Image Machine

Chapter Summary

The present chapter explores the aesthetics of machines with regard to such systems of technical vision. It starts off with a presentation of the concept of “operational images,” which describes a type of image that is not made for human perception but forms part of technical operations. From here we venture into a more general discussion of the technicality and mediality of images. We look at mechanical, electronic, and digital systems that have been used to produce images in the context of art since the 1950s, and pay special attention to a number of art projects that seek to test human perception and the mediality of human vision. In an attempt to apply the dimensions of machine aesthetics outlined previously—its associative, symbolic, formalist, kinetic, automatic, interactive, and autonomous aspects—we explore how visuality is negotiated, and how the machine—as a relation to technology—is constituted, by such artistic vision machines.

For a closer analysis of technical vision systems, it is important to distinguish between the three levels of optics, vision, and images. By “optics” we understand the physical system through which light is captured and prepared for technical or physiological processing. In an electronic system, optically derived input will be translated into electronic impulses which can in turn be coded as analog or digital signals. “Vision” is the process of interpreting the optically derived impressions or data. In such an understanding of vision, the term refers to both the physiologically based process in living beings and the technically based process in electronic and digital apparatuses which process visual information. An “image” is a particular form in which such optically derived visual information appears.

The attempt to find an authoritative definition of the image is a daunting task. Both the art historical and the philosophical literature on the topic are prolific and have recently been expanded even further, given the attention that the images of popular culture have received in cultural and visual studies since the 1980s and the recognition of scientific and technical imagery by image studies since the 1990s. This expansion is reflected in W. J. T. Mitchell’s insistence on distinguishing analytically between images and pictures, the latter referring to the material objects, the picturing surface, as against the more conceptual idea of images, which can take graphic, optical, perceptual, mental, or linguistic forms.⁵ The German art historian Hans Belting has proposed the concept of an “image anthropology” that encompasses interdisciplinary research into the social usages and meanings of images in a global and anthropological perspective.⁶ For Belting, images have to be understood in relation to the media in which they are constructed and brought to appearance, and with regard to the bodies which perceive and make sense of them. A broadening of the field of art history toward a general discipline of historical image studies (historische Bildwissenschaft) has also been advocated by the art historian Horst Bredekamp and others in their research into “technical images,” which include visual representations from scientific, medical, and engineering contexts, looking at the apparatuses, media, and interfaces through which such images are made, presented, and used. Such a broad approach requires an extended definition of images that takes their technical production into account, as well as the instrumental, scientific, and diagnostic procedures in which they are deployed and analyzed.⁷ As the American philosopher Mark Hansen has put it, “The image can no longer be restricted to the level of surface appearance, but must be extended to encompass the entire process by which information is made perceivable.”⁸

Such an extended definition of images would not include all forms of optically derived data sets, like for instance the optical sensors in Nicolas Schöffer’s CYSP 1, Gordon Pask’s Colloquy of Mobiles, or David Rokeby’s VNS, which provide the input for interactive behavior of technical systems, since here the input data are processed without reference to a coherent visual unit. In order to describe the “processes by which information is made perceivable,” it seems useful to draw on the definition of images that the art theoretician Gottfried Boehm has proposed mainly for the traditional objects of art history, but which can be adapted also to an expanded understanding of the field of images.

Boehm has introduced the notion of “iconic difference” (ikonische Differenz) that points to both the semantic and the visual difference of the image from its surroundings:

We speak of an image only when the visual contrast organizes the field of its appearance in such a manner that it can manifest itself as a manageable extent sui generis, i.e., when the visual contrast connects with the size and the borders. It requires the artful division of the gaze which—simul et singulariter—refers the configuration to the field, and accommodating both the individual and the simultaneity of the image.⁹

This interlacing of visuality and materiality is studied through the material conditions of images, how images are made, what images are made of, and how the changing technical conditions of making and using images have changed the very notion of what an “image” is. In his discussion of the relation in principle between image and medium, Boehm has suggested, as a minimal definition of image, “the simultaneous perceivability of the representation and the represented, of the medial premise and the iconic shaping.”¹⁰

