Many years from now, but surely fewer than one wants to think, those of us who survive ecological collapse and the technocratic reformation of the global economy will remember Pierre Huyghe (b. 1962, Paris). Whatever conscious being roams the Earth in that year will look back at UUmwelt as an indicator of what the art in their brave new world could look like and how it could work. For us, right now, today, at Serpentine Gallery in London, I’m not sure we’re completely ready to comprehend its relevance, but it may also be that UUmwelt has come too late.
UUmwelt begins with a door, behind which is a heavy black curtain. This double barrier is there to keep in the bluebottle flies. An entire colony has been deposited into the gallery, and the flies, which have a two-week lifespan, will breed generations throughout the run of the show. They flitter about the rooms of the gallery, dimly lit, and come into view when seen as silhouettes against small windows set high on the wall, or in front of orange-hued lights arranged in circular patterns on the ceiling, or, most prominently, when swarming grotesquely around five LED screens. Each screen within the galleries cycles through thousands of images in a vertiginous array of melding forms, patterns, and colors—the result of a series of psychological, mechanical, and algorithmic translations, which are explained in a flyer provided by the gallery.
Huyghe worked with a research institute in Kyoto, Japan, where he asked participants to imagine specific images (he won’t say what, except that they relate to a future in which animals and humans converge) while inside an MRI. A picture of the brainwave pattern captured by the MRI, which serves as a map of each source image, was then fed through multiple neural networks, which cross-reference the patterns against a database of known patterns, in an attempt to reconstruct the original image. The structure of a neural network, in simplistic terms, is modeled after the structure of a brain. It uses a vast array of discrete algorithms, designed to fire like synapses when they identify a specific characteristic, for example a shape or color,and so a larger taxonomy can be distinguished from the complex pattern of triggered “neurons.” (The truth is that this metaphor, like high school physics, is a lie of oversimplification—because, for example, biological brains don’t store memory or process information in the same way computers do—but it is helpful in understanding how things work, and is repeated often enough that Huyghe’s seeming adoption of the metaphor isn’t too gross a distortion.) Machine learning, broadly speaking, is the process of aligning this artificial network with the biological one, by sharing thought; more specifically, it is the process of teaching the individual nodes of a neural network the threshold at which to fire. When the MRI image is fed into Huyghe’s neural network, it’s filtered through a multilayered information bank, and, all things stable, the neural network would make a guess and propose a static image of what it believes the human subject was mentally picturing—it would display an image reverse-engineered from the image map.
In the case of UUmwelt, this process is interrupted by the conditions of the gallery’s environment, which Huyghe interjects into the neural network, in effort—as he described to Hans Ulrich Obrist, Serpentine’s creative director, in a recorded lecture—to create “a collective production of imagination between two kinds of intelligences.” The result is that artificially generated images displayed on the screens are rendered mutable by such things as sunlight passing through the windows, flies breeding in real time, lights programmed to shift in hue, as well as a series of scents and an audio track developed from the same brain-wave scans. Even the visitors are captured by hidden sensors and refracted back into the neural network. The result of this complex, even incomprehensible, process is a multitude of revulsions. Thousands of morphing images ooze and meld across the LED walls, the buzzing flies distract and harass you, the scents can be unpleasant or trigger unusual associations, and the walls of the gallery have been sanded down to reveal the painted layers of previous exhibitions—remnants of Arthur Jafa and Leon Golub—in amorphous patterns that appear like mossy overgrowth, while visitors kick up dust left on the floor. The sensory tumult provoked in me an uncomfortable and guarded response, possibly an evolutionary one: an immediate recognition of something unidentifiable but occasionally familiar, something insurmountably other. UUmwelt is a rickety bridge across the uncanny valley.
