? Do you think some of the meaning and feeling of these gestures, which means so much to you and me, will carry over to them through pure form?

Some of these gestures mean a lot to us because we know the specific source, but many of the gestures have entered the broader cinematic culture. Just one example is John Woo’s very famous “double gun” gesture, which has entered Hollywood films like The Matrix. So even an audience who doesn’t have the specific context of Hong Kong cinema, would find some of these gestures and postures very familiar. And Hong Kong cinema itself has often appropriated from the cinema of other traditions. It’s a long stream of influence and transformation. The great Hong Kong cinema of the ’80s and ’90s was just one point in this long stream. This flow, which is sometimes subterranean, is what interests me the most in cultural history.

In relation to the question of context and decontextualization—the last time I visited Hong Kong before now, I went to a few museums to look at the exhibits. Not just art museums, but also cultural museums. It struck me that so many of these exhibitions were framed in relation to the “golden age” of Hong Kong cinema, or the “golden age” of Canto-pop. Growing up I enjoyed a lot of these products, and so I could join in the celebration of this amazing culture. Nevertheless it left me melancholic. This notion of the “golden age” as something unsurpassable and already in the past. That experience affected me quite a bit. A lot of that went into the conceptualization of Night Charades. So while I tried to draw from the past, I plugged it into this system that we we usually associate with the future, sometimes in fear. I wanted to see what happens out of that meeting.

When I first read about the project, I assumed that it was primary about cultivating nostalgia. But hearing you talk about it now, it sounds like you’re trying to do the opposite. Or maybe, playing with the tension between those two things.

Nostalgia is one of the key words that I had in conceptualizing Night Charades. I find nostalgia melancholic and problematic, but it’s also undeniable. I feel this nostalgia as well. So while I’m trying to counter it, I’m not doing so out of direct and fierce opposition, but because I am also part of it.

You mentioned that this is the first time you’ve used AI. These types of AI systems, whether its a text generator or an image generator, are designed to reproduce the median of their traning data. Obviously that can mean the reproduction of the historical biases or societal assumptions or stereotypes embedded in the data set.  Was that something that you considered with Night Charades?

It’s true that the existing AI platforms are built on specific models that rely on data sets born out of the base aggregates of culture. There is a tendency to reduce things to types. This is one way of thinking about it, and we have to be conscious and alert. But there is also some kind of power in the aggregate. I don’t know if archetype is a good word for it, but something close to that. When I am using these platforms, I play into both sides of this. You draw on types that have the power to communicate widely, that can resonate with large numbers of people, but you are also dealing with the biases inherent to that. It’s something that one just has to be conscious of throughout the process.

What interested me in Night Charades was the combination of the specific qualities of Hong Kong cinema with these large aggregates and biases. Hong Kong cinema provides a very interesting counterpoint to these gross stereotypes and biases. For example, in the movies of Stephen Chow, who I’m big fan of, there are a lot of scenes of people picking their noses on a large screen. And just like feeding that into an AI system and seeing what comes out of the combination of the typical aggregate of the system with something so specific to his films and, in a way, low culture. It produced very interesting and quite original combinations.

You’re almost creating your own archetypes for Hong Kong.

Yes, but through the AI processes and the additions that we threw into the pipeline that one could say have a closer relationship to high fashion [like the pleated fabrics]. The combination then becomes very strange. You have lowness in the sense of Hong Kong cinema and then you have something that is typical of the AI data sets, which tends toward glossy surfaces. This high fashion design and clothing, and the way it reacts with the air and movement, reminds me a lot of Renaissance and Baroque paintings, especially Caravaggio. Night Charades, for me, is this strange and maybe slightly monstrous combination of different types of cultural references.

And then once you put it on this massive screen, it becomes a character within the city itself.

It’s a character within the city, but the size of the screen and its position in the city is interesting. I like to think that it is so huge that everyone can see it, but no one is watching it. It’s so big that you don’t really watch it the way you watch a film in the cinema or even in a gallery, where you focus on a narrative that you follow for a duration. With a screen as big as that of the M+ facade, the light that is generated from the screen becomes ambient light. It becomes atmospheric and environmental and, in a way, architectural. The image is transformed into something else. For Night Charades, I didn’t think it was possible or a good idea to create a narrative that one needs to follow. The concept of a charade was interesting to me, in that sense, because it is a sequence of small, little acts. It doesn’t demand or require the focus of following a straight narrative, but it’s a sequence of different performances. It’s possible to read into this sequence. You could make your own narrative out from it, if you choose, but you can also detach from this sequence and just focus on the individual.

One thing about AI that I’ve been thinking about often lately is how it allows the artist to remove themselves from the center of the work. I’ve noticed you do this in other works. With The Critical Dictionary of Southeast Asia (2017–ongoing), you surrendered editing control to an algorithm to weave together different combinations of texts, music, and images that explored Southeast Asian identities. Do you see a kind of promise in AI for artists that they might be able to release themselves from being the sole creator or arbiter of truth?

What interests me with AI is the uncontrollability. I like to work with processes in which I have no mastery, or where mastery is an impossible task. My works are born out of this encounter with limits and obstructions and obstacles and then folding these limits and obstacles into the process to find the form of the work. I’m not sure, to be honest, how much of this was conscious to me over the years, but I have now realized it is a constant part of my work. I suppose I just like to make things difficult for myself. Limits, difficulties, and obstacles are interesting to me because one of the criteria I have for making a work is how much of the world I am able to fold into it. The more of the world the work folds into it, the denser it becomes, the more it interests me. Sometimes, these obstacles are technical, like these AI platforms that you can’t quite control. Sometimes it’s political—certain things that are unsayable, and we have to find some way to fold it into the work. With AI, it is like a wave, and you are riding it, but you can’t quite control it.