Such an analysis that conceives an image as delimited and mediated can also be applied to technically produced images. The aesthetics of such images hinges on a variety of aspects, one of which is the apparent automatism which brings about and transforms the images. Each media-technical dispositive offers a variety of points where such automatisms can occur.¹¹

With such a definition in mind, the exploration of the aesthetics of “machine images” seems a paradoxical endeavor, given that it seems to make little sense to ascribe to a technical apparatus the ability to perceive and understand an image as such. Rather, we have to assume that such an apparatus will use the optically derived data without reference to either the delimitation of the field or the mediality of the image. Both are aesthetic aspects which may pertain to the identification and decoding of signals, but not to a holistic visual impression.

Instead, I would propose approaching “images made by and for machines” as objects of human perception and interpretation. The notion of the “image,” like that of the “machine,” is most usefully understood as a companion concept to the human. In the same way as the machine is a designation employed by a subject in order to address an aspect of the apparatus that constructs both subject and machine, in the same way an image is predicated on a human subject for whom the “image” is a particular form in which the world reveals itself.¹²

I’d argue for the “image” as an anthropological category whose understanding has to take into account machine vision and algorithmic operations on optically derived data as “competition” in the visual field. Such a focus on the image as an anthropological category is not only seconded by Hansen’s emphasis on the perceivability of images, but also by French theorist Paul Virilio’s analysis of the “vision machine” with special regard to the effect that “visionics,” the technical systems of automated vision, are having on human perception: “After synthetic images, products of info-graphic software, after the digital image processing of computer-aided design, we are on the verge of synthetic vision, the automation of perception. What will be the effects, the theoretical and practical consequences for our own ‘vision of the world,’ of Paul Klee’s intuition’s becoming reality?”¹³

Just as the machine is constituted as a technological Other, the “sightless vision” of visionics constructs objects as an actively gazing Other of human perception. “Unless you are Lewis Carroll, it is hard to imagine the viewpoint of a doorknob or a button on a cardigan. Unless you are Paul Klee [who had noted, ‘Now objects perceive me’], it is not easy to imagine artificial contemplation, the wide-awake dream of a population of objects all busy staring at you.”¹⁴ It becomes possible to imagine such an artificial contemplation when, two decades after Virilio made this remark, one experiences Seiko Mikami’s installation Desire of Codes which confronts us with a staring population of robotic cameras feeding a system that processes the images they take of us—and that falls into its dream state when it seems to be alone.

Operational Images

Around 2003, the German filmmaker Harun Farocki coined the term “operational images” for “images that are not simply meant to reproduce something, but instead are part of an operation.”¹⁵ For several years, Farocki had been doing research about automated vision systems for a trilogy of two-channel video installations, titled Eye/Machine. The third part of the series, Eye/Machine III (Auge/Maschine III) (2003), and the related film version War at a Distance (Erkennen und Verfolgen) (2003), deal specifically with different systems used in industrial and military contexts, in which images are not used for humans to survey and control certain operational procedures, but the technical system itself analyzes the images (figures 4.3, 4.4). The historical and contemporary examples that Farocki shows include, for instance, directing a construction robot in a car factory to hit the head of a screw, or a computer vision system that enables a cruise missile to hit its target.¹⁶

Figure 4.3 Harun Farocki, *Eye/Machine I (Auge/Maschine I)* (2000), video still. Copyright © Harun Farocki GbR.

Figure 4.4 Harun Farocki, *Eye/Machine III (Auge/Maschine III)* (2003), video still. Copyright © Harun Farocki GbR.

