home
Lost Information,
Projected Transparency
brian bartz

I. An Inversion
In 1997, The Bureau of Inverse Technology, an ongoing art and research collaboration between Natalie Jermijenko
and Kate Rich, launched the BITplane. A counter-surveillance project conceived of at the dawn of Silicon
Valley's rise to global prominence, BITplane was a small radio-controlled airplane with an attached camera,
which the duo flew over Palo Alto and Mountain View to invert the gaze of surveillance technology back onto the
geography of its makers. The campuses of Sun Microsystems, XEROX Parc, Stanford Research Institute, and Lockheed
Martin — all pioneers in the burgeoning military-surveillance-tech-industrial-complex — were put on
view during the stalwart little drone's maiden voyage, its mission to understand the "threat of the
camera to information space."
1
The project demonstrated a peculiar contradiction, as Jerimjenko explains in an interview at the Bard Drone
Center.
2 What motivated the project was that these corporate
campuses were all, "no-camera zones," and forbade UAV operation in their airspace, implying that the
military technologies and company secrets housed within these facilities were easily capturable through the lens
of a camera. And yet, as the BITplane made clear through it's scarcely legible black-and-white images, there
was very little information to be easily gleaned off the surfaces of the architecture below. As Jerimijenko
phrases it, "What could you actually see? What information could you take from the plane? Of course, the
answer is not much, as contemporary drones have so aptly demonstrated: lots of images but not much actual
trustable information."
3
..............
Twenty years earlier in the mid 1970's, CIA analysts reexamining images from the second world war faced a
similar, if inverted, contradiction. As explored in Harun Farocki's Images of the World and the Inscription
of War
4, those analysts discovered that aerial reconnaissance
photos of factory construction taken over Krakow had inadvertently captured evidence of the as yet undiscovered
Auschwitz concentration camp; in this case the aerial images were laden with actionable information, but because
nobody was looking for it at the time, it was as if it was not there. It took thirty years for anybody to notice
what had been captured. As narrated in the film: "The Nazis did not notice that someone had photographed
their crimes, and the Americans did not notice that they had photographed them. The victims also noticed
nothing. Notes as written into a book of God."
5
How might one account for these two cases: action leading to no information, and information leading to no
action? Might these two examples illustrate the changing relationship between an image and it's capacity to
store or convey information in an increasingly machinic era? What were these two shots fired off into the
abyss?
top: annotated video still of the Lockheed Martin campus captured by the BITplane.
bottom: Film still from Farocki's Images of the World and the Inscription of War, depicting the aerial
photograph of Auschwitz finally reviewed and annotated in 1977.
II. Invisible Polygons Still Exist
To complicate things further still, let us turn our attention to an even more contemporary model for such a
relationship: computer generated graphics. These kinds of images entail the meticulous modeling of objects in
three dimensional space. In essence, this modeling process represents the conversion of information into a 3D
form which is potentially legible to human vision. However, that visual legibility is not as much a given in the
virtual space of a computer as it is with a camera inscribing light from the world. As Jacob Gaboury argues in
the Hidden Surface Problem
6, It is necessary to delineate this kind
of computationally generated, three dimensional image from prior visual forms (like photography), because to
even generate such an image, engineers have had to mathematically construct abstractions such as vanishing-point
perspective and the opacity of surfaces. These features are not inherent to the computational image in the same
way that they are to the photographic. As such, the imposition of these familiar visual constructs which make an
image human-readable were not an inevitability so much as a calculated decision.
Gaboury spends particular time engaging the development of so-called "hidden surface" algorithms
throughout the 80s, which proved immensely difficult for the field to produce. Largely concentrated in the
University of Utah, these pioneering engineers, many of whom would go on to work for the same companies
surveilled by the BITplane, struggled immensely with the question of how to algorithmically determine whether
one vertex appeared behind another in a computer generated image. While it was relatively trivial to model three
dimensional objects in virtual space as well as render them through a mathematically constructed perspective
akin to that of a camera, what proved incredibly difficult was determining if, from that constructed
perspective, a given surface of an object would appear behind another, and would thus necessitate obfuscation in
the output image.
