Trevor Paglen: A Study of Invisible Images

Metro Pictures, New York

‘Our eyes are fleshy things.’ So begins Trevor Paglen’s essay ‘Invisible Images (Your Pictures are Looking at You)’ published in The New Inquiry in December 2016. It’s a banal observation to be sure, but in a time of rapid image transfer and algorithmically curated visual culture, it’s an important reminder: Human vision, Paglen notes, is at least on an anatomical level rooted in the body and therefore bound up with all of the senses of the ‘flesh’—touch, hearing, smell— along with all the pleasures thereof. What to do, then, within a technological complex fed by images that are no longer prepared for these ‘fleshy things’ and are instead ‘made by machines for other machines, with humans rarely in the loop?’ Such is the domain of ‘invisible images,’ a shadowy realm that has occupied the newly appointed MacArthur Fellow in his present research projects. It is a domain, he writes, no longer dependent on human vision, or any form of sensation for that matter, but on the ‘activations and operations’ of artificial intelligence systems. While AI networks may still look to information encoded within image material as training grounds for surveilling and organising bodies, gestures, and social interactions (Paglen often refers to the readily available image banks within Facebook, Instagram, or Police databanks), they have begun, Paglen writes, to do away with the anthropocentric construct of the image altogether. With the advent of ‘machine-tomachine seeing,’ they have started instead to develop ways to exchange information directly— that is, invisibly. And yet, those archaic, ‘fleshy things’ and the bodies attached to them are not completely left out of the loop: ‘Invisible images are actively watching us, poking and prodding, guiding our movements, inflicting pain and inducing pleasure.’ To investigate the pains, pleasures, and anxieties resulting from the inhumanly fast operations, activities, and movements of these ever evolving machinic overseers represented the central goal of Paglen’s recent exhibition, his first to explore this particular realm of invisibility (though, to those familiar with his research into ‘black sites’ and surveillance systems, it was by no means his first foray into the question of invisibility in general). The show included sixteen imagebased works and a video installation. At the front desk, a substantial printout of notes by the artist was on offer, an essential guide to both the theoretical and technical ideas guiding the project.

Trevor Paglen: Machine Readable Hito (2017). Adhesive wall material. 490.2 x 140 cm. Courtesy of the artist and Metro Pictures, New York.
Trevor Paglen: Machine Readable Hito (2017). Adhesive wall material. 490.2 x 140 cm. Courtesy of the artist and Metro Pictures, New York.

The show opened with a large digital print, Human Eyes (Corpus: The Humans) (all works 2017), installed in the reception area of the gallery and visible from the street. Drawn from a series Paglen terms ‘Hallucinations’ (more on these later), the image alarmed for its ‘fleshy’ viscerality, depicting two disembodied, unblinking eyeballs peering out from a pink, digitally-rendered epidermis. Its placement in the gallery was strategic, introducing visitors to the barely comprehensible and deeply uncanny concept structuring the following three galleries: that a new sort of bodiless being has learned to look back at us through algorithmic non-eyes.

In the first gallery, Paglen curated a selection of images that spoke to the prehistory and early days of ‘invisible images.’ Included here was Megalith, a grid of roughly 70,000 examples of handwritten numbers (0-9) gathered by machines in order to recognize human writing gestures. The links to Minimalist and Conceptual practices were clear. But if Sol LeWitt’s wall drawings or Eva Hesse’s diagrammatic drawing acts of the 1960s and 1970s spoke to a latent irrationality underlying systems of order and regulation, Paglen’s print evinced a machine logic in which the irrationality of the handwritten mark—and its attachment to the libidinal drives of the body—no longer has a place in a society of control.

On an adjacent wall, Paglen included It Began as a Military Experiment, a series of ten photographic portraits of employees at the Defense Advanced Research Projects Agency (DARPA). Selected from thousands of volunteers for an early foray into facial recognition software, these ten subjects represented what Paglen calls in the accompanying text the ‘Adams and Eves’ of machine learning: the first subjects fully enrolled into the fallen world of machine omniscience (‘Checklist with Artist’s Notes: A Study of Invisible Images’). Paglen submitted the portraits— curated so as to include different genders, ethnicities, and age groups—to newfangled forms of AI, running the 1990s-era images through face recognition software that overlay diagrammatic patterns over the physiognomic features. Such pseudo-drawings also lined the surface of Four Clouds, installed on a facing wall. These four cloudscapes trafficked in a romanticist aesthetic similar to Paglen’s earlier series of drones navigating weather patterns. Those earlier images resemble Turner-esque compositions in which technology, nature, and atmosphere have merged into some new ecosystem of control. In the new cloud series, Paglen dug deeper into the long history of drawing, asking his AI to find patterns in the clouds, and thereby riffing on age-old artistic practices of cloudwatching.