In an essay that Farocki wrote after completing the film, he cites a passage from the French philosopher Roland Barthes’s Mythologies where Barthes speaks not about images but about words that are part of an operation, for instance when a woodcutter names the tree that will be cut down.¹⁷ Farocki was skeptical about the actual technical relevance of such vision systems and affirms that he found no evidence in his research that such systems had actually been used in either of the US Gulf Wars in 1991 and 2003, and that he came across only one technical system that actually processes visual data for identifying a target, despite billions of dollars’ worth of research and development spending over decades.¹⁸ Farocki’s focus is therefore more on the ideological function of the pictures of such weapon systems shown to humans, which create the myth of “precision-guided weapons” and instill fear, serving to give the weapons industry the appearance of a failsafe and economical system—whose development requires further funding and political support. Paul Virilio had claimed that the images of vision-guided missiles that were shown during the First Gulf War were targeted at the television viewers who saw them, a claim that in Farocki’s interpretation means that the visual regime of operational images is primarily conceived as advertising for the weapons industry.¹⁹ Farocki insists that in the material that he presents in Eye/Machine III, it is therefore important to distinguish between “the aesthetics of commodities” on one side and “machine aesthetics” on the other.²⁰ The latter is characterized, first and foremost, by the disavowal of what is not of importance. In order for it to work, the vision system has to reduce the complexity of the current image and compare certain stored patterns of a territory or an object to such a reductive version of the image. Farocki writes that it is part of this “machine aesthetics” that this disavowal of detail and complexity “creates a strong counter-reaction” in human observers.²¹

Paul Virilio may have been the first to engage with the epistemological, strategic, and aesthetic aspects of such technical vision systems. In a text first published in 1988, The Vision Machine, he discusses the mental impact of current and potential future developments in the sightless vision of “visionics.”²² Virilio is particularly concerned with the epistemological status of mental images in comparison with virtual images, those “synthetic images created by the machine for the machine,” which are enigmatic and exclude the human observer. In this context, Virilio uses the term “operational,” though not in the exact sense that Farocki will later define. Virilio points to the transgression of synthetically produced images into the field of the factual, and to the way in which the virtual begins to have real effects: “To my mind, this is one of the most crucial aspects of the development of the new technologies of digital imagery and of the synthetic vision offered by electron optics: the relative fusion/confusion of the factual (or operational, if you prefer) and the virtual; the ascendancy of the ‘reality effect’ over a reality principle already largely contested elsewhere, particularly in physics.”²³

In Virilio’s analysis, this becoming factual of the virtual is exacerbated by the temporal aspect of these new technologies. Like human perception, the vision machine operates in time, but it does so at a speed that decouples the machine reality from human reality: “These synthetic-perception machines will be capable of replacing us in certain domains, in certain ultra high-speed operations for which our own visual capacities are inadequate, not because of our ocular system’s limited depth of focus, as was the case with the telescope and the microscope, but because of the limited depth of time of our physiological ‘take’.”²⁴ Forecasting future events, which had been one of the major challenges of early cybernetics, becomes a crucial task of the new vision systems, as well as increasing attempts at deception and the overriding of reality by virtuality, not least in the military and propaganda scenarios of the Cold War, where nuclear deterrence was based on the visibility and invisibility of one’s own nuclear military potential. This also concerns decoy technologies, which Virilio describes in detail because they exemplify the operative replacement of real phenomena by deceptive images which, for instance, divert a missile from its real target onto a virtual decoy target.²⁵ An important reason such deceptions are possible is the mediality of perception which hinges on the specific perceptive parameters of a particular system, and the necessity of what Farocki called the disavowal, i.e., the reduction of what enters the field of vision to aspects that can be recognized and processed for operative decision making.

Virilio emphasized the fact that technical visuality is fundamentally different from the way humans see:

Don’t forget, though, that “image” is just an empty word here since the machine’s interpretation has nothing to do with normal vision (to put it mildly!). For the computer, the optically active electron image is merely a series of coded impulses whose configuration we cannot begin to imagine since, in this “automation of perception,” image feedback is no longer assured. That being, of course, the whole idea.²⁶

From a human perspective, this is a blind gaze, a gaze that does not take in the complexity of a visual field and its aesthetic multidimensionality but only looks for precoded cues. “Blindness is thus very much at the heart of the coming ‘vision machine’. The production of sightless vision is itself merely the reproduction of an intense blindness that will become the latest and last form of industrialization: the industrialization of the non-gaze.”²⁷ However, this blind, technical gaze and its statistical way of analyzing and interpreting images is not simply different from a human way of seeing; Virilio saw the danger of it increasingly becoming the basis for human visual perception and understanding: “The usual criticism of statistical thought, as generating rational illusions, thus necessarily comes down to what we might here call the visual thought of the computer, digital optics now being scarcely more than a statistical optics capable of generating a series of visual illusions, ‘rational illusions’, which affect our understanding as well as reasoning.”²⁸

