Art(ist) in the Machine
A new artwork at the MoMA delivers a message about art created with, and by, technology.
If you can have a conversation with a simulated person presented by an AI program, can you tell how far you’ve let your sense of personhood degrade in order to make the illusion work for you?
―Jaron Lanier, You Are Not a Gadget
Last week my son Dino and I traveled to New York City, and, as usual, our first visit was to the Museum of Modern Art (MoMA). Soon after entering, we encountered an ample open space where people sat and looked at images projected on a large wall. As we approached, I noticed lines of code on a computer screen on a wall, and I instantly recognized the work of the artist Refik Anadol, whose creations I have followed for years.
Anadol is a pioneer of what might be called “data art”—visual art built from various forms of information ingested and processed by computers. The work at the MoMA, called “Unsupervised,” is part of Anadol’s “Machine Hallucinations” series that he began in 2016. For “Unsupervised,” Anadol and his team—which included engineers from Nvidia, the computer chip company—“fed” a proprietary Artificial Intelligence (AI) machine 380,000 high-resolution images from more than 180,000 artworks (spanning over 200 years of art) stored in MoMA’s galleries.
The AI was programmed to absorb the images and metadata and then create new images in the highest resolution currently possible. The AI does so in response not just to its own programming but also to two external factors: real-time weather data just outside MoMA and the motion of the people in the space in front of the screen. Because both external factors are constantly shifting, the images visitors see are constantly new. The video below provides a sense of what the AI is doing as it processes all its information, a process recorded both in video and as snapshots of the AI’s “state of mind.”
“Unsupervised” AI process visualization (Video: MoMA)
For Anadol, the spontaneity and unpredictability of the AI’s creation are the main attractions. “[My use of AI] like its own entity,” he recently said. “We don’t know what kind of forms it can create.” As with any constantly changing form, the effect on the viewer can be hypnotic and calming. We sat in front of the giant screen for several minutes transfixed by the ever-changing forms and color combinations morphing before us.
At first, it was fun to try—albeit without much success—to link what I was seeing at any moment to a specific work of art or artist. Connecting to familiar works of art was impossible, however, because there is no direct visual connection between what the AI is showing and the works it has ingested. Whatever “elements” a Rothko painting or a Brancusi sculpture contributed to AI’s artistic storehouse have long been dissolved into something altogether new and different. After a few minutes of looking at “Unsupervised,” I found myself becoming restless and bored. Once the novelty of the creation had worn off, I realized that the ever-shifting shapes did not really have anything to say to me, or to most of my fellow visitors who, like me, seemed to lose interest in the creation after only a few minutes or seconds. I contrast my reaction to this work with the way I felt when I saw Mark Rothko’s grey and black color field painting elsewhere in the museum: I never grew tired of looking at it, and its deep colors seemed almost to vibrate with life the closer I approached the canvas.
“Unsupervised” may be a technical tour de force, but it does not rise to the same level as art, per se. Though novel, it does not address anything essential; it seems, in fact, not to know what it is trying to be or say. Reflecting on my conclusion about the work, I recalled the first Anadol piece I came across, his 2017 creation for SALT Research that used AI to search and sort “relationships” among the 1.7 million documents in the library’s collection. Entitled “Archive Dreaming,” the “AI data sculpture” installation “dreamt” the connections, and a visitor could stand before it and interact with specific documents and help the AI map the relational journey.
“Archive Dreaming” exhibition design (Image: www.rafikanadol.com)
Perhaps because I am a writer and not a painter, “Archive Dreaming” made more sense to me than the work I saw with Dino. I could imagine spending hours looking through the pages of the archive and seeing what connections the AI made in another of the library’s works. Crucially, in “Archive Dreaming,” the original works remain recognizable, and this fact helps the viewer process and internalize the relationships the AI discovers.
The more I thought about “Archive Dreaming,” the more I recalled Jorge Luis Borges’ short story, the Library of Babel. In the story, a vast archive contains books that comprise every possible permutation or combination of 25 symbols: 22 letters, the comma and full stop, and the space. All the books are precisely 410 pages, and each is different from all the rest. Contained within the vast collection are unique books called “Vindications” that hold special knowledge beneficial to humanity but that the collection’s librarians cannot find in the vast multitude of works. The Library is both helpful and useless, finite and infinite. Anadol’s “Archive Dreaming” struck me similarly: it is fascinating but serves no purpose other than to invite one to search for yet another relationship, hoping to find meaning amidst all those words.1
The fact that the original works of art are totally lost in “Unsupervised” may be what keeps it from being more than it is. We come to museums, in part, to behold creations that seem to have been produced by super-humans—objects of supreme skill and creativity that speak across time and space. The messages in MoMA’s art treasures are lost in Anadol’s AI translation, which is unfortunate for a hallmark piece in a museum of MoMA’s caliber. As New York Times' art critic Travis Diehl noted: “The work’s enveloping abstraction scuttles the possibility of conceptual dignity offered by an institution founded to embrace challenging new art.”
Unkindness to Anadol’s MoMA installation is not limited to a single critic, with one even humorously describing it as “an extremely intelligent lava lamp.” I share some of the critic’s disappointment, and it would be easy enough to call “Unsupervised” a giant screensaver or high-tech curiosity and be done with it. Then again, I think Anadol’s piece may be more prescient, if we look not at what it is today but what a future iteration may be years from now. Perhaps Anadol’s installation’s unintentional message is that AI left “unsupervised” can create amazing transformations, but they are somehow less than human, and so further dehumanize us—the exact opposite of what great art should accomplish. As the writer Jaron Lanier (whose quotation opens this essay) has noted:
Individual web pages as they first appeared in the early 1990s had the flavour of person-hood. MySpace preserved some of that flavour, though a process of regularized formatting had begun. Facebook went further, organizing people into multiple-choice identities while Wikipedia seeks to erase point of view entirely. If a church or government were doing these things, it would feel authoritarian, but when technologists are the culprits, we seem hip, fresh, and inventive. People accept ideas presented in technological form that would be abhorrent in any other forms.
If Lanier is correct, “Unsupervised” may be a deep warning disguised as shallow art: in a new world defined and ruled by AI, where the artist is subsumed by, and exists only in, the machine, MoMA’s future audiences may never make it past the foyer.
Unsurprisingly, Anadol said in a TED Talk that Borges’ story inspired “Archive Dreaming.”