This same gag reflex was first triggered in me several years ago when similar pictures, this time static, were uploaded into the public consciousness. Reports on Google’s DeepDream neural network surfaced after an engineer had asked a pattern-recognition network to show him what it had learned about images. For years, humans have been training neural networks to identify images, and we wanted to know the characteristics it had developed. This one revealed dreams of electric sheep. DeepDream’s dissected renderings of disembodied squirrels and chimeric dogs, which appeared in major newspapers and magazines, were derided as kitsch, further evidence of the supremacy of art and human ingenuity. Fear of the singularity, the mythical convergence of biological and synthetic intelligences, was allayed. But I, and I suspect Huyghe, too, was fascinated by this strange new kind of image, this category of creativity that takes place away from human eyes in the form of an imagistic ur-language developed by an emergent intelligence. For Huyghe, this developing thought is engaged in a process of becoming, as something can only become in relation to another. Once an intelligence begins to draw distinctions around something, it also begins the process of shaping that thing into a distinct entity. This negation, this casting off—I must be this because I am not like that—is the way by which a being draws the boundaries of everything that inhabits this world, its Platonic state:
There will be a moment where, of course, that as a primal technology, used only to visualize imagination, you can start to think, “What is […] particular to that thing that does not need to be actualized?” Because by actualizing something, then of course you have to go down the ladder of […] that matter, that thing, that painting, or that decision. […] You co-weave the ladder of actuality.
In their hysteric elucidations, these algorithmic webs (specifically, convolutional neural networks, the same kind Huyghe uses in UUmwelt) revealed themselves to be engaged in a process of taxonomic distinction comparable to the one we had developed 17,000 years ago in the caves, when Homo sapiens drew itself up as a species distinct from the animals. It’s possible to walk out of UUmwelt believing the significance of Huyghe’s “ecosystem,” as he calls his recent installations, stems from humanity’s position (“as viewers,” to put it in parlance) relative to this dividing line—that we recognize ourselves on the insignificant side of this taxonomic border, not as what severs but as what is severed. But the significance of Huyghe’s return to the caves is not analogous (artificial intelligence : human intelligence :: human intelligence : animal intelligence), nor is it material, procedural, or historical (artists have embraced images not “made by hand” since at least the advent of photography). Rather, Huyghe’s gross ecosystem is a complex instantiation of another Frenchman’s transgression, at Lascaux.
Georges Bataille’s texts on prehistoric art, on which he’d been writing since the early 1930s, are sometimes treated as irreconcilable with the writer’s “literature of transgression.” But Bataille saw represented in the cave paintings a violent sheering of consciousness, the sacred awareness of the tat tvam asi (the Sanskrit knowledge, “Thou art that”) shattered by imagination. He believed that “Lascaux Man,” a breed of Homo sapiens capable of conceiving itself as a “finished individual being,” emerged from Homo faber, which was defined through tools and work, and this process of becoming was documented in rituals and visual art. “Resolutely, decisively, man wrenched himself out of the animal’s condition and into manhood,” he wrote in 1955. “That abrupt, most important of transitions left an image of itself blazed upon the rock in this cave,” a visible record of the dual births of art—rebellion from work—and the conception of humankind as a being distinct from animalism.
To Bataille, this emergent taxonomy presented the first prohibition and allowed the first transgression: humankind, once made distinct from the animals, can never go back again. The existence of this prohibition is what defined Lascaux Man, as “for an animal nothing is ever forbidden.” In Bataille’s literature, art is a way of pressing the boundaries of socially constructed prohibition, transgressing it in pursuit of the sacred—a concept which lapsed Catholics always believe in, though for him in somewhat gnostic terms. This desire to transgress prohibitions is the source of his fascination with beheadings and sacrifice, with death by a thousand cuts, with incestuous and perverse sexuality—as methods, perhaps even as rituals, for approaching what the Surrealists sought: Benjamin called it “profane illumination.”
The artistic ritual that developed at Lascaux was both an “abjuration of animality,” according to Bataille, and a magical embrace of that subsistent profanity, and the cave paintings provided the earliest evidence for his conception of a cultural advance through transgression. In the paintings at the Well, where the only humanoids appear at Lascaux, he saw a permissible return to this state but which is tempered by a violent finality: a spearman with a bird’s head, his penis erect, has disemboweled a bison, which in turn has gored the hunter. This depiction of mutual “murder and expiation” evidences “the religious nature of transgression which indisputably invested the hunt with significance,” Bataille wrote in Erotism. Only sacrifice could absolve the hunter with the bird-like head, who has admitted his animality and killed one of his own. “Nothing finer,” he wrote, “has been done since.”