Virilio’s thinking is speculative in the best sense: it departs from facts in order to extrapolate and imagine what the cultural and mental effects of certain technical developments might be. This development goes back to the coupling of photo and film cameras to aerial vehicles and weapons, a history which Virilio addressed in War and Cinema,²⁹ and which was also taken up by Farocki in his films. Both their accounts reference a more technical treatment of the development of computer vision only indirectly, focusing instead on representations and understandings of such machine vision systems.

Initially, these systems were developed in a military context to support the interpretation of aerial photographs taken of enemy territoy.³⁰ Post-Second World War research into cognitive psychology and artificial intelligence was used to devise systems for pattern recognition, and, from the mid-1950s onward, for image processing. Pattern recognition is based on the conceptualization of, for instance, spoken language, texts, or images as information, and their translation into stacks of data that can be analyzed to detect specific patterns that point to the existence of certain objects or structures. Both the encoding of images and the analysis of the resulting coded data are thus crucial tasks of computer vision systems. As a complement to pattern recognition, the field of image processing was developed to enhance or prepare images for their interpretation by humans or computers.³¹

In the United States, institutional development, for instance with the founding of the National Photographic Interpretation Center (NPIC) in 1961, was paralleled by conceptual research in which, by 1965, the mathematician Lawrence G. Roberts made important proposals for formalizing optical information from perspectival views of three-dimensional objects for computer analysis, which would also play an important role for 3D computer graphics.³²

Research in other areas, like radar, ultrasound, and magnetic resonance, explored possibilities of perceiving nonvisual qualities of the object under observation and thus accounted for yet another dimension of the blindness of the “vision machine”: here the model sense was not vision but touch, a remote touch that does not have the limitations of visual perspective, though it has of course its own medial particularities.³³ The decisive criterion for the success of such perceptual systems was, and probably is, whether the derived information can be made operational. Media historian Lev Manovich summarizes the conflation of comprehension and agency that determines such a strategic and technological conception of computer vision systems: “In the best tradition of … pragmatism …, cognition is equated with action. The computer can be said to ‘understand’ a scene if it can act on it—move objects, assemble details, destroy targets.”³⁴

Several authors have more recently attempted to describe the operationality of images less in such pragmatic and technological terms, with their focus on “images by and for machines,” than in terms of the efficacy and agency of images in social and scientific contexts. Art historian Horst Bredekamp has introduced the concept of the “picture act” (Bildakt) to highlight the way pictures provoke or engender actions. The picture act is related to the notion of the effect of an image (Bildwirkung), but it puts the picture itself in the position of the agent that affords an effect or response.³⁵

Bredekamp’s conception of “picture acts” is based on the work of the German philosopher Sybille Krämer, who has been developing a theory of the operativity of writing, in writing as well as in computer programs. Krämer distinguishes between “phenomena of operational pictoriality [operative Bildlichkeit]” and “operationally efficacious images [operativ wirksame Bildern].” Such operationally efficacious images include the type of computer images in, for instance, remote control and automated guidance systems, which Farocki and others call “operational images.” They are technically integrated into a given technical apparatus. In contrast, the images of operational pictoriality are, for instance, diagrams, graphs, and maps. They generally combine pictorial aspects with those of writing and can be put to use without themselves being operational.³⁶

Bredekamp’s picture acts are performed by images in the sense of operational pictoriality, i.e., by the potential of pictures to be used and acted upon. Instead of such a sidelining of operational images, the media historian Thomas Keenan has argued for a politicization of image operations, whether they are done by machines or by humans.³⁷ Keenan introduces Allan Sekula’s account of aerial photographs in the First World War, taken and interpreted by a unit led by modernist photographer Edward Steichen, as “instrumental images,” and the ensuing attempts by aerial reconnaissance units to automate their reading and interpretation.³⁸ Keenan discusses several examples of operational images and refers to their definition by Farocki as images that function without being seen by human eyes: “They need no interpretation in order to work; or, rather, they include something like interpretation as part of the image-making process. They are about as purely instrumental as images can get, and to that they add a feedback loop—what happens in the image guides, produces, creates effects in the world that is imaged.”³⁹

