It is now four years since Refik Anadol was awarded The Lumen Prize for his work, Melting Memories, comprising “data paintings, augmented data sculptures and light projections.” He has since introduced a wider public to the power of human-machine collaboration, inducing a flow state in audiences at major institutions and public spaces around the world. 

Given the importance of aura to the history of modern art, his MoMA project, Unsupervised, was ideally suited to inducting a new, “born digital” generation into the art of the museum. By assimilating 138,151 pieces of metadata from MoMA’s own collection, the work extends Anadol’s long-standing “Machine Hallucinations” with a meditation on art history. It also reveals data as a vehicle by which collected pasts can train a brighter future, in theory at least. Here, the artist challenges us to look beyond the surface with Alex Estorick.

‍Alex Estorick: Do you see a greater willingness right now to accept practices that might once have been classed as “new media” within the bounds of contemporary art? Are we living in an expanding art world?

‍Refik Anadol: I think that’s definitely a big “yes.” My journey with data started almost 15 years ago in 2008. Back then, Peter Weibel, who founded the ZKM Center for Art and Media and was a mentor to many media artists, was also a part of my undergraduate school. I’ve been incredibly inspired by ZKM and Peter Weibel’s approach to media art, which he never called “new media.” 

We had a deep discussion about why new is a very limited word in a world that is constantly changing.

That same year, I coined the term “data painting,” which was very dear to me and exactly the practice, exactly the world, that I wanted to explore both in my heart and in my mind. Over the last 15 years, I’ve witnessed a lack of acceptance of work made by humans and machines as a form of art. But of course the more I have deep dived into such concepts and uncovered new worlds with my team, every single gatekeeper, every single naysayer has been proved wrong. 

At the moment, having a show at MoMA, and planning major shows across many different museums, the good news is that the world is expanding and media art is being accepted. This applies not only to me as an artist working with AI — digital art now has profound acceptance, and a future. 

‍AE: Your project, Unsupervised (2022), has introduced a new generation of MoMA visitors to digital art and to the potential of machine intelligence as a creative collaborator. What can we learn about the invisible activities of a machine from its hallucinations — from charting latent space? 

‍RA: First of all, Unsupervised is a whole new world. I’m grateful to my mentor Casey Reas, a pioneering media artist who, alongside Ben Fry, invented Processing. In my mind, that was the starting point for creative coding. I’m one of his students, and I’m also one of his colleagues at UCLA’s department for Design Media Arts, where I’ve taught for the last eight years. 

At MoMA, it has been incredibly important to me to meet with curators Paola Antonelli and Michelle Kuo. The project is completely uncharted in many dimensions. It started during the pandemic in the form of a blockchain exhibition and subsequently in the physical world, where we’ve been able to unleash many more dreams. 

I’ve always been looking to produce a living artwork that is real-time interactive with the world around it. To that end, Unsupervised not only explores machine learning algorithms in a more interactive way, it also relies on audience interpretation while exploring the concepts of chance and control. 

We used NVIDIA’s newest DGX station — an AI supercomputer — to train and inference with AI. That took us to the edge of computation, transforming the idea of AI data sculptures and paintings to new heights. For any living artist, MoMA is one of the most important museums in the world and I’m grateful every single day for this unique project. For our studio, it has not only put us into history, it has also allowed us to bring new generations to the world of museums. For that we are deeply grateful both to MoMA and to the visitors who have allowed this project to become an historic work. 

‍AE: The MoMA project also allows visitors to obtain an NFT as part of the experience. What, for you, is the relationship between public art and private ownership in Web3? 

‍RA: The MoMA show received an incredible audience and has now been extended four times into a year-long show. But we also love the concept of the memento because we believe that both the project and the role of the blockchain are historic. Thanks to MoMA, we’ve been able to experiment further in this regard. 

I believe that Web3 is a much more open, honest, straightforward, and transparent space in which to create work. 

In Web3, I’ve created works that are more accessible than is common in the classical art world, while building a new community around my work. Right now, we have 15,000 people in our Discord, who I’m very happy to be connected with. Another example is our project, Living Architecture: Casa Batlló (2022). Before being auctioned as a single edition at Christie’s, it was a public art installation experienced by 65,000 people.

