Refik Anadol: art in a latent space

Pau Waelder

Refik Anadol. Unsupervised: Machine Hallucinations MoMA (2022)
Museum of Modern Art, New York

Refik Anadol (b. 1985, Istanbul, Türkiye) is a media artist and lecturer, whose meteoric career has taken him from creating video mappings on building façades in several European cities, to being one of the first artists in residence at Google’s Artists and Machine Intelligence Program, the founder and director of Refik Anadol Studio RAS LAB in Los Angeles, a lecturer for UCLA’s Department of Design Media Arts, and a successful artist with global recognition in the contemporary art world. In just 15 years, Anadol has amassed numerous awards and presented his site-specific audio/visual performances at iconic museums and events such as the 17th International Architecture Exhibition – La Biennale di Venezia, Walt Disney Concert Hall, Centre Pompidou, Daejeon Museum of Art, Art Basel, Ars Electronica Festival, and the Istanbul Design Biennial, among many others.

He works with a large team of designers, architects, data scientists, and researchers from 10 different countries and has partnered with teams at Microsoft, Google, Nvidia, Intel, IBM, Panasonic, JPL/NASA, Siemens, Epson, MIT, Harvard, UCLA, Stanford University, and UCSF, to apply the most innovative technologies to his body of work. He is represented by bitforms gallery in New York. 

bitforms is participating in the Art SG art fair in Singapore from 12 to 15 January 2023, presenting a selection of generative artworks by Refik Anadol. Niio supports the exhibition as technical partner, in collaboration with SAMSUNG, and is proud to present the artcast Refik Anadol: Pacific Ocean, which features excerpts from three pieces by the Turkish artist. The following article offers a brief introduction to the main aspects of Refik’s work.

Refik Anadol. Unsupervised — Machine Hallucinations — MoMA Dreams — F
Image sold on Feral File as NFT. 100 editions, 1 AP

Building a latent space

A trailblazing artist in the field of art and artificial intelligence, Refik Anadol uses large amounts of data and machine learning techniques to create his generative artworks and site-specific installations. His creative process often implies the creation of a data set from an archive of images, sounds, and documents or from measurements taken by sensors, radars, and other devices. The data set feeds a series of machine learning processes that generate an endless succession of audio-visual compositions, which can fill a large screen, a whole room, or the façade of a building. 

At the heart of the machine learning models that transform the original data into something else lies what is called a “latent space,” in which clusters of items are formed from similarities between them, which give rise to a set of variables. The latent space is therefore a space of possibilities, somewhat unpredictable, that contributes to shaping the final outcome. In Refik’s work, it is not only part of the machine learning model but also a concept that helps understand his generative pieces and installations as spaces in which creation is constantly exploring its latent qualities. Spaces in which the artwork is never finished. 

His recent installation at the Museum of Modern Art in New York, Unsupervised: Machine Hallucinations MoMA (2022), clearly exemplifies the conception of the artwork as a latent space. Using the public metadata of The Museum of Modern Art’s collection, which comprises more than 130,000 pieces including paintings, drawings, photographs, and video games, the artist and his studio created a series of artworks that result from the interpretation of this data by means of a machine learning algorithm. The initial series was sold as NFTs in the exhibition Unsupervised on the online platform Feral File, the project being further expanded into the generative artwork installed at MoMA’s lobby. Casey Reas, artist and co-founder of Feral File, aptly described the artwork in terms of its latency: “What I find really interesting about Refik’s project with MoMA’s dataset, with your collection, is that it speculates about possible images that could have been made, but that were never made before” [1]. The artwork can thus be seen as a space of possibilities, but also as a simulated environment that becomes particularly meaningful in the context of the building that houses it.

Casey Reas: “What I find really interesting about Refik’s project with MoMA’s dataset is that it speculates about possible images that could have been made, but that were never made before”

Refik Anadol Studio. WDCH Dreams, 2018.