On the gallery’s end wall, Paglen installed his playful Machine Readable Hito, a grid of hundreds of portraits of fellow artist and writer Hito Steyerl playacting for face recognition AI. Her seemingly infinite expressions were submitted to algorithms that attempted to decode her gender, emotional state, age, and so on, leading to guesses (listed as percentages below each portrait) that could verge on the slapstick. Such casual humour in the face, as it were, of technological control brought an added question to Paglen’s researches: what moments of humour, even pleasure, might arise from slippages between seemingly irrational human gestures and the forms of machine learning trained to decode them? Might such moments of misrecognition generate critical, oppositional laughter from within systems of control? These questions haunted the final room, which included the most provocative series in the show: the ‘Adversarially Evolved Hallucinations,’ or simply, ‘Hallucinations.’

Before entering this last gallery, however, visitors passed through the film installation Behold These Glorious Times! Bathed in a seductive electronic score composed by Holly Herndon, viewers watched AI cycle through grids of images drawn from ‘training libraries’: banks of images that machines use to learn how to recognize objects, faces, and gestures (‘Checklist’). The result frightened in at least two ways: while the film confronted viewers with a seemingly random selection of the infinite movements and behaviors that make up human social interaction, it also revealed the coldness of a machine perspective that transforms, at inhuman speeds, all manoeuvres into instruments of control. The installation, in short, reminded viewers of a central premise of the show— that machines don’t see images but instead, as Paglen discussed at a recent talk at New York University, do things with them (‘Invisible Images: Aesthetics and Politics of Autonomous Vision’). His ‘Hallucinations’ series seemed to offer one, potentially humorous, tactic for a human mode of détourning the lightningquick movements of the machinic gaze.

The process Paglen invented for constructing these images is complex. Essentially, he created a ‘training set’ similar to those in Behold These Glorious Times! but drew his categories from human discursive structures and the ‘irrational things’ populating them (‘Checklist’). These included ‘Omens and Portents,’ ‘Monsters of Capitalism,’ ‘American Predators,’ and the very Freudian ‘Interpretations of Dreams.’ Building image ‘corpuses’ based around these and other categories, Paglen trained an AI, which he called the ‘Discriminator,’ to recognize iterations of these fantastical beings and events. A second, ‘adversarial’ AI (the ‘Generator’) was then tasked with drawing pictures capable of tricking the first AI into recognizing ‘ false’ versions of these categories. Over millions of interactions, the Generator and Discriminator dialogued with one another until the latter assumed, wrongly of course, that it was seeing something it had been trained to recognize. Paglen termed the resulting images ‘ hallucinations’ because they emerged mirage-like from out of the domain of invisibility that structured his AI adversaries.

The work at the entrance of the exhibition, Human Eyes, was generated in this fashion (from a ‘corpus’ called ‘The Humans’), and it emblematized the paradoxical self-reflexivity of Paglen’s endeavour. Here was an almost painterly depiction of the instruments of human vision but one hallucinated by non-visual, nonhuman means. The final gallery included ten further iterations, and the results mixed horror with dark laughter. On one wall, Paglen paired Highway of Death (Corpus: The Aftermath of the First Smart War) with the decidedly Foucauldian A Prison Without Guards (Corpus: Eye Machines). The former presented the viewer with a matte-grey sky hanging over what appeared to be a barren, crudely-rendered desert landscape, its wide expanse streaked with jarring red-pink pigmentation. In its hallucinated neighbor, the AI’s empty prison was all enclosure, a suffocating grey interior loosely conforming to perspectival space. Situated within the long history of the landscape or interior scene, these works were made uncanny by the knowledge that they had been generated by the automated logic of machine actors, who also, Paglen was quick to remind us, surveil and patrol, IRL, the very sceneries they have hallucinated here. Across the way, Paglen curated a number of other images drawn from the various corpuses that guided his AI adversaries. One of the most striking was Comet (Corpus: Omens and Portents), a representation of a cascading white form falling onto a thin sliver of horizon. Paglen’s earlier images of drones and black sites, which also trafficked in painterly traditions of the sublime, were clearly forerunners. But a new form of awe now seemed to have displaced the nature-based sublimity that still haunted those earlier series: What entranced here was not the magnificent scale of the vista unfolding before us but the incomprehensible, alien speeds and computations that went into manufacturing and preparing this image for human eyes.