‍AE: Having trained AI models with numerous vast data sets in recent years, what have you learned about your audiences around the world and, specifically, how different data and metadata register with human viewers?

‍RA: When I had my residency at Google AMI in 2016, the first question I challenged myself with was: “if a machine can learn, can it dream? Can it hallucinate?” To me, that was much more inspiring than AI mimicking reality. I found it much more artful and more profoundly imaginary to think about the cognitive capacity of dreams and hallucinations, which are based on our real memories. 

In the last seven years, we’ve worked with more than four billion images, sounds, and texts. We’ve also trained more than 300 AI models. I’m happy to say that our focus was only on the collective memories of humanity, such as nature, space, urban [life], and culture. All the data and metadata is unique and each has created a unique experience. [...] Even though the fundamental aesthetics are similar, they have also produced very different movements, patterns, colors, and behaviors. 

I’m obsessed with the idea of data pigmentation or fluid dynamics, which is a very common aesthetic in my work [whereby] hundreds of millions of molecules touch and transform each other in a non-Newtonian manner. 

‍AE: How do you respond to those who argue that your works evacuate the meaning, indeed the politics, of a particular data set in favor of its visual resonance?

‍RA: There are multiple ways of looking at our work. If you’re on the surface, most likely you are only seeing shiny pixels and a bunch of particles moving [from] A to B. But the more that people understand the context of machine-human collaboration and how much work has been done behind the scenes to train neural networks, create data sets, and all the connections necessary to generate the work itself, that’s where I find the depth. 

Bringing that depth to the surface is the journey of the audience. 

I’m very grateful for all the messages I’ve received from around the world, which include personal stories about how people felt when they went through the flow state that our works can create, [which is] an impact on the mind. Following the major success of the piece at MoMA, we are [currently] working with Adam Gazzeley and a team of neuroscientists at Neuroelectrics to quantify the impact of the work on the mind. 

‍AE: At a time when humans are being reconstructed as data bodies, what do you see as the potential social implications of your work, which creates new visual worlds based on collected pasts? What alternative futures are perhaps implicit in your work, which combines art with scientific research?

‍RA: Working with AI means that we are working with past data that we are experiencing now, but which has potential for the future. I call this state “remembering the future.” [Claude] Monet was inspired by the atmosphere to create beautiful clouds using shades of color. But when I am inspired by the clouds, I have data in the form of 110 million images of clouds with which I can train an AI. Then I can use a thinking brush to paint with. What people sometimes miss is that using information from life with AI and data is not so different to artists’ imaginative processes over the centuries. 

To me, one the most profound parts of using AI is the ability to play with multiple time dimensions. 

If I think about a traditional artist, there’s a high chance that, every morning, the brush, canvas, and pigment are the same even while the ideas and imagination are changing. To me, every morning is a new morning, with a new AI model, a new data set, and most likely a new algorithm — scientific research is embedded in our work. That is the beauty of working with computation, data, and AI.

‍AE: As Trevor Paglen reminds us, more images today are produced by machines for machines than any other. As an artist whose work crosses between human and nonhuman domains, how can we ensure the legibility of the latter to the former?

‍RA: There’s only one way to answer this question and that is by demystifying AI and machine decisions in general. That can create a much safer and more secure place for humanity. This is a powerful technology that doesn’t forget. As an artist, my role is to take responsibility. In every exhibition we do, we dedicate one major space to demystifying where the data comes from, which algorithms we use, which AI domains we are researching, etc. 

For our recent project, “Living Paintings,” at Jeffrey Deitch Gallery in Los Angeles, we dedicated a major wall — the largest artwork actually — to the process. Similarly, at MoMA, we dedicated a screen to share how the machine makes its decisions. I regard this as one of the most important, meaningful, and purposeful sides to how we create artworks in a given context and discourse.

‍AE: Where in the world is Refik going next? We’d love to hear about any forthcoming projects that you might be working on.

‍RA: For a year, we’ve been working on a major project and challenge — a magnum opus all of its own — that I’ll be able to explain more about in November. Earlier this year, we also announced our research into the Amazon rainforest at the World Economic Forum. For this, we are partnering with Google and NVIDIA to develop an AI model that will be a gift to humanity.