At the Museum of Modern Art in New York, Pierre Huyghe’s UUmwelt turns the sculpture garden into an unstable field of machine-generated mental images. The work, on view through November 29, does not use artificial intelligence as a shortcut for making pictures. It treats the image as a living situation shaped by brains, sensors, biological data, visitors, and computation.
Huyghe developed the project with neuroscientists in Kyoto. A participant imagined a series of objects while an fMRI scanner recorded brain activity; an artificial neural network then produced thousands of visual interpretations of those signals. The results flicker across freestanding screens as fragmentary forms that seem to approach recognition and then slip away.
Between thought and reconstruction
The premise is seductive: could a machine show us what another person imagines? But UUmwelt refuses the certainty of mind reading. The images reveal an act of reconstruction, not direct access to thought. They expose the distance between a neural signal, a model’s statistical operations, and the human desire to recognize a subject.
That distance is where Huyghe’s art lives. His title expands the German idea of an umwelt—the perceptual world available to a particular organism—by adding another “U.” The alteration suggests an environment in which human and machine cognition continually disturb one another rather than occupying separate realms.
An artwork that keeps changing
The installation is not a fixed film. Sensors register visitors’ gazes, while data from virtual simulations of cancer-cell mutations helps determine how new images are made and sequenced. MoMA describes the work as drawing a metaphorical connection between algorithmic proliferation and metastatic growth.
That association gives the work an ethical tension. Generative systems can multiply images with extraordinary speed, but multiplication is not the same as meaning. Huyghe makes the system visible as an ecology of inputs and transformations. The artwork is not only what appears on the screens; it is the network of people, scientific techniques, software, organisms, and institutional space that allows those appearances to occur.
What artists can take from it
For artists working with AI, UUmwelt offers a demanding model of authorship. The artist’s role lies in framing a question, assembling collaborators, choosing the system’s inputs, shaping its behaviors, and deciding how viewers enter the work. Technical novelty is subordinate to the conceptual structure.
It also demonstrates why transparency can deepen rather than diminish mystery. Knowing that fMRI data, neural networks, sensors, and cellular simulations are involved does not solve the images. It clarifies the conditions under which the uncertainty was produced.
Huyghe’s current presentation is hosted by MoMA, one of the world’s leading museums of modern and contemporary art. Its placement in the sculpture garden matters: the work meets weather, bodies, plants, architecture, and passing time. A screen-based installation becomes environmental, and an apparently disembodied technology is returned to a physical world.
The result is less a picture of artificial intelligence than a proposition about perception. Every image is made by a situation, and every situation contains forms of agency we may not fully see.