7
An illustration of a hidden surface algorithm at work. On top, a wireframe rendering of
intersecting 3D forms is less legible to the human eye, but technically conveys the most information. On the
bottom, after the algorithm and shading are applied, the form is more legible to the human eye, but obscures
much of the original information necessary to produce the forms which populate the image.
Therein lies what, on the surface, seems to be a contradiction in the generation of computer graphics; in order
for a machine to generate an image that is legible to human vision, it necessarily entails the removal of at
least some of the original information which was used to describe its forms in the first place. In the same way
that the BITplane could not see beneath the roofs of those tech campuses which it counter-surveilled, nor into
the computer hard disks which actually held most of their proprietary secrets, so too does a computational image
hide more information than it reveals. The difference being that this newer type of image was artificially
generated from a complete body of information rather than constrained by the physical capacity for light to
penetrate opaque surfaces. To my mind, this amounts to an ideological choice on the part of the engineers, one
in which a computer is endowed with all the information, deciding algorithmically what to disclose. This kind of
selective obfuscation and tight control over the accessibility of information is a pervasive organizing
principle underlying many contemporary technologies.
8
III. Projecting Transparency
Perhaps it was that those companies did not fully understand the new visual paradigm they had already set in
motion when they decided to prohibit the use of cameras on or around their campuses. Perhaps they had not
realized that, through the work of their own engineers, the world had moved on from the model described in
Farocki's film, one in which someone seeking information from an image might well find it if they knew what
they were looking for and were paying sufficient attention. Or perhaps it was a paranoid understanding that new
paradigms usually arise before the death of the old, that a camera could still catch them off guard. Continued
caution regarding cameras was still warranted, as evidenced in Andrew Norman Wilson's film Workers Leaving
the Googleplex
9, in which the artist videotaped a group of
minimum-wage laborers working on a book scanning project of questionable legality quickly shuffling into their
cars after their shifts ended. He was promptly fired and threatened with litigation for recording it.
10
But despite the continuing (albeit shrinking) possibility of camera-based security breaches, these companies
eventually came around, in a way; in the decades since BITplane was launched, the rise of mainstream privacy
debates and the general opacity of Big Tech has forced many such companies to outwardly project an illusion of
absolute transparency, despite their immense repositories of state and corporate secrets. As Key MacFarlane
notes in The Greenhouse Effect
11, tech campuses have increasingly
made a turn towards "green architecture," which, incidentally, entails their construction out of glass
and transparent materials. And according to MacFarlane, this type of transparency has as much to do with
generating trustworthiness as, "making brutal stratifications of class, race and gender appear transparently
natural."
12
Sometimes this outwardly projected transparency looks like Google taking journalists on tours through data
centers, Apple building high-ceilinged stores that you can see right through from the outside, or the trend of
new tech HQs being constructed entirely from ergonomically curved, eco-friendly glass panels, likely produced
using heavily exploited labor and highly extractive supply chains. Regardless of how this pseudo-transparency is
manufactured and projected, all we have left to look at is a 3D video of a Google server rack in all its opaque
glory.
They realized, somewhere along the way, that most of their data did not need to be visually accessible,
that their best kept proprietary secrets are not going to be ones that a person with a camera could walk away
with, but would exist instead as bits and bytes. And as hidden surface algorithms illustrate, once information
is encoded computationally, its translation into a visually legible form tends to be tightly controlled and thus
inflected with the biases and intentions of those who control the bits and bytes. As such, the accessibility and
trustworthiness of information in an image —which was never really a given— is especially not a
given now.
top: Amazon's Sphere campus in Seattle. bottom: Visitor center for Apple's new
ring shaped campus: Apple Park. Transparency has become an underlying architectural principle for new tech
campuses.
