A New Experiential Gallery Just Might Change Your Mind About AI Art

“I think we are literally in a renaissance,” says the artist Refik Anadol, in a characteristically optimistic comment, when asked how he sees this moment in art history, with artificial intelligence ascendant yet controversial as a medium. “We just don’t have a name for it yet.”

Anadol, known for technological installations that probe the relationship between humans and machines, has reason to be happy. On June 20, Dataland, the cutting-edge downtown Los Angeles gallery he cofounded with studio partner Efsun Erkılıç, opened its doors to an eager public. Billed as the first “museum of AI arts” in the world, it welcomed more than 10,000 visitors to the opening exhibit in the first two weeks, Anadol tells WIRED.

The set piece is his most ambitious to date, an immersive architectural vision titled Machine Dreams: Rainforest. Its interactive digital displays, directly responding to the visitors’ movements and biometric data (tracked by wearable devices), produce ever-shifting images and soundscapes drawn from Anadol’s Large Nature Model, an AI system built using natural science archives from prestigious research institutions like the Smithsonian.

“For three years, we started from scratch and trained our own AI models, and we worked with our own data sets,” Anadol says. He and his team traveled to the Amazon and other rainforests to capture raw material that would fuel the model’s hallucinated versions of those environments. “We have 5 petabytes worth of raw data that we collected by ourselves,” Anadol says. He’s proud that Dataland made a point of sourcing this trove with the consent and participation of researchers, whereas Silicon Valley’s major AI firms have faced backlash and lawsuits over what many creators say is unlicensed, extractive use of their content as training data.

Anadol adds that Google DeepMind gave Dataland access to “experimental low-energy” resources, allowing the gallery to run on Google Cloud and maintain “sustainable compute.” (Anadol has collaborated with the tech giant since becoming the first person awarded the Google Artists and Machine Intelligence Artist Residency in 2016.)

Ethics, environmental responsibility, and a wholehearted effort to produce what feels like a living, breathing ecosystem with artificial intelligence: These commitments are crucial if Anadol and Dataland want to redefine “AI art.” The very phrase is a nonstarter for many creatives and critics of the generative “slop” that has infested visual media at every level. Anadol is perfectly aware of people rejecting that stuff, and he hardly blames them. “I mean, 100 percent, the majority is right,” he says, noting that when someone hears about AI art, “their first assumption is like, prompt engineering, or a bunch of eight-second clips.”

His new gallery aims to prove that there are grand possibilities beyond all that. “The reason for Dataland is to understand and explain and explore—and tell the world that there is not only one option,” he says.

Machine Dreams: Rainforest makes an impressive case for AI as a tool to unlock surprising modes of artistic engagement. (Anadol claims that it is his first project where “it is impossible to record what it feels like,” and it’s no exaggeration.) After going through a somewhat finicky entrance procedure involving a waiver and a specialized app, I receive a smartwatch and a U-shaped plastic collar to be worn over my shoulders. As these are calibrated, the shoulder harness emits the first of the exhibit’s many striking scents: here in the heart of this hypermodern building, I can suddenly smell trees.

The largest gallery is empty but for a couple of pillars, with ultravivid fusions of nature scenes and computer-chip textures swimming across the walls and floor; there’s a 40-minute cycle of varied effects that I quickly realize are partly dependent on the movements of everyone in the room. I enter during a sequence simulating heavy rain and thunder and see that each visitor has a watery circle around their feet that travels with them, becoming an aqueous streak when they walk faster. The paths of droplets in the downpour shift if you wave a hand in front of them. There’s even the clean smell of a summer storm emanating from my biosensor.

Anadol says that while you have the option to move through his artwork as a “ghost”—that is, without the biosensors—the wearables, which are modified medical grade devices, offer a way to put your fingerprint on it, at least temporarily. “For almost 5,000 years, we as humanity, we look at artworks and we feel something,” he says. “At Dataland, as a laboratory of imagination, our first deep question is: Can artwork feel us back?”

Anadol laughs when I wonder aloud whether the gadgets tracked my trip to the restroom. “No, no, no, no, the restroom has zero connection,” he says. Moreover, he says, the museum “forgets” your data the moment you leave, though it remains available to the visitor by way of a personal token received at the exit. “Data is a form of memory,” Anadol says, and the gallery seeks to treat it with respect. “So it’s the opposite,” he notes “of what we have in the whole world,” where invasive surveillance has become the norm.

Farther on lies the Infinity Room, which invites me to follow a glittery hummingbird as it soars over a fantastical neon forest—the flight puts one in mind of the aerial acrobatics in the Avatar movies, and at times I find myself a little off-balance when the POV careens one way or the other. I’m more intrigued by an adjacent wing, the Latent Gallery, where I can take a peek behind the curtain of the Large Nature Model. At one console, for example, I can scroll through the training data by category. I click “Frogs,” and the wall in front of me turns into a massive grid of frog photos included in the model.

Anadol says he wanted people to have a glimpse of the profound wealth of information underlying the often surreal and impossible permutations of actual nature generated by the installation. (The Smithsonian’s Encyclopedia of Life alone supplied data on more than 2 million species.) “These are the places where we demystify our algorithms, demystify our data sets, demystify our training experiments,” he says. Anadol says he wanted to see “what happens if we go next-level and tell the audience and the visitors exactly the reason for that machine dream or hallucination.”

And hallucination is certainly the point, which also serves to differentiate Dataland’s debut show from the ultra-realistic AI-generated images and deepfakes that contribute to our culture of misinformation. “It's about the idea of a machine that falls in love with nature—that's the rainforest—and we enter the machine dreams as a concept where we smell the machine's hallucinations, we taste the machine dreams, we hear millions of bird songs that are impossible to hear in a human world,” Anadol says. The model may incorporate scientific libraries and real-time weather data, but the place it constructs with these building blocks is, in most respects, totally alien.

Finally, you have the Sanctuary, based on a concept Anadol has been working toward for several years. As a group enters the room, the biometric information from each individual—everything from heart rate and skin temperature to the path they have taken through Dataland and the speed of their tour—is compiled to generate a swirling, abstract, 3D representation of the room’s collective energy, never to be seen again.

Anadol says that here and elsewhere, the gallery takes “emotion as an input.” He’s been especially pleased to observe how frequently the biosensors pick up on visitors’ goosebumps, which he calls an “incredibly relevant” sign that the sensory biomes of the exhibit are having a pronounced effect on those who pass through them. “The artwork can feel this information,” he says. “I saw people with tears, I saw people with joy, I saw people with excitement,” Anadol adds. “And I think, if that's not art, what is?"