The algorithm and the mona lisa: When museums dream in code
Imagine an autumn evening at the Louvre. The last visitors leave the galleries, the guards lock the doors, and in the gathering darkness, something strange happens. The paintings do not sleep. Behind the display cases, sensors come to life, algorithms awaken, analyzing every brushstroke, every crack in the varnish. The Mona Lisa, that enigmatic face that has endured five centuries, begins to speak—not with words, but with data. Her smile, once interpreted by legions of historians, suddenly becomes the playground of an artificial intelligence. It generates infinite variations of her expression, futuristic versions where Mona Lisa wears augmented reality glasses, cubist reinterpretations, digital clones that populate virtual galleries. She is no longer a painting. She is a matrix.
By Artedusa
••13 min readWelcome to the era of algorithmic museums, where art no longer settles for being contemplated: it reinvents itself, duplicates itself, questions itself under the impulse of machines that learn, hallucinate, create. But what remains of the museum’s magic when a work can be generated in seconds by a simple text command? When a visitor, equipped with connected glasses, sees the walls come alive beneath their feet? When a curator must decide whether to archive a painting… or the code that created it?
This is not science fiction. It is already happening.
01The night a painting without an author shook Christie’s
On October 25, 2018, an auction at Christie’s in New York marked a turning point in art history. Not because of a Picasso or a Warhol, but because of a canvas titled Portrait of Edmond de Belamy, signed with a mathematical formula: min G max D Ex[log(D(x))] + Ez[log(1-D(G(z)))]. An equation, not an artist’s name. Yet this blurry portrait of a man in a frock coat, his features slightly distorted as if seen through steamed glass, sold for $432,500—nearly 45 times its initial estimate.
Behind this work was Obvious, a French collective of three young men who had trained an algorithm—a Generative Adversarial Network (GAN)—on 15,000 portraits painted between the 14th and 20th centuries. The result? A machine capable of producing hybrid faces, halfway between Rembrandt and a fever dream. But the most unsettling part was not the work itself. It was the question it posed, like a stone thrown into the pond of purists: who was the author of this painting? The programmer who wrote the code? The algorithm that had "chosen" the features? The thousands of anonymous artists whose works had served as raw material?
The reactions were immediate and passionate. Some saw a revolution, a democratization of artistic creation. Others, a fraud. Art critic Jonathan Jones, writing in The Guardian, called the work "soulless" and "fake artistic currency." Yet something had changed that night. For the first time, a work generated by artificial intelligence entered the art market—and the world of museums could no longer pretend to ignore it.
02When the museum walls begin to breathe
If you have ever visited TeamLab Borderless in Tokyo, you know what it means to walk inside a waking dream. The walls are no longer walls: they are liquid screens, digital canvases that respond to your presence. A forest of virtual flowers blooms beneath your feet, luminous fish swim around you, and if you reach out, butterflies land on your skin. The space has no limits—neither physical nor temporal. You are both spectator and actor, and the work exists only because you are there.
Founded in 2001 by Toshiyuki Inoko, TeamLab has pushed interactive art into a radically new dimension. Here, there are no frames, no display cases, no "do not touch." Visitors are invited to become part of the work, to dance, to shout, to leave their digital imprint on the space. The algorithms that animate these installations do more than project images: they learn. They analyze visitors’ movements, adjust colors, modify the trajectories of light particles in real time.
But this magic comes at a cost. TeamLab had to close its flagship museum in Tokyo in 2022, not for lack of success, but because maintaining its installations—hundreds of projectors, kilometers of cables, energy-hungry servers—had become a logistical nightmare. Can a digital museum, as ephemeral as a breath, survive the wear of time? The question is all the more pressing because TeamLab’s works are not designed to be collected. They are experiential, almost organic, like fireworks that fade the moment you stop feeding them.
Yet more traditional institutions are beginning to draw inspiration from this approach. The Centre Pompidou, with its 2021 exhibition UAM (Un Musée en Mouvement), attempted a similar experiment: rooms where works, generated by algorithms, evolved based on real-time data—weather, Parisian traffic, visitors’ moods captured by cameras. The museum was no longer a frozen sanctuary, but a living organism, in perpetual metamorphosis.