The room as Merzbau

Architecture, and more generally a real or simulated three-dimensional space as a container, are key elements of Refik’s work. Artworks such as WDCH Dreams (2018) or Seoul Haemong (2019) use the exterior surfaces of the Walt Disney Concert Hall in Los Angeles and the Dongdaemun Design Plaza in Seoul as canvases, while Infinity Room (2015) and Pladis: Data Universe (2018) are conceived by the artist as “Temporary Immersive Environments.” The artist stresses this connection frequently in his interviews: “I’m interested in exploring the architectural domain as deeply as I can,” he has stated recently. “All my art works tend to have a physical connection to public space” [2]. However, the architectural space is not conceived in terms of static shapes and volumes, but as something fluid and malleable, a work in progress. 

Kurt Schwitters’ celebrated Merzbau installations from the 1920s and 1930s come to mind as an illustrative example of Refik’s conception of space. Schwitters began to alter the space of his studio in Hannover by putting together small artworks, found objects, and debris into structures that he would glue and fix with plaster, building columns and shapes that protruded from the walls. The sculpture was never completed, the artist always kept adding elements and reshaping the space [3]. In a similar way, the space occupied by Refik’s artworks is permanently reshaped through a process that stems from an accumulation of found materials, data that loses its original shape and merges into something new. In the Temporary Immersive Environment series, he specifically seeks to immerse the viewer in a “non-physical world” that questions their perception of space and their own presence [4]. The room is expanded and multiplied through optical effects, with the aim of creating a viewing experience that goes beyond staring at a flat projection.

Refik Anadol: “I’m interested in exploring the architectural domain as deeply as I can. All my art works tend to have a physical connection to public space.”

This conception of space as integrated into a Gesamtkunstwerk, the “total work of art” that has been the aspiration of opera composers, architects and filmmakers, is not, however, the only connection with architecture in Refik’s work. Interestingly, while Schwitters sought to merge all of his artistic practice under one term (Merz), erasing distinctions between painting, sculpture, and architecture, Anadol describes some of his generative art works as “data paintings” and “data sculptures.” These references to classical formats speak of a different type of space, confined within the limits of a screen or a wall, which nevertheless intervenes in the surrounding space by means of a trompe-l’oeil effect that creates the impression of three-dimensional shapes pulsating beneath and beyond a solid, thick frame. Artworks such as Virtual depictions: San Francisco (2015), displayed on an L-shaped media wall inside the main lobby of the 350 Mission building in San Francisco, seek to create an imaginary space that stands out spectacularly, but at the same time embeds itself into the surrounding architecture. The connection between the artwork and its location, though, is not only expressed in terms of how the screen is placed on the wall, but also in the data that gives meaning to the fluid elements that inhabit the virtual space.

Refik Anadol Studio. Future of the City, 2020.

Data is not just a bunch of numbers

Coming back to the concept of latent space within machine learning models, it is important to remember that Refik Anadol’s artworks do not only have an aesthetic dimension, as colorful shapes in fluid transitions or enormous mosaics of distinct elements, but also a conceptual dimension, expressed by the data that feeds the whole process leading to the site-specific installations and performances. Speaking about his project Quantum Memories (2020), the artist states the importance of this data and the meaning it conveys: 

“For me, data is not just a bunch of numbers. For me, data is actually a memory. From that perspective, I’m always looking for what kind of collective memory that we are holding as humanity, and how can we use these memories and turn them into a pigment or a sculpture that represents who we are as humanity.” [5]

Conceiving data as memory resonates with his ongoing work with all kinds of archives, from the 1,700,000 documents found in the SALT Research collections to the 587,763 image files, 1,880 video files, 1,483 metadata files, and 17,773 audio files in the Los Angeles Philharmonic Orchestra’s digital archives. Every bit of information in these files has its own history and meaning, and has a certain association with another file, enabling the clusters and variables that will emerge in the latent space. The artworks may appear as abstract compositions or massive collages, but they are actually visual representations of underlying stories and invisible structures. As the artist puts it, they aim to “make visible the invisible world of data that surrounds us” [6]. This statement may lead to considering Refik’s work as a form of data visualization, but it goes well beyond this task, into what media theorist Lev Manovich has described as “represent[ing] the personal subjective experience of a person living in a data society […] including its fundamental new dimension of being «immersed in data»” [7].

Lev Manovich: “The real challenge of data art is how to represent the personal subjective experience of a person living in a data society.”