Keenan thus points to the fact that such operational images are not passively read, but are coded and thus active elements of an assemblage of image production, interpretation, and agency; that they are in fact part of the operational apparatus:

Reading becomes mechanical coding. The ideal goal of such an interpretative machine would be to incorporate the “reading” of the image into the very technology that generates it in the first place, to produce images that arrive before the eye bearing their own translation into the terms required for intervention … and then to link that directly to the means of intervention. … The ideal mode of reading them would be not to have to read them at all.⁴⁰

While this description may echo the pragmatism of the developers of computer vision systems quoted by Manovich, Keenan adds that such systems are not neutral, but that their formation and the use of images in them—“the interpretive grid created by Steichen’s unit, or a contemporary analytic algorithm”—are always framed by social and political contexts, which makes their meaning, in Sekula’s expression, a matter of “political maneuvering.”⁴¹ Even the purely technological interpretation of visually acquired data still implies such an act of “maneuvering,” that it is still politically informed. Keenan does not understand operational images in a strictly technical sense, but applies the term to different types of efficacious images and to different contexts in which they can be put to work. Besides describing technological and nonhuman systems, the concept of operational images thus becomes available to political contexts where, for instance, human rights abuses are to be uncovered by practices of “counter-forensics.”⁴²

The German art historian and image theoretician Tom Holert has similarly suggested a more inclusive definition of operational images which, he says, have a function of determining and programming behavior: operational images are “images which have become components of a functional, technical environment and which determine a more or less automatic activity or behavior.”⁴³ Like Keenan, Holert considers images and their potency (Wirkmächtigkeit) in a context of public communication and political struggle. With regard to image censorship and propaganda, for instance, it becomes clear “that the issue is not images but asymmetrical visibilities, and that the availability of visual data always also implies the possibility of its absence.”⁴⁴

However, Holert also argues for an expanded conception of images or pictorialities (Bildlichkeiten) that takes into account the consequences of their algorithmic state as operational images and that may be more suitably expressed mathematically than visually.⁴⁵ A current sense of visuality implies the particular aesthetics of this algorithmic dimension, and considers such images “as complex interfaces of human-machine interaction, … in order to stop expecting them to tell the truth about reality.”⁴⁶

In a similar vein, Thomas Keenan has proposed to extend the notion of the image to this kind of optical processing.⁴⁷ With regard to the automatic, optically controlled docking procedure at the International Space Station, the information that is being processed by the system could then also be called an image, which in this case is a set of optically derived digital data that are processed according to certain algorithms and patterns, a technical matching of spatial positions and movements according to optical clues.

A strong notion of operational images will no doubt have to integrate the algorithmic functionality of the digital systems that process them and in which operational images are not merely reference data but are themselves part of the coding and decoding procedures. However, such a technical understanding of operational images does not mean that the category of the image is actually useful for describing the technical functionality of such vision systems. Rather, I would argue that the notion of the “image” is best employed in analogy to that of the “machine,” which, in the conception developed here, is understood as a metaphor through which the unaddressable apparatuses of subjectification are made addressable. Accordingly, a description of optically derived, digitally and electronically coded and processed data sets as “images” indicates that these data sets are considered, and their efficacies contemplated, by human subjects. The fact that they are also, and differently, seen and acted upon by machines is an aspect of the estrangement and uncanniness that underpins the aesthetics of the machinic.

Machines that Make Images

In most instances when, in the visual arts of the twentieth century, machines were given the task of making images—by means of mechanical, electronic, or digital devices—there was less doubt about the human observer as the proper recipient of such works. Only in the rarest instances were such images operational. A frequent point of discussion, however, was the aesthetic status of the technical devices and the media through which works were created or displayed. In a number of cases, the machines making the images were seen as at least as important as their visual products, and such apparatus-based artworks were criticized for being ambiguous about their potential artistic creativity.