03Is the artist dead? The ghost in the machine
In 2019, the ZKM (Center for Art and Media) in Karlsruhe, Germany, organized an exhibition titled Open Codes. At its heart stood a strange work: Memories of Passersby I, by Mario Klingemann. A wooden box, topped with two screens, continuously displayed portraits generated by an artificial intelligence. Faces that had never existed, expressions that seemed both familiar and deeply alien. Some visitors spent hours in front of the installation, fascinated by these faces that were born and died before their eyes, like soap bubbles.
Klingemann, a self-taught German artist, is one of the pioneers of generative art. For him, AI is not a tool, but a collaborator. "I don’t really control what the machine will produce," he explains. "I give it rules, constraints, and then I let it explore. Sometimes, it surprises me." In Memories of Passersby I, the algorithm was trained on thousands of historical portraits, as well as images of contemporary faces. The result? Unsettling hybrids, where the features of an 18th-century aristocrat blend with those of a stranger met on the street.
But this human-machine collaboration raises a dizzying question: if AI can create such convincing works, what is the artist’s role anymore? For some, like Refik Anadol, another master of algorithmic art, the answer is clear: the artist becomes a data architect. In his work Machine Hallucinations, Anadol used millions of architectural images to create dreamlike landscapes, cathedrals dissolving into the sky, cities pulsing like living organisms. "I don’t create images," he says. "I create collective dreams."
Yet this optimistic vision collides with a darker reality. In 2022, controversy erupted when an artist, Jason Allen, won an art competition with Théâtre D’opéra Spatial, a work generated by MidJourney, a text-to-image AI tool. Other participants cried foul. Allen defended his work: "Art is dead, damn it. It was killed by corporations. AI is the renaissance." The debate was launched: is AI-generated art still art? And if so, who is its author—the human who wrote the prompt, or the machine that interpreted their words?
04The museum as a laboratory: when AI restores masterpieces
In 2021, the Rijksmuseum in Amsterdam used artificial intelligence to solve a four-century-old mystery. The Wedding at Cana, a masterpiece by Veronese, had been cut into pieces by Napoleon’s troops in 1797. Only part of the canvas had survived, and curators had wondered for decades what the original work looked like. Using an algorithm trained on hundreds of Renaissance paintings, the AI reconstructed the missing parts with astonishing precision. Colors, faces, architectural details were recreated from historical data, offering visitors an unprecedented glimpse of the work.
This technical feat opens dizzying possibilities. What if, tomorrow, museums used AI not only to restore but also to reinvent the works of the past? Imagine a version of Rembrandt’s The Night Watch where the characters move, where shadows shift with the light. Or a Mona Lisa whose expression changes with the viewer’s mood.
But this revolution also raises ethical questions. How far can we go in manipulating works? In 2020, the Louvre launched Mona Lisa: Beyond the Glass, a virtual reality experience that allowed visitors to "enter" Leonardo da Vinci’s studio. Some loved this immersion. Others cried sacrilege: "The Mona Lisa doesn’t need special effects to be magical," wrote one critic.
Yet museums have little choice. Faced with competition from screens and digital experiences, they must reinvent themselves. The museum of tomorrow will no longer be a place of passive contemplation, but a space for interaction, creation, even controversy. And AI will be one of its main tools.
05The ephemeral and the eternal: how to preserve a digital dream?
In 2018, the MoMA in New York acquired a work titled Unsupervised, by Refik Anadol. An immersive installation where algorithms generate abstract landscapes in real time from the museum’s data. The problem? No one really knows how to preserve it. Unlike a canvas or a sculpture, Unsupervised depends on code, servers, technical infrastructure that can become obsolete in a few years. What will happen when the computers running the work are too old to repair? When the programming languages used are forgotten?
This is the paradox of digital art: it is both eternal and profoundly ephemeral. A work like Jeffrey Shaw’s The Legible City (1989) can no longer be exhibited today because the computers that ran it have disappeared. Museums face a dilemma: should they archive everything, risking turning their storage into graveyards of dead technologies? Or accept that some works are doomed to disappear, like performances or ephemeral installations?