Some of Refik’s installations directly address this condition of being immersed in data by means of projections that surround the viewer with visualizations of the data collected from archives, or in real time from sensors and other sources, as in Latent Being (2019) or Future of the City (2020). Others present data that relates to environmental systems that we usually ignore but that have a profound impact on our planet, and therefore in our lives. This is the case of Pacific Ocean (2022), the series presented by bitforms at Art SG in Singapore and on Niio as an artcast featuring three video excerpts.  

bitforms gallery booth at ArtSG Singapore presenting a series of artworks by Refik Anadol. Photo courtesy of bitforms.

Collecting data from High Frequency Radars (HFR) located in the Pacific coast of the United States, the artist has created a series of visualizations of ocean currents that seems abstract and realistic at the same time: the ebbs and flows of granular elements in shades ranging from dark blue to emerald green clearly evoke the surface of a raging sea, but they are also somehow unreal, impossibly merging numerous currents from different directions in beautifully chaotic, and even violent, clashes. HFRs are used to measure ocean currents and understand their impact and response to climate processes. The data collected from networks of radars in coastal zones around the world is crucial to protect the marine ecosystem and predict changes that will affect life on our planet. Seen from this perspective, the artworks acquire a somewhat unsettling tone and inspire an awareness of the ecosystems that we so often ignore, putting into question our anthropocentric view of the world.

Refik Anadol. Pacific Ocean A, 2022

Machine dreaming

Anthropocentrism and our inability to understand the agency of non-human entities and systems around us are underlying subjects in most approaches to art created with artificial intelligence. The perennial question of whether it is the artist of the machine that creates the artwork is still debated after 60 years of algorithmic art, now reinforced by the spectacular achievements of machine learning models in producing realistic images and coherent texts. In Refik Anadol’s work, the use of artificially intelligent systems leads to two interesting aspects of artistic creation: the notions of control and authorship. 

Terms like “machine learning,” “supervised learning,” “reinforcement learning,” and “training model” speak of the intention to use artificial intelligence as a tool to obtain predictable results, in which the machine is meant to produce a specific output. This perception of the machine as a mere instrument, fully controlled by a human, contradicts the way artists have used generative algorithms and AI systems to create their artworks. Nowadays, artists working with artificial intelligence understand machine learning as a way of exploring post-anthropocentric creativity, therefore using AI to reach beyond the confines of human imagination and let the machine bring in the unexpected, the incongruous, the unsettling, and even the impossible. In Refik’s work this approach is made clear in the use of machine learning models to create “dreams” and “hallucinations.” He has described AI as “a thinking brush, a brush that can think, that can remember, and that can dream.” This statement implies an interesting balance between letting the system loose and keeping it under control. In Archive Dreaming (2017), the installation is allowed to “dream” when a viewer is not interacting with it, so that this state is interrupted when a human takes control. In other installations, such as Machine Hallucination (2019), the system can create its own associations and reimaginings of the contents of a very precise dataset, so that its “unconscious” is nevertheless under a certain level of control. 

Refik Anadol: “the most important thing for me is creating a thinking brush, a brush that can think, that can remember, and that can dream.”

The question of authorship stems from the perceived control over the final output: if the artist had no control over it, is he the author of the artwork? Interestingly, while the Dadaists and Surrealists already integrated randomness into their artistic practices and many other artists have incorporated unpredictable processes or external agents into their work, authorship tends to be more fiercely contested when a computer is involved. Refik Anadol’s authorship is nevertheless palpable in the aesthetic and conceptual foundations of his work, which remain consistent throughout his career despite considering himself part of a large team of experts and working with increasingly complex AI technologies. He conceives the process as a collaboration, both when dealing with software and hardware and when teaming up with designers, coders, and researchers to develop a project. There are, however, crucial moments when decisions are made, and these are the moments when the artist states his authorship:

“There’s a collaboration between machine and human. With the same data, we can generate infinite versions of the same sculpture, but choosing this moment, and creating this moment in time and space, is the moment of creation.” [8]

Out of infinite possibilities, making a choice that determines the next step in the process and shapes the final output is a prerogative of the artist, who is finally the author of the artwork that emerges from a latent space.


