I had a poster of Duchamp’s Nude Descending a Staircase, No. 2 in my bedroom as a child and would stare at it before sleep. It would come alive before my eyes—the fractured figure seemed to move down the canvas. That hallucinated movement is where this work begins.
In the 1870s, Eadweard Muybridge froze motion into sequential photographs. In 1912, Duchamp compressed those sequences back into a single canvas and scandalized the Armory Show with a painting that depicted time rather than a body. I wanted to take a third step: not freezing motion, not compressing it into stillness, but extending Duchamp’s fractured figure into actual duration.
Duchamp is a rare case—an artist inspired by a technological forerunner like Muybridge for the subject of a painting, and each generation’s emerging technology enables the next artist to extend what the previous one could only imply. I recorded ten people descending a staircase, then fed their collective movement through a machine learning program to distill a single generic figure performing a generic descent. The algorithm’s imperfections produced visual artifacts that echoed cubist fragmentation—limbs dissolving and reconstituting mid-stride, shimmering with instability, as though Duchamp’s fractures had been set in motion rather than resolved. What the machine reveals is that cubist fragmentation was always a theory of motion waiting for its technology.
Ten individual bodies averaged into one: the algorithm distills particularity into data, just as the systems shaping our present flatten individual experience into generalized patterns. Whether this distillation represents inclusiveness or erasure is a question the work holds open.
The work’s deepest resonance is circularity. John Cage composed “Music for Marcel Duchamp” in 1947 for a sequence in Hans Richter’s film Dreams That Money Can Buy, in which Duchamp’s original Nude Descending was animated. I selected Cage’s composition as the soundtrack—and so the historical circuit closes: Cage wrote music for Duchamp’s animated nude, and now the same composition accompanies a machine learning reanimation of that figure. His aleatory approach, composing through chance operations, resonates with my own use of machine learning’s unpredictable outputs.
The work premiered during Frieze New York in May 2019 at the Oculus—Santiago Calatrava’s station at the World Trade Center—across twenty-one screens, one four stories tall, another 280 feet long. The venue was essential. Hundreds of thousands of commuters pass through daily; art in a transit space becomes an encounter rather than a pilgrimage. I have always valued placing work where people come across it in motion, without seeking it—something like a digital readymade, where selection and context matter more than conventional craft.
The figure descends perpetually, looping without resolution. A loop is not a narrative—it does not move from beginning to end. It offers continuous transformation, a kind of time you feel rather than measure. What began as frozen photographs became a fractured painting, then algorithmic motion that retains the analytical quality of cubism while embodying the duration Duchamp could only diagram.
Marco Brambilla