Among the most explicit mechanical image machines are the Metamatics drawing sculptures by Jean Tinguely. From the mid-1950s until the 1970s, Tinguely built a total of 17 sculptures, or apparatuses, with which drawings can be produced. Metamatic No. 8 (1958), for example, is a “tabletop model” of 50 centimeters height and 110 centimeters width (figure 2.5). It is a technical contraption on a tripod, with several flat sheet metal elements (circles, a rounded triangle, a spiral) that are painted black and attached to a support construction. An electrical motor is built into the pedestal on which the sculpture is placed, which is connected to the sculpture by a transmission belt. It drives a conveyor system that moves the different elements. In the right half of the sculpture is a vertically placed black metal sheet of 30 by 20 centimeters to whose front a piece of A4 paper can be attached. A metal arm in front of this sheet can be equipped with a color pen. The arm is attached to one of the moving elements, and as the motor is switched on by means of a button in front of the sculpture, the entire construction begins to move and shake, the pen erratically hitting the sheet of paper within a radius of about 15 centimeters. By applying pens with different colors in successive sessions, the sculpture can be made to produce polychromatic drawings.

In the series of drawing machines, Tinguely varied the construction in size and form; some sculptures use human physical power, like the bicycle-driven Metamatic No. 6 (1959),⁴⁸ but the general principle of a kinetic metal construction to which a pen and a piece of paper can be attached remains the same. The sculptures are never anthropomorphic.⁴⁹ The construction that holds and guides the pen determines the possible structuration of the drawing; the physical latitude of the construction and its actualization determine the graphic form. The act of drawing is an event that happens in time, and the duration of the event—timed by the duration of operating the mechanism—determines the density of the drawn color lines and strokes. Each drawing is unique.

What does it mean to “make an image”? We are faced with a complex assemblage that encompasses the artist-engineer who constructed the machine; the machine with its particular physical structure that determines its operation and that includes both functional elements and “nonfunctional” elements which form part of its overall sculptural appearance and performance; the user or operator who sets the process of drawing in motion; and the resulting drawing. Tinguely provides a deconstruction of the act of image making—even though he reserved a special place for himself. He was aware of the currency of the artist’s name on the art market, so the drawings were treated as artworks, signed and dated.⁵⁰ In the context of the 1959 art fair, Tinguely presented Metamatic No. 17, and at the same time he filed a patent for the principle of the drawing machine. He positions himself as artist, but also as inventor and engineer who leaves the operation of the machine artwork to exhibition visitors.

In contrast, the use of photography and film by artists in the twentieth century most often meant the deployment of industrially produced devices, save for the instances of experimental creation where the construction of special equipment or the transformation of given devices became part of the artistic practice. This was also the case with artists who worked with electronic equipment from the 1950s onward. While this initially meant the use of oscilloscopes and radar screens, mainly from military contexts, the artistic development in the 1960s and 1970s was closely tied to the use of television and video technologies.⁵¹

Nam June Paik had first used television sets in his exhibition “Exposition of Music—Electronic Television” (1963), manipulating the standard functionality of the electronic devices. Paik explored the aesthetic latitude of the TV set, both as a sculptural object and as an instrument for generating images. The minimalist Zen for TV, for instance, which shows a single white line of pixels on the screen of a television set that is placed on its side, was the result of a technical manipulation of the vertical deflection plates in the cathode ray tube of a television set, which is really not an electronic but a physical, namely magnetic function of the CRT monitor. In 1968–1969, together with the electrical engineer Shuya Abe, Paik constructed the Paik-Abe-Video-Synthesizer, one of the earliest examples of an electronic device that could not only manipulate recorded video images but also create synthetic images.