[1] Refik Anadol, Casey Reas, Michelle Kuo, and Paola Antonelli. Modern Dream: How Refik Anadol Is Using Machine Learning and NFTs to Interpret MoMA’s Collection. MoMA | Magazine, November 15, 2021.

[2] Dorian Batycka. Digital Art Star Refik Anadol’s First Supporters Were in the Tech World. All of a Sudden, His Work Has Become White-Hot at Auction, Too. Artnet, May 18, 2022.

[3] Gwendolen Webster. Kurt Schwitters’ Merzbau. A dissertation presented to the Open University, Milton Keynes, 2007.

[4] Refik Anadol. Liminal Room. Refik Anadol Studio, November 19, 2015.

[5] Claudia Pelosi. Machine intelligence as a narrative tool in experiential art – Interview with Refik Anadol. Designwanted, December 19, 2020.

[6] Refik Anadol. Virtual Depictions: San Francisco. Refik Anadol Studio, October 1, 2015.

[7] Lev Manovich. Data Visualization as New Abstraction and Anti-Sublime., 2002.

[8] Refik Anadol, Casey Reas, Michelle Kuo, and Paola Antonelli. Modern Dream. op.cit.

Get to Know Anne Spalter: Academic pioneer, artist, collector, curator and author

You have an MFA in painting from RISD: Rhode Island School of Design.  What came first for you, art or technology?

I studied traditional analog art in high school and at RISD as an undergraduate before transferring to Brown University to study mathematics. I was as surprised as anyone to find myself using the computer to make art after graduation. In fact, I have to admit that I was a total computer-phobe and pretty much thought art made with a computer had to probably be evil (which, unsurprisingly, seemed to be the general institutional sentiment when I later returned to RISDI for my MFA).

I really began to appreciate the visual power and convenience of the digital world when I was working in New York after college. I had a computer in my cubicle and was working a gazillion hours a week. The only way I could work on my art was inside that machine; it became my tiny studio. I began to explore what I could do with art software like Photoshop… and when my boss walked by… presto… like magic… I could easily click back on Excel.

After a while, I realized my future probably didn’t rest in banking [duh] and I applied to graduate school in Painting at RISD–of course, without mentioning that I’d been using a computer to make art, since that was still a pretty subversive insult to the painting tradition back then.

Once I was back at RISD, I wanted to continue exploring the digital realm. There weren’t any classes offered, though, so I actually ended up getting “volunteered” to teach one to my fellow graduate students. In the course of developing curriculum for what became the first digital fine art courses at RISD and Brown, I ended up writing the widely used textbook, The Computer in Visual Arts (Addison-Wesley). This was a multi-year undertaking and it brought together aesthetic, technical, and art-historical aspects of the field.

I was also fortunate during this time to work with Andy van Dam in his Computer Graphics Research Group in Brown’s Computer Science Department where I was a researcher and Artist in Residence. It was a bit like an old fashioned apprenticeship and I learned about the technical side of the field.

After 15 years immersed in academia, I thought, if I don’t give my art career a chance it’s probably never going to happen. I took a sabbatical in 2008 and ended up not returning.

Precession at SPRING/BREAK Art Show, curated by Elizabeth Keithline, New York, NY, 2016. Wall mural, video screens, canvas prints. Courtesy of Anne Spalter.

What do you think is most misunderstood about digital art and what would you like people to know?

Ahh, where to begin. Perhaps the most common myth, similar to the “anyone can point and click” accusations that plagued photography, is that somehow the computer is making the art for you and the digital artist is merely pressing a mouse button and sitting back with a toothpick while the machine does all the hard work.

In reality, it is exactly as difficult to make art with the computer as with a paint brush or any other medium because the hard part doesn’t lie with the technical device or medium but in the artmaking aspects–the choice of content and how to express it. A great artist can make art with a crayon–as Picasso did drawing a dove–and a lousy artist can fail to make art with the world’s most powerful supercomputer. If anything, I’d say it takes more hours working with a computer than most traditional media because it is a new technology and involves a lot of tedious problem solving on a daily basis.

As we move into a new world of machine learning and AI, this may change and the computer may truly be making art, but I have not seen convincing examples of that yet.