The most striking, and darkly comic, juxtaposition in the gallery involved Vampire (Corpus: Monsters of Capitalism) and Porn (Corpus: The Humans), both of which were centered on opposite walls so that they faced off across the room. The vampire, Paglen notes in his text, has often served as a metaphor for, among many other things, capitalism run amok, and in this uncanny portrait, AI has given us its best rendition of the immortal monster’s physiognomy. Made up of a lone, seemingly surgically removed eye, a bloodsoaked maw, and grey, feathery skin, the image, which included both a profile view and a frontal one, peered, many-faced, across the gallery. The leering placement of the image, which struck one as both campy and visceral, pointed to the sinister humour of a machine consciousness that had been pushed to ‘hallucinate’ a being that is itself a hallucination of human imagination. There was also something obscene about this ‘Monster of Capitalism’ staring across the gallery at an image of machine-hallucinated ‘porn.’ Here was the fruit of a situation in which the operations of power that these machines normally serve—namely, surveillance and marketisation—have been momentarily suspended, given over instead to a broken, dysfunctional erotic imagination. A kind of dark, critical laughter could be detected here, and yet, if any humour was to be found, it was one accompanied by horror. With the latter image in particular, Paglen’s adversaries brought out the centrality of violence within pornography, giving us nothing but a mutant image of disfigured pink meat, splayed within some red interior. Paglen, in other words, made the vampiric male gaze that structures pornographic imagery encounter its new adversary: a sort of undead, zombie porn fantasized by machine overlords.

The eroticism of machine-to-machine interaction seemed to be at play in Paglen’s ‘Invisible Images’, intensified by the clear links to painters like Francis Bacon, Max Ernst, or Yves Tanguy. But searching for art historical references would be to miss the point: Though they might draw on its myths and icons, Paglen’s are, at their core, not images drawn from the ‘ fleshy’ realm of human visual culture. They are instead active datapoints exchanged between machines, temporarily made over into images for the benefit of human visitors to an exhibition. Paglen’s provocation is existential, then, forcing us to reevaluate both the centrality of the exhibition format as well as the function of the artist within a postvisual culture of information and machine learning.

How, for example, are we to conceptualize Paglen’s artistic subjectivity here, especially when we take into account his cross-disciplinary activities as artist and writer, researcher and technologist? Mediating between the realm of machine learning and the human world of the senses, Paglen seems keen on making visible the invisible machinations of power that organize everyday life at all levels. The urgency of these concerns has intensified for a present in which technology and neoliberal capital have conspired to forge neo-fascist organisations that no longer keep their racist, xenophobic, or sexist ideologies ‘invisible.’ As Paglen noted recently, these latter ideologies have become internalized by ‘machine-to-machine’ interactions, no longer requiring human input at all (‘Invisible Images’). In this logic, neo-fascist forms of populism that have seized the reins of power in the United States and elsewhere might be seen as expressions of this underlying machinic order.

What might this mean for the system of art in which Paglen’s work participates? In the artist’s crosshairs, I think, is not only the regime of human visuality that has structured western culture since at least the Renaissance but, more specifically, the institution of art and the privileged mode of display and curation that buttresses it: the exhibition format. Though the Metro Pictures show amounted to a seemingly traditional exhibition of pictures (and a video installation) staged within a white cube space, it was haunted, I think, by a kind of institutional critique. If the gallery system depends on the staging of discrete exhibition events catering to the visual capabilities of human bodies, it has already begun, Paglen’s show implies, to adapt to new, ‘invisible’ forms of image processing and control. This has and will occur, no doubt, alongside systems of policing and marketisation that have from the beginning made use of ‘invisible images’ to better surveil, control, and capitalise activities of pleasure, sociality, or resistance.

Within the privileged spaces of the art system, the exhibition—as a form that almost categorically depends on the presence of the body in space— might become a key interface in which to test new artistic practices (and new artistic subjects) capable of intervening in the shadow world of invisible images. For help, Paglen might look to the anti-aesthetic exhibition design practices of Marcel Duchamp and other members of the historical avant-gardes (indeed, during his NYU talk Paglen discussed his growing interest in Surrealism, and in René Magritte’s ‘treacherous’ images in particular). For Duchamp and his colleagues, vision was never presumed to be the domain of the human being alone. Those ‘fleshy things’ in our heads and our presumed ownership over them had to be negated and instead revealed to be intimately intertwined with whole media ensembles. The camera, screen, factory, machine and of course, the eye all had to be considered as equal partners in a complex of technological gestures. But this was only the first step. Behind it lay a utopian attempt to produce lived opportunities for spectators to construct resistant forms of subjectivity within systems of control. In conjuring the invisible movements of power, Paglen takes a significant step in this direction, though we will have to wait and see whether this path results in critique and resistance or mere ‘hallucination.’

Trevor Paglen: A Study of Invisible Images was on view at Metro Pictures, New York, 8 September – 21 October 2017. Paglen’s essay, ‘Invisible Images (Your Pictures Are Looking at You)’ is accessible at your-pictures-are-looking-at-you/