Some institutions are trying to find solutions. The ZKM in Germany has created a department dedicated to preserving digital works, where experts work to emulate old systems to revive forgotten pieces. Others rely on documentation: filming works from every angle, recording interactions, preserving source codes. But these methods have their limits. How do you capture the essence of a work that exists only in the moment?
For some artists, this fragility is intentional. Rafael Lozano-Hemmer, known for his interactive installations, created Disappearing Act, a work that gradually erases itself with each public interaction. "Digital art is like life," he says. "It is born, it lives, and one day, it dies." A philosophy that echoes Renaissance artists, for whom the work was inseparable from its context. But in the age of AI, this ephemerality takes on a new dimension: what if the museum of tomorrow is no longer a place of memory, but a space of flux, where works are born and die in real time?
06The visitor of the future: between wonder and surveillance
In 2023, the British Museum launched a pilot project: Ask the Museum, an AI-powered chatbot that answers visitors’ questions. Ask it about an Egyptian mummy, and it will tell its story. Inquire about a Greek pot, and it will explain the techniques used to make it. But behind this user-friendly interface lies a more troubling reality: every interaction is recorded, analyzed, used to refine the algorithms.
Museums are increasingly tempted to use AI to personalize the visitor experience. Imagine an audio guide that adapts to your interests, an app that suggests a custom route based on your mood, or rooms that reconfigure in real time to avoid crowds. But this personalization comes at a cost: the end of the collective experience. If every visitor sees a different version of the museum, what remains of shared discovery?
Worse still, these technologies raise privacy concerns. In 2022, an investigation revealed that several European museums were using facial recognition cameras to analyze visitor behavior. Some see this as a tool to improve accessibility (by detecting, for example, people with disabilities). Others, an Orwellian drift, where every glance, every hesitation, is turned into data.
Yet one thing is certain: the visitor of tomorrow will no longer be a mere spectator. They will be an actor, a co-creator, even a subject of study. And museums will have to choose: do they want to be temples of knowledge, or laboratories of the future?
07Art after the apocalypse: what if AI were our last museum?
In 2021, Refik Anadol presented Machine Hallucinations at the Serpentine Gallery in London. A work generated from 100 million images of architecture, transformed into dreamlike landscapes by artificial intelligence. But behind this hypnotic beauty lies a chilling question: what if, one day, AI were the only one to remember our civilization?
In a world threatened by climate change, wars, and technological collapses, museums could become the last guardians of human memory. But what will happen when humans are no longer there to maintain them? When buildings crumble, when servers shut down? Could AI become the last museum of humanity?
Some artists are already exploring this hypothesis. Ian Cheng, with his series Emissaries, imagines virtual ecosystems that evolve autonomously, like artificial life forms. Laurie Frick, meanwhile, creates works from personal data, turning our digital lives into abstract paintings. "Art," she says, "may be the only thing that survives our species."
But this apocalyptic vision also has its flip side. What if AI, instead of preserving our memory, rewrote it? If, in a thousand years, an artificial intelligence generated "lost works" by Picasso or Van Gogh, drawing on what it knows of their style? Who could say whether these paintings were authentic? Will the museum of tomorrow be nothing more than a vast deepfake, where the boundary between true and false has definitively vanished?
08Epilogue: the algorithm’s smile
Let us return to the Mona Lisa. Tonight, in the Louvre’s storerooms, an artificial intelligence continues to work. It analyzes every pixel of the painting, every crack in the canvas, every glint of light on Mona Lisa’s lips. It generates thousands of variations, futuristic versions, surrealist interpretations. What if, one day, one of these versions were displayed next to the original? If visitors had to choose between Leonardo’s smile and the algorithm’s?
Perhaps the answer lies in another, lesser-known but equally unsettling work: TV Buddha by Nam June Paik. Created in 1974, this installation shows a Buddha statue watching its own image on a television screen. An infinite mirror, where man and machine contemplate each other. The museum of tomorrow may resemble this: a space where we see ourselves through the eyes of AI, where art is no longer something to possess, but an endless conversation between past and future.
Then, the next time you visit a museum, look closely at the walls. They may no longer be what they seem. They breathe, they dream, they watch you in return. And somewhere, in the shadows of the servers, an artificial intelligence is creating the next work that will shake the history of art.
Unless it has already done so.