The Brooklyn Rail

Pierre Huyghe’s exhibition, In Imaginal at Marian Goodman Gallery, is the US debut of a selection of intermedia works that premiered at the Punta della Dogana, Venice, in 2024. The installations utilize deep learning processes that are networked to trigger each other. In a 2024 interview with curator Anne Stenne, Huyghe remarked that his practice is directed by counterfactual worldbuilding—creating “other possible and impossible worlds” that betray “an occult side.” Huyghe’s artistic process involves systems-theoretical relations. Initially, he models first-order, human-computer interactions, which license his video projections and sculptures. A number of these—chiefly, Huyghe’s pairing of the Cosmoseeded (2001) lamp sculpture and the film, Annlee – UUmwelt (2018–24)—are set into a second-order relation, with their signals responding to one another in intercomputer feedback loops. Drawing on the principles of second-order cybernetics and systems theory, the masks composing Idiom (2024), installed in various corners throughout Marian Goodman’s two gallery floors, operate on a third-order register. Their semiologically unstructured robotic utterances respond both to one another’s grumbling speech acts and those exacted by human gallery-goers.

The video, Annlee – UUmwelt plays in the first gallery. Three bulbous, arciform sculptures, collectively titled Mind’s Eye (2021) are installed behind the film screen, with each mound constructed from what is described as “aggregates of synthetic and biological matter.” The video and the sculptures have their source material in Huyghe’s recurrent “Annlee” character, a purple-haired, almond-eyed anime sprite/avatar. In 1999, Huyghe and his close collaborator, Philippe Parreno, purchased the rights to Annlee from the Kwork animation studio’s catalog. Since then, Huyghe has often returned to this motif, albeit the resemblance between the original anime and his most recent video imagery and sculptures are heavily obscured. The Annlee – UUmwelt video interpolates Annlee through several registers of deep image reconstruction, her eyes multiplied across her contorted, pixel-pocked body and her cranium ballooning into crazed, twisting forms. In creating this video, Huyghe collaborated with Kyoto University’s Kamitani Lab, which took fMRI scans of a percipient who was prompted to watch or imagine a series of images. The scans were correlated to a dataset of images and trained in a neural network such that the subsequent AI would produce second-order images based on the training data. The sculptures are the byproduct of isolating a static frame and rendering it three-dimensional. Where two Mind’s Eye formations recall stalactite and mineral-punctuated boulders, one of the sculptures, the aptly titled Mind’s Eye Annlee (2021), betrays facial indices. A set of equidistant black eyelets are set in a coral crystal ovoid such that one can faintly make out Annlee’s visage. In admitting its source material through several stages of networked mediation, this is the most interesting of the three sculptures.

The adjacent room is dimly lit, with a flickering axial light source looming from the ceiling. This roseate, fleshy bulb is a phosphorescent lamp titled Cosmoseeded. It occasionally shuts off, in coordination with the Annlee – UUmwelt video’s halting. In the corner, one finds the first of several Idiom masks (donned by performers on select days).

Upstairs, Huyghe’s film, Camata (2024), is the centerpiece of the show. The film is self-editing, auto-generating, and continuous. Like any AI-generated artwork, the “end-to-end” training involves a constitutive dataset that is not (or at least not entirely) AI-generated. In Camata’s case, these include various scenes filmed in Chile’s Atacama Desert, the world’s driest non-polar desert, which serve as the unstructured dataset for the algorithm’s supervised task. Skeletal remains found in this desert, which Huyghe photographed in 2016, first drew him to this ethereal locale. The skeleton’s birth is estimated to be between 1834 and 1894, and its identity is believed to be that of a soldier who fought during the War of the Pacific (1879–83). The film stages a funeral rite for the remains. The camera pans from long-range shots of the clay-orange desert field to close-up scenes of variegated minerals, stones, salt flats, and mining tunnels, returning, semi-regularly, to various images of skeleton pieces flanked by tattered overcoats, a ragged boot, and unidentifiable cloth shreds. Three robotic arms and a carbon slider track circumscribe the skeleton. Two of the arms are equipped with a motorized hand that positions glass spheres in various configurations around the sand-interred skeleton, never quite touching it.

Staging the camera-arm amongst the whirring robotic, Camata’s long-range shots simultaneously frame the rite and the process of filmic recording. Moving from second-order to first-order vantages, the perspective trades distanced observational glances with interior views in which we, occupying the circumscribed cameras’ intimate standpoint, draw close in and away from the glass orbs, cerulean mineral stones, and ulna bones. In its self-editing, Camata consolidates false continuities, aberrant camera movements and absorbed framings that pass from intimately observed sand burrows to bird’s eye perspective studies of the desert vista.

Where the film makes viewers aware of camera movement, reflecting on the camera-eye’s way of seeing, one is reminded of structuralist cinema like Michael Snow’s La Région Centrale (1970–71), which similarly isolates the phenomenon of camera movement—the vernacular that film critic and director Alexandre Astruc, in his famous L’Écran française article, “Du Stylo à la caméra et de la caméra au stylo” (1948), called the camera-stylo (camera-pen). Camata, in its juxtaposing a formalist study of the camera-stylo with the thematic content of the ritual, reveals an inherent tension that extends to Huyghe’s work tout court. At first pass, Huyghe appears to be a post-humanist interested in effacing the human’s intentional processes and decrying anthropocentrism, revealing pure mechanical processes that are mind-independent and ontologically constitutive. But his program also remains ethnographic, anchored as it is in a set of humanist meanings, ranging from the significance of a found skeleton to the funeral rite. Relatedly, the AI training datasets remain intention-bearing by dint of their human source material. Mental pictures ground Kamitani’s fMRI imaging and, in turn, inform both the Annlee – UUmwelt imagery and the Mind’s Eye sculptures.

In terms of subject matter, Huyghe’s self-editing and self-generating devices are significantly more complex than those deployed by those early to mid-twentieth-century Dadaists and Surrealists who similarly sought to incorporate chance operations. In both form and content, Huyghe’s sculptures and video installations sit at the cutting edge of new media affordances. Unlike his Dadaist or Surrealist antecedents, Huyghe’s theoretical framework is as far from Sigmund Freud, Carl Jung, and Jacques Lacan as it is from the esotericism of Pyotr Ouspensky, Raymond Roussel, and Erik Satie. In his apparent rejection of anthropocentrism and sympathy towards panpsychism, Huyghe’s theoretical interests are closer to Tristan Garcia’s Object-Oriented ontology, Alfred Whitehead’s Process Philosophy, or Johanna Seibt’s research on integrative social robotics and general process theory. In fact, Garcia theorizes Huyghe’s work in a recent essay, “An Accretion on the Ruins of the Will” (2024), writing that they “blur the boundary between an artefact—the product of an artist’s will—and a multitude of decentered human, animal, vegetable, machinic, and algorithmic effects, amidst a vast network of relations.” Yet if the intentional content—the artwork’s aboutness—requires operationalizing training data comprised of human-authored imagery, is the machine-man boundary truly blurred? Furthermore, if Huyghe is making the human artistic decision to use generative AI, does the artist’s will not ultimately outshine the algorithmic means through which it is expressed? The best works in the show, like Camata, not only produce such fecund lines of inquiry but offer an answer: that generative AI is an artistic tool amongst tools, a historical instance within a broad-spanning genealogy that countenances the paintbrush and petroglyph stone tools within its ranks. Thus, just as the Dadaists and Surrealists of yore trafficked in chance operations that nevertheless revealed human-centric meaning and expressed the artist’s will, so too does Huyghe demonstrate this sustained, perhaps implacable principle.