I’d say another misunderstanding, and one that photography and video art and some other art forms share but that has for some reason particularly plagued digital art, has been that “it’s not art because it isn’t done by hand.” For reasons that are not entirely clear (given the history of other genres for which this is true), this stumbling block keeps reappearing. Ironically, I have often felt more like I was drawing and painting when working with digital video than I did wielding a physical pencil or brush. One would think in this day and age that art could be accepted for it’s conceptual and aesthetic qualities and the mediation of the hand would not be brought into so many discussions–but it remains an issue.

I could go with these misunderstandings forever, but I’d say these are the most common.

Adrift on Titan (Miami Marbles series) at PULSE Contemporary Art Fair Miami Beach, FL, 2016. The first PULSE PROJECTS Special Commission, Miami Marbles is a mixed augmented reality (AR) installation combining AR components, via a custom app, with nine physical helium-filled spheres—ranging from seven to 16 feet in diameter— printed with digitally manipulated footage of Miami Beach; Courtesy of Anne Spalter
Adrift on Titan (Miami Marbles series) at PULSE Contemporary Art Fair Miami Beach, FL, 2016. The first PULSE PROJECTS Special Commission, Miami Marbles is a mixed augmented reality (AR) installation combining AR components, via a custom app, with nine physical helium-filled spheres—ranging from seven to 16 feet in diameter— printed with digitally manipulated footage of Miami Beach; Courtesy of Anne Spalter.

You use custom software to create your work. Have you always developed/used your own tools?  Tell us about that process.

At this point I have custom software, but there was a process behind the evolution that led me there. I mostly used off-the-shelf software (back to Photoshop 1.0 and even its rudimentary predecessors like Letraset Realist) but starting with my kaleidoscopic video works, the standard software didn’t offer all of the features I felt I needed.

I began working with a wonderful programmer, Nathan Seilikoff, on custom plugins for Adobe AfterEffects and Photoshop. These let me work with more parameters for the patterning and motion, and also control them better. I did take programming courses but, basically I’m a slob and spent an inordinate amount of time chasing down stray semicolons. Learning how to program does help me understand what is possible, however, and to communicate with people who are good programmers. That said, I am happy working with programmers like Nathan to develop my custom software.

New York Dreaming, The Fulton Center, New York, NY, 2016-2017; 9 corresponding videos on 52 screens throughout the Fulton Street Transportation Hub; Courtesy of Anne Spalter.

As an artist, what do you think are the biggest challenges in exhibiting digital art?

Some galleries simply will not exhibit digital art. Of course, many artists use the computer at some point during their creative process (from image research to composition work and beyond), so this is an increasingly meaningless statement.

For those that do exhibit new media, the biggest challenges are making everything run smoothly and supplying instructions for non-technical people to easily keep things running. Surprisingly, there are still not any widespread “entry level” mechanisms for basic digital art display (e.g. a simple and easy-to-use screen or projector that would seamlessly loop digital files), and thus a new media show usually involves a crazy set up of extension cords and media players; multiple remotes with line-of-sight issues; and other things that are baffling to gallerists unaccustomed to such technology. Things are even worse if an actual computer must be kept running the whole time a show is up. Many spaces turn their power off at night requiring everything to be reconfigured each morning.

Beyond technical display issues, it is also sometimes challenging to explain to viewers that a work could be shown in their home or institution differently from the way they are experiencing it in the gallery or museum. For example, it is sometimes hard to explain to  people that a video work would look fine on a different sized screen, or on a screen even though they are seeing it projected. Prospective clients also often balk at having to choose their own screens or other equipment, even with advice from the artist or gallery.

Wonder Why, 5K digital video, 2017; 7 minute loop; Courtesy of Anne Spalter.

As a collector and artist, what  do you think are the biggest challenge in collecting digital art?

We collect mostly early works that are plotter prints on paper. As such, we avoid most of the archival and storage issues of collecting new media as essentially they are india ink on paper and this is a well known entity. We do have some works that are video, i.e., files, and those are backed up multiple times–both on physical hard drives and in the cloud.

Some of the challenges are the same as any art collection I think–storage, organization, documentation, etc. We have begun to put the collection online for research purposes–to share it with a broader audience and you can see our efforts so far at Many works are not up yet as they need to be photographed, and we are still entering data. Implementing any new cataloging system is always an incredibly daunting task, though.