In the late 1960s, many artists started to explore the aesthetic potentials of electronic signal processing in video, first with analog, later with digital and hybrid analog/digital systems. Among them were Steina and Woody Vasulka, whose work in the 1970s is exemplary of the spirit of experimentation and technical invention which dominated those years.⁵² Between 1975 and 1978, Steina Vasulka developed a series of works under the title Machine Vision which experimentally combined optics with electronics. The physical devices assembled video cameras with flat and spherical, still and motorized mirrors and lenses that would move and turn in front of the camera lens, continuously changing the camera view and elucidating principles like the pan or zoom. The series included the installation Allvision (1975) of two cameras pointing at the two sides of a convex mirror sphere, each providing fisheye images of one half of the exhibition space—the two cameras together thus taking images of the entire space (figure 4.5). The live-image feeds from all installations were displayed on a three-by-three monitor wall. The video images of the Machine Vision installations were also used in a series of tapes made by Steina in the same period. One of them, Orbital Obsession (1977), shows a series of performative experiments with the different installations and the real-time modulating, keying, and switching of the image inputs by means of electronic video equipment.⁵³ These effects—rather limited and clunky in comparison with the image manipulations that would soon become possible through digital processing—are typical of the aesthetics of electronic video as used by artists working in the 1970s and 1980s, which harked back to the formalist and kinetic work of artists like Moholy-Nagy or the experimental photographers at the Bauhaus in the 1920s and 1930s.

Figure 4.5 Steina Vasulka, *Allvision* (1975), installation from the series *Machine Vision*. Photo by Kevin Noble. Courtesy of the artist.

More recently, the French artist Julien Maire has used mechanical and electronic techniques to build installations that combine sculptural and media-technical aspects and thus present a unique blurring of associative, kinetic, and automatic aspects of machine aesthetics. His installation Exploding Camera (2007) forms an interesting case when considering the aesthetics of electronic media (figures 4.6, 4.7; plates 4, 5). It also raises the question of the artistic status of the technically created image, which we will later address in a more theoretical perspective.

Figure 4.6 Julien Maire, *Exploding Camera* (2007), installation view, Haus der Kulturen der Welt, Berlin. Copyright © the artist.

Figure 4.7 Julien Maire, *Exploding Camera* (2007), sequence of screenshots. Copyright © the artist.

There is a jumble of electronic gear with some LED lights on a table, and a TV set with a CRT tube next to it. The TV monitor shows blurry still images. Sometimes there are slight movements in those images, but these could also be mere shifts in the color spectrum, technical effects rather than representations of visually recorded physical movements. Different images are presented, one after the other. The images change by being shifted downward; one image is moved out of view at the bottom as a new image appears at the top of the screen. Occasionally, several such images are moved past the screen, then the movement stops and one image is held on the screen for a while. It lights up, tinted in changing colors; objects appear in single or double form. Sometimes a flash of light burns a point-shaped hole in the surface of the screen image and slowly disappears again, an electronic afterimage receding.

On the tabletop, we see a mess of thin cables connecting multiple LED lights which are fixed on small metal stands, around 50 centimeters above the tabletop. The LEDs are of different sizes and colors, go on and off at irregular intervals, and point downward at what looks like the dissected body of a video camera. The camera is wired up, but its case is opened and the components scattered around. Somewhere in the middle there is a CD-size disk, occasionally spinning. Small, frameless slides are fixed to the rim of the transparent disk. Underneath the disk, a button-size device can be recognized as the video sensor of the camera. When we realize that the images on the TV screen change synchronously with the physical movement of the disk, and that the color and light effects on the screen images take place synchronously with the different LEDs lighting up, we understand that the video sensor must still be functioning, transmitting the optical information that it receives as electronic signals to the cathode ray tube of the TV set. Without optical lenses, the optical input to the video sensor is blurry; only rough shapes can be recognized, what they show can only be guessed. The images on the screen are still images which are animated by the changing configurations of light effects from the LEDs.⁵⁴

There are sounds playing which are synchronized with the movement of the images, a mix of almost musical electronic mood sounds and occasional sounds of motors, shots, and explosions. These evocative sounds seem to bind the abstract images and color and light effects together into a filmic narrative: we see helicopters, we see an aerial view of large oil tanks, we see the sun rise in the desert, and together with the sounds, a movie begins to unfold in our head, about political tension, about military activities in the Middle East, about looming death.

