“Voices,” a recent exhibition by the French artist Philippe Parreno at Munich’s Haus der Kunst, presented visitors with an omnium-gatherum of sights, sounds and weather conditions. Some rooms flickered with films — a winsome cuttlefish here, bits of Goya’s Black Paintings there. Others blinked as light sculptures switched on and off. Dancers moved as enigmatic guides through the galleries; a motorized wall, fitted with a spotlight like a small sun or a giant eye, rolled about in a vaguely predatory way. There was a live video feed from Spain’s Tabernas Desert, and heat lamps kicked in periodically to match Munich air to Spanish temperatures. Reverberating throughout was the calm, competent voice of the German news anchor Susanne Daubner, speaking gobbledygook.
Together these elements created the sense — half unsettling, half reassuring — that everything was connected and nothing was quite as it seemed. And indeed, the exhibition’s wall texts explained that the Daubner voice was coming from an A.I.-controlled system (built on hours of recordings) and that the language it spoke was an invention that responded to real-time data about conditions in the exhibition space. The Spanish landscape was real, but the video was computer-controlled, cutting between multiple robotic cameras and adding color effects in response to environmental sensor data from the site. Coordinating these handshakes between here and there was an unprepossessing rack of electronics, also on display, titled simply “Brain.”
“Art is a game between humans,” the artist Jan Svennungsen wrote in his recent book on generative A.I., “Art Intelligence.” We humans look to art for reflections of human experience, and for most of human history, art has been made by humans; even when we employed sophisticated tools, we could feel secure that the ideas being expressed were our own. A.I. muddies those waters. It may work as a kind of smart pencil, executing images beyond an artist’s own manual skill, or as a sparring partner whose suggestions and misunderstandings push the artist’s ideas in new directions — or it might offer a way to destabilize how we think about intelligence itself.
Millions of users have already experienced A.I. art creation in its most ubiquitous form: the fast, easy fun of typing in wacky prompts and watching funny pictures emerge from text-to-image models like Dall-E, Midjourney and Stable Diffusion. More ambitiously, there are entities like Keke (billed as a “truly autonomous A.I. artist”) and Botto (partly guided by human stakeholders), which churn out digital images for sale as NFTs. But however novel this technology may feel, the images themselves lean toward familiar-looking pastiches — rubbery surrealism, futuristic dreamscapes and pretend brushwork, all rendered with the generic facility of an illustrator whose mind is somewhere else.
What else would you expect? Fast, easy and fun is seldom a recipe for serious art. What draws people to museums is the lure of originality and particularity — things that are unlike other things. Most text-to-image models, however, are trained on material scraped from the internet (including copyrighted work by other artists); their goal is to imitate and approximate. And while art made by hand may captivate viewers with erasures and pentimenti — evidence of ideas being born — the key creative decisions in text-to-image models are usually made in advance (the prompt) and post-creation (Botto’s website once boasted of generating 70,000 images in a week, with just one surviving for sale). The evolutionary model is less like human babies and more like spider eggs.
The effect is less like a call to arms and more like the world’s most awesome screensaver.
The same tools can, however, be trained to grander and more riveting, if not inherently profound, ends — as with Refik Anadol’s “data sculptures.” His team spent years compiling some 100 million images of reefs for “Artificial Realities: Coral” (2023), training an A.I. model to generate images of the machine’s interpretations of coral structures and picking the ones they found inspiring. In the finished work, lollipop-colored polyps surge and sway, seeming never to repeat along a 500-square-foot wall of LEDs — part Frederick Edwin Church, part “FernGully,” part liquid light show. As a paean to one of the world’s most threatened ecosystems, “Coral” has claims to seriousness: “It’s not just about shiny pixels,” Anadol told an interviewer. And yet shiny pixels are the hook, and as undeniably mesmerizing as those undulating, make-believe corals are, the effect is less like a call to arms and more like the world’s most awesome screensaver.
Where are the chill winds that blow through so many conversations about A.I. — the enormity of its potential social disruptions, the troubling opacity of machine reasoning? One job of contemporary art, after all, is to register not just society’s aspirations but its dismay. In Parreno’s work and that of his colleague and compatriot Pierre Huyghe, A.I. takes a very different role: a combination éminence grise and switchboard operator that enables exhibition elements to talk to one another and adapt to changing circumstances. Huyghe’s “Liminal,” shown at the Pinault Collection’s Punta della Dogana during last year’s Venice Biennale, is an A.I. entity that evolves from a state of ignorance by processing information from its surroundings — everything from temperature and humidity to visitor behavior. Huyghe describes it as “a presence without a body,” though its visible manifestation is an A.I.-generated projection of a young, fit, naked woman with no face, groping about in the dark.
Visitors to Huyghe and Parreno exhibitions are faced with a similar task: working to form a coherent picture from scattered inputs. There is no way for a viewer to really understand how weather data might be affecting the Tabernas video feed or whether the sounds that emerge from “Liminal” or the A.I. newscaster are, as the artists claimed, embryonic language. You can believe the proffered explanations or not. Either way, you’ll know you’ve made a choice, which is undoubtedly the point.
Both artists have long made art that invokes the wonder and unease of human intelligence bumping up against alternatives. Huyghe’s exhibitions frequently house living animals, as conscious of us as we are of them. (“I don’t want to exhibit something to someone, but rather the reverse: to exhibit someone to something,” he has said.) Parreno’s Tate Turbine Hall show in 2016 featured giant film screens rising and falling, lights flashing and massed speakers turning in perfect unison — the Rockettes came to mind but so did “Triumph of the Will” — all directed, via algorithm, by the state of a humble yeast culture. In nature, these artists remind us, everything both observes and is observed.
A.I. puts a discomforting spin on this, because we don’t know how the machines’ observations — the word “surveillance” has been applied to Huyghe’s work — may be employed. Parreno’s “Voices” apparently “listens and modulates,” which is cool but also creepy. This feeling of being at the mercy of incomprehensible forces is an old one in art. For 19th-century painters of the sublime like J.M.W. Turner and Caspar David Friedrich, nature was humanity’s foil, immeasurable in its scale and obscure in its motives. Two centuries on, very little in the physical universe remains immeasurable — but if you’re looking to invoke an implacable, nonhuman power, A.I. fits the bill nicely.
Great art has always presented viewers with questions imperfectly answered: It provides space for the essential human activity of trying to make sense of the world. For Huyghe and Parreno, A.I. isn’t a quicker, weirder way of making a picture. It’s a way to build substructures of connectivity — links between elements that dictate what happens, even though they cannot be seen and can be only partly intuited. Think of it as a new playing board for what is still very much a game between humans.
Susan Tallman is an art historian and critic, the author of several books (most recently “Kerry James Marshall: The Complete Prints”) and a regular contributor to The Atlantic and The New York Review of Books.
Source photographs for illustration above: Jana Ilic Stankovic/iStock, Getty Images; Lucas Schifres/Getty Images.
The post A.I. Enters the Museum appeared first on New York Times.