IV. Operational Images, Uncanny Images
Near the end of his life, Harun Farocki became preoccupied with what he called "operational images," a
term he coined to describe the increasing volume of images which are created by machines and are only legible to
machines
13. They are the visual detritus of our age: computer
vision algorithms passing scores of abstractly-patterned jpegs between one another to better understand their
inputs, QR codes littering the virtual and physical world, microsecond graphs produced by high frequency stock
trading bots, a body detection algorithm running on a predator drone deciding whether or not to launch a
missile, ultimately based on its interpretation of small pixel-clusters; the list goes on. Humans are not
necessarily part of the semantic chain any longer; an image is often the output of one machine and the input to
another, even when the consequences are all-too-human in terms of their resultant violence.
But on the other end of this spectrum, we find ourselves climbing out of the deepest depths of the uncanny
valley, wherein the computationally generated images we do create for human consumption are nearing a complete
mimesis. We appear to be asymptotically approaching the ability to generate images which are hyper-compatible
with our understanding of reality, to an extent that was perhaps unthinkable to those early engineers at the
University of Utah. Once we cannot distinguish a render from a photograph, the informatic distinctions between
the two become flattened.
A computer generated image from the Ikea Catalog. They are rendered using a
contemporary light-simulation technique called 'ray tracing.' Around 75% of images in the Ikea
Catalog are now made using 3D modeling software, as it is logistically much easier than
photography.
The types of images being produced computationally are thus stratifying further with each day, either cutting
humans out entirely, or supplying us with images almost indistinguishable from those produced by cameras, still
our deeply flawed metric for realism and truth. In this ever-growing gulf, what have we lost? What have we
found? And what will get buried? What is clear is that, at either end of this operational to uncanny spectrum,
our capacity to consume and interpret visual information becomes deeply problematized.
This is, at the end of the day, a story about an asymmetry of information, which, without coincidence, comes
into being concurrently with the rise of data as capital. Such a paradigm is entirely contingent on accumulating
and centralizing as much information as possible and deciding very carefully what its subjects can and cannot
see of its superstructure. It's a large scale compartmentalizing of data into abstract and horizontally
positioned clusters that one can only interface with using proprietary protocols. The design makes it impossible
to get a bird's eye view on it all, too complex for any individual to fully comprehend.
Maybe we are all a little like the BITplane now, roving about the world while processing countless empty images,
attempting with futility to gain that vertical perspective which might lend us clarity. And even in those images
which aren't so empty, which do tell us something meaningful, can we even trust that they are disclosing to
us anything but a partial truth? A carefully curated reality? Once the images are rendered and distributed,
their hidden surfaces are as opaque to us as what lies beneath the roof of the Lockheed Martin headquarters; 3D
rendering is as much a mode for generating images as it is a lossy compression format that necessarily sheds
information. Can we use computer graphics as a way to highlight the fault-lines, exaggerate the contradictions
of our new visual schema? Maybe at least artists can recuperate computer graphics away from the overly-smooth,
frictionless computer renderings increasingly saturating our world, folding such tools back onto themselves to
reveal something about their constitutive parts, their component ideologies.
___________________________________________
Notes:
1. Bureau of Inverse
Technology
2. Interview at the
Drone Bard Center
3. ibid.
4. Harun
Farocki's Images of the World and the Inscription of War
5. ibid.
6. Hidden Surface
Problems by Jacob Gaboury
7. ibid.
8. See: Abstraction in the computer engineering sense. "The essence of
abstraction is preserving information that is relevant in a given context, and forgetting information
that is irrelevant in that context." - John V. Guttag
9. Andrew Norman
Wilson's Workers Leaving the Googleplex
10. Interview
with Andrew Norman Wilson
11. The Greenhouse
Effect by Key MacFarlane
12. ibid.
13. Operational Images
by Trevor Paglen