As an artist creating digital video works that others collect, I have struggled to find an optimal way to present work for easy display and use and integration with clients’ existing collections. Several of my collectors have told me my work was the first new media piece they purchased; I think that is because I go to great lengths to try to make it easy to install and maintain the work. My ongoing search for solutions in this space led me to Niio. Their approach is the only hardware agnostic one I know of —letting clients use their existing screens without the need to  deal with extra remotes.

In addition, unlike virtually all the made-for-art displays available, Niio addresses file quality issues that have always bothered me, letting clients view the optimal version of the video. The Niio server also stores archival versions of the file, and addresses distribution and backup issues. It is difficult to communicate some of these features to collectors who are not technologically inclined, but they are supremely important.

Beacon, 1080p digital video, 3 minute loop, 2018; Courtesy of Anne Spalter.

As a collector and artist, how do you tackle the topic of  preservation?

[see above for the Collection]

As an artist I live in fear that I will lose files! This is the equivalent of a fire in the studio. I make local backups to a RAID array, multiple cloud backups, and off-site backups.

AR – Are you a fan as an artist? As a collector?  Any works in your personal collection?

I really love AR. I think AR has a great combination of convenience and aesthetic quality and ease of use. We own one of Claudia Hart’s AR works. I thought Will Pappenheimer’s Privateer (in Boston Harbor) was super. And, of course, I did an app for Pulse Contemporary Art Fair Miami Beach in 2016–Miami Marbles–which you can download from the app store.

What do you think will help establish the stature of digital art in the context of the global art world?

Fortunately this is already happening! Although we are all impatient, it really hasn’t been that long since the invention of digital computers and the advent of digital art to, now, shows at major museums featuring new media works.  I think we all live in internet time and expect things to happen almost instantaneously. Artists already know that the computer is a part of the art-making process and use it without hardly thinking about it.

For collectors and the critics and THE art world status quo to accept it will just take a few real leaders to give it THEIR seal of approval. Shows like the Thinking Machines, curated by Sean Anderson and Giampaolo Bianconi currently up at MoMA in NYC, help accomplish this. Not only is it a thoughtful interesting show but it brings together digital and analog works under the aegis of a larger theme and it doesn’t comment on the difference. The show treats all the artists equally — de facto as part of art history. This is the type of thing cements digital art into the canon.

About Anne Spalter:

Digital mixed-media artist Anne Spalter is an academic pioneer who founded the original digital fine arts programs at Brown University and The Rhode Island School of Design (RISD) in the 1990s. In her studio practice, Spalter uses custom software to transform source footage—captured by the artist during multisensory experiences such as riding the Coney Island Cyclone; walking through an open-air flower market in Bangkok; and gazing down from a helicopter over downtown Dubai—into kaleidoscopic, algorithmically manipulated Modern Landscapes.

Spalter, who studied mathematics as a Brown undergraduate before receiving an MFA in painting from RISD, has a longstanding goal of integrating art and technology. With additional cross-disciplinary masteries including a 2011 Sensei designation in Kenpo Karate, Spalter’s influences in the studio are as diverse as Buddhist art, pure mathematics, Futurism, and Action Painting.

Spalter’s work is housed in the permanent collections of the Victoria and Albert Museum (London, UK); the Albright-Knox Art Gallery (Buffalo, NY); the Rhode Island School of Design Museum (Providence, RI); and others. In March 2016, Spalter received accolades from Forbes, Surface, Whitewall, and others for her large-scale installation Precession at SPRING/BREAK Art Show. Later that year, she was tapped by PULSE Contemporary Art Fair for its inaugural commissioned installation series, debuting at PULSE Miami Beach 2016. Also in late 2016, MTA Arts commissioned Spalter to create a 52-screen digital art installation, New York Dreaming, in one of its most crowded commuter hubs (on view through Summer 2017 in Fulton Center). Spalter currently sits on the board of the New York Foundation of the Arts (NYFA).

To learn more about Anne Spalter and to experience her artwork, please visit:

Anne will be exhibiting her work with curator Natalie White at this year’s SPRING/BREAK Art Show 2018 in NYC, March 7-12.  Get passes now!