Huyghe is not so pure as to believe he corrects this original sin of the cave, but he does hint at reclaiming something lost. Neural networks, like pre-Lascaux Homo faber, communicate in an imagistic protolanguage, and in his use of these algorithmic webs and their inflection with human consciousness, Huyghe muddles the images’ origins and makes them translatable from one category of consciousness to another. A clue as to the artist’s intent is found within the exhibition’s title, a pun made from the German word umwelt, meaning “environment.” In conversation with Obrist, Huyghe invokes the biologist Jakob von Uexküll’s use of the word (it was later picked up by Heidegger, Giorgio Agamben, and other Continental philosophers) to describe “a world that is self-oriented,” and which is navigated by entities who are “co-present” but hold different conceptions of experience, determined, in part, by their biological characteristics—so a human’s umwelt is structured by its being bipedal and having two eyes on the front of its head, two ears pointed outward, a nose too far to easily lift scent from the ground. The second U in UUmwelt, Huyghe said, suggests a bridge across these epistemologies (it’s a small idea, but in it I hear Oṃ, the sacred sound of the universe).
UUmwelt’s mode is a technological form of consciousness, one currently in the middle of its own birth—via machine learning—through an experience of which we can model an understanding of our own. This symbiosis is achieved essentially through confusion: you can’t be sure where the machine’s thoughts intersect your own, and the umwelten converge. Through this “back and forth between biotic and abiotic,” as curator Rebecca Lewin phrased it, Huyghe muddies the boundaries of consciousness, renders them porous and malleable—and a distinction that cannot properly distinguish is a useless one after all. In this coproduction of images, UUmwelt returns you to a semblance of the consciousness before the first transgression.
There is a kind of neural network called a generative adversarial network that is composed of two forces. One side, the generative net makes predictions about what an image should contain, given a known characteristic. (Given a picture is Cubist, how likely is it to contain a fractured surface?) The other half, called a discriminator, checks the guess for authenticity, and allows only the permitted images to pass. In technological terms, Huyghe’s UUmwelt plugs in the human mind into a similar interchange, in such a way that it and the AI alternate between generating and discriminating, an endless feedback loop in the middle of which one is left suspended in a state of stasis. Spend long enough there, as the generations of bluebottle flies do, and perhaps a new umwelt may evolve, an epistemology to fit their unstable conditions.
In the ecosystem’s schizophrenic guesswork, patterns and recognizable figures start to appear. As Huyghe’s neural network responds to the conditions of the environment, the human brain responds to the image through pareidolia. From the Ancient Greek para, which means “beside” and is the root of the Latin parere, meaning “to give birth to,” and eídōlon, meaning “image,” pareidolia is the phenomenon through which people interpret familiar patterns where they don’t actually exist. When it runs rampant, something like Dalí’s paranoid criticism, a “delirium of interpretation,” can emerge. Pareidolia is also the term for the spurious identification of features by a neural network, which hallucinates the features it has been trained to see. But Huyghe’s UUmwelt disrupts this process of pareidolia, and in this way the artificial intelligence refuses to return the final image as much as it denies the same closure from you. If a face forms in the pixel mosaic, or something resembling one closely enough for pareidolia to complete the image, it’s never for longer than a fraction of a second. UUmwelt may represent a hysterical search for meaning, but it’s ultimately a futile one because neither human nor machine has the beginning of a clue, because nothing stable ever actualizes, not on screen and not in the imagination.
The Serpentine lobby, at least, is a familiar space, all white walls and a high desk decorated with pamphlets, guidebooks, and catalogues for sale. Behind the desk, a clean-cut man gives directions to Kensington Palace, while nearby a couple speaks in German. In a photograph, palm trees reach above buildings and a low, gray haze. It’s unmarked by a placard, but I know it’s by Arthur Jafa, probably an image from last year’s exhibition. As I wait there for my partner, a small bluebottle fly buzzes—I hear it above, around the spotlight—swirling, alights and settles on my shoulder. “Mr. Bataille loves flies,” said Breton. “Not we.” I turn to catch it and it’s gone. Outside, we’re once again tourists, infatuated with the unfamiliar, and it’s October, which means, for a brief window this time of year, ladybugs have descended on London.