New York

Trevor Paglen, A Prison Without Guards (Corpus: Eye-Machines), 2017, dye sublimation print, 32 x 40". From the series “Adversarially Evolved Hallucination,” 2017.

Trevor Paglen, A Prison Without Guards (Corpus: Eye-Machines), 2017, dye sublimation print, 32 x 40". From the series “Adversarially Evolved Hallucination,” 2017.

Trevor Paglen

Metro Pictures

Trevor Paglen, A Prison Without Guards (Corpus: Eye-Machines), 2017, dye sublimation print, 32 x 40". From the series “Adversarially Evolved Hallucination,” 2017.

Back to school: Trevor Paglen produced his important early work on military black sites and extraordinary rendition while pursuing a doctorate in geography at the University of California, Berkeley. Having already received his MFA, he drew on geography’s analytic tools to develop an artistic practice premised on the hunch that, however “secret,” clandestine government programs would always leave material traces—facilities, flight records, post office boxes—that could be located, documented, and made visible to a broader public. This year, Paglen spent several months as an artist-in-residence at Stanford University, collaborating with computer scientists to explore how computers are being trained to “see” images without human intervention.

Machine vision figures prominently in military technology, aspects of which were detailed in videos by the late Harun Farocki. For his Stanford research, Paglen has largely shifted his attention away from the security state to the private sector. It was thus apropos, if not downright on the nose, that he debuted his new work by staging an event in San Francisco, the city most associated with (and responsible for) image recognition’s capitalization. There, the Kronos Quartet performed string music while a phalanx of cameras and projection screens revealed in real time how software interpreted their movements. Intentionally or not, the setup brought to mind the final chapter of Vilém Flusser’s prescient Into the Universe of Technical Images (1985), which invokes chamber music as a metaphor for a world where automation diminishes human agency so incrementally and ebulliently that the loss is never registered. 

For this exhibition, “A Study of Invisible Images,” Paglen demonstrated the workings of machine vision by applying it to two of art’s most traditional genres, landscape and portraiture. Photographs of cloud studies were superimposed with vector lines, marking a software program’s attempt to translate their curls and tufts into actionable coordinates. In Machine Readable Hito (all works 2017), 360 photographs of the artist Hito Steyerl were each individually annotated with automated gauges of her age, gender, and mood. Perhaps the most haunting image in the exhibition was a rendering of Frantz Fanon’s “face print”—a composite of the unique characteristics that facial recognition software would use to identify him from one picture to the next. Whereas portraiture typically seeks to capture the essence of a subject’s personality, here Fanon was defined purely by the edges of his nostrils and the distance between his eyebrows.

Computers learn to recognize specific objects by analyzing massive sets of sample images. The twelve-minute video Behold These Glorious Times! dramatized this process, putting the viewer in the position of a newly conceived artificial intelligence program reading rapid-fire sequences of firearms, license plates, fingerprints, hand gestures—all typologies indicative of machine vision’s instrumentalization by military, police, and commercial interests. For the series “Adversarially Evolved Hallucinations,” 2017, Paglen trained AI programs with alternative image sets, such as American predators (eagles, stealth bombers), omens (eclipses, comets), and metaphors for capitalism (octopuses, vampires). He then connected these programs to AIs capable of generating images. The back-and-forth between them ended when the first AI recognized (or rather, hallucinated) an object in the picture produced by the latter.

Paglen’s image-set selections for “Adversarially Evolved Hallucinations” paralleled the antic taxonomies of the Borgesian encyclopedia that Michel Foucault recounts at the start of The Order of Things (1966)—categories organized according to a shattering logic that he hails as “the stark impossibility of thinking that.”“A Study of Invisible Images” makes clear that automated vision might now be constructing a new order of things around us. Quite literally, its algorithms could one day determine what objects, faces, and bodies are legible and significant. With the “Hallucinations” series, Paglen takes a step past his usual strategy of mobilizing art to reveal hidden power structures. Instead, he draws on art’s capacity for purposeful purposelessness to imagine otherwise. It’s a venerable cliché that art teaches “us” to see the world differently. Might it also teach our machines?

Colby Chamberlain