How long does it take to truly see a work of art? A 2017 study suggested museum visitors spend an average of 27 seconds with a piece—barely enough time to register color and composition. Yet when media artist Refik Anadol unveiled Unsupervised at the Museum of Modern Art in 2022, audiences reportedly lingered for an average of 38 minutes.

That disparity—between the glance and the immersion—lies at the heart of the storm surrounding Anadol’s work. When the piece resurfaced in public debate on 60 Minutes, it reignited a volatile question: when artificial intelligence remixes humanity’s visual memory, does it create art—or merely simulate it?

Inside Unsupervised: Data as Pigment

To create Unsupervised, Anadol fed an AI model metadata from more than 138,000 artworks in MoMA’s collection. The machine processed titles, dates, and classifications, translating institutional history into a continuous stream of shifting abstraction. On the museum’s lobby wall, colors bled and reassembled. Impressionistic atmospheres dissolved into gestural turbulence. Echoes of van Gogh seemed to melt into Monet, then surge toward de Kooning—an algorithmic genealogy in perpetual motion.

Anadol describes his method in painterly language.

When I think about data as a pigment, it doesn’t need to dry. It can move in any shape, any form, any color, and texture.

– He told correspondent Sharyn Alfonsi.

For him, data is not cold information but malleable material—an extension of the palette.

Visually, the installation resembles an enormous digital fresco. Swells of color pulse like breathing organisms. Edges liquefy. Forms emerge only to collapse into chromatic mist. The effect is hypnotic—part cathedral projection, part psychedelic screensaver. Visitors lounged on couches. Some danced. Many filmed. The museum lobby became less a transit space than a digital amphitheater.

Glenn Lowry, then MoMA’s director, characterized the project as an extraordinary success. It brought bodies into proximity with abstraction at a time when painting often feels hushed and rarefied.

The “Lava Lamp” Critique

Not everyone was entranced.

Art critic Jerry Saltz famously described Unsupervised as a “massive techno lava lamp,” likening it to a high-end screensaver. On 60 Minutes, he conceded that AI will one day produce art of consequence. For now, he argued, much of what we see amounts to “an average of averages”—a smoothing out of human invention into algorithmic consensus.

Saltz’s critique strikes at authorship. If the system recombines preexisting data, does it generate insight—or only pattern? Is immersion equivalent to meaning? A lava lamp can captivate for hours without saying anything at all.

His skepticism also gestures toward art history’s cyclical nature. The spectacle of light and motion has precedents—from psychedelic projections of the 1960s to early video art experiments. If novelty lies primarily in scale and processing power, is that enough?

The Ethics of the Archive

The debate sharpens when questions of consent surface.

Artist and writer Molly Crabapple, also interviewed on 60 Minutes, described AI training datasets as “the greatest art heist in history.” Billions of scraped images, she argues, have been absorbed without permission. Where museums guard individual masterpieces, algorithms ingest entire visual cultures.

Anadol counters that he now works exclusively with what he calls ethically sourced datasets and insists on treating AI as a collaborator. His ideal is a 50–50 partnership between human and machine: the artist sets parameters, curates outputs, and sculpts the final experience.

Yet the tension persists. Even when datasets are institutional—like MoMA’s metadata—they represent accumulated labor: curators, artists, historians. The algorithm’s “creativity” emerges from a scaffold built by countless human hands.

Spectacle, Attention, and the Museum of the Future

Beyond authorship and ethics lies a subtler transformation: the recalibration of attention.

If visitors spend 38 minutes with an AI-driven projection, what does that suggest about contemporary spectatorship? Unsupervised functions less like a painting and more like an environment. Its scale and movement invite bodily response—lounging, dancing, filming. The museum lobby becomes social, performative, shareable.

In this sense, Anadol’s work aligns seamlessly with digital culture. It thrives on documentation. It circulates online as loops and clips. Its mutability mirrors the endless scroll.

But spectacle has always shadowed innovation. The question is whether AI art can move beyond astonishment toward sustained critical depth. Can machine-generated abstraction grapple with grief, power, memory—the enduring preoccupations of art?

A New Frontier—or Familiar Glow?

Anadol maintains that we are discovering a place we have never been before. Saltz insists we have visited this territory already; it merely glows brighter now.

Perhaps both are correct.

Unsupervised occupies a threshold moment in art history. It exposes the museum as database, the archive as raw material. It reframes authorship as orchestration. It reveals how quickly technology can recalibrate aesthetic experience.

Whether one sees a visionary leap or a luminous gimmick, the installation has achieved something undeniable: it has made the art world argue—publicly, urgently, and at scale.

The lava lamp metaphor lingers, half-dismissive, half-accurate. Lava lamps mesmerize because they transform physics into poetry—heat into motion, wax into drift. Anadol’s algorithms perform a similar alchemy, turning data into spectacle.

The unresolved question is not whether AI can generate images. It already can. The deeper inquiry is whether those images can carry intention, risk, and responsibility.

As museums navigate this new terrain, the glow of Anadol’s shifting abstractions continues to pulse—provocative, polarizing, and impossible to ignore.