A movie has been choreographed for us from some technical debris, a miniature media spectacle created with a sculptural installation of a “dying” yet still functioning media apparatus that is able to produce in real time an experimental film. The video images are produced live by direct illumination of the camera’s active light sensor, filtered by found slide images.

With the title of the installation, Exploding Camera, Julien Maire refers to a historical event that is iconic for the contemporary role of public media. In the year 2001, two days before September 11, Commander Massoud, then the most powerful opponent to the Taliban regime in the Afghan civil war, was murdered. Two al-Qaida suicide bombers posing as journalists killed the northern Afghan warlord with an explosive device that had been hidden in a video camera which they brought to what was supposed to be a television interview. The historical connection between Massoud’s assassination and the attacks on the World Trade Center in New York on 9/11 is documented, though the event of 9/9 has been widely forgotten—certainly because of the magnitude of the events that overshadowed it a few days later, and maybe also because this figurehead of an independent Afghanistan did not fit into the narrative that led to the following military campaign in the central Asian country by the US-led alliance.

Julien Maire’s Exploding Camera is a media sculpture that takes this historical example to reflect on violence and image construction in the electronic mass media.⁵⁵ The image production takes place in the physical gallery space, while the transmission and display of the electronic image happen in the virtual electronic space of the technical infrastructure, including the cathode ray tube of the TV monitor. The installation deploys film slides, LED lights, a digitally controlled electronic video sensor, and an electronic video monitor to portray the fatal dialectics of construction and destruction in the politics of media representation. It confronts us with the speculative perceptions of a machine eye that lingers in a state of near death. The images that unfold on the screen are an electronic interpretation of light signals caught by the video sensor, unprotected by the focusing, distancing, objectifying lens. The apparatus, which is there both as the main site and medium of image production and as a spectacle, allows us to look critically at the construction and the technical constructedness of mediated experiences. Exploding Camera deconstructs the electronic media image and turns the exhibition space into an experimental stage reminiscent of a battlefield.

The historical importance of artworks like Zen for TV, Allvision, or Exploding Camera lies, first, in the exploration of an iconicity that is based on the given technical structure and specifications of the electronic video system: an exploration that was pushed forward both by the invention of new technical tools that broadened the range of visual expression and by the appropriation and transformation of existing technology. Between Maire, the Vasulkas, and the early Paik, we see that the electronic image can be tweaked and created on three levels, the levels of input, signal processing, and output. Every electronic image is constructed from these technical components.

Second, these transitional electronic video images were important because they made the technological construction of television images visible—epitomized perhaps by the early TV-Décollages of Wolf Vostell (since 1958) and the excessive multimonitor video installations of Nam June Paik. They formed the basis for a critique of the image which, in Chris Marker’s seminal film Sans Soleil (1983), is represented by the video images made by Hayao Yamaneko that are transformed by a video synthesizer in order to create a distance from the real objects and that can thus be more truthful, because they appear as what they really are, namely images, and not the transportable and concrete form of a reality that is always already unreachable.⁵⁶

Besides mechanical, photographic, and electronic techniques, algorithmically based digital systems offer yet another form of artistic image making in which part of the creative process is delegated to a technical system. In the early phase of the 1960s, the resulting images were mostly still graphics, due to limited calculation power and limited options for displays and other output devices.⁵⁷

In a discussion of “the images of new media,” Gottfried Boehm in 1999 formulated the assumption that digital images are “explications of calculation processes” which, in principle, “can take on any shape” and “resemble the programs on which they are based.”⁵⁸ He claimed that “the new media are literally ‘media’, i.e. the conditions of possible images,” and that it was questionable “whether and how the new media are able to achieve genuine iconic manifestations.”⁵⁹

Against such a restrictive approach, I want to argue in this chapter that in fact iconicity is always based on technical conditions, conditions of possible images, and that it is a historically arbitrary decision to “naturalize” images resulting from some technical conditions—like the material conditions of painting—while bracketing others. By contrast, Boehm’s phenomenological definition of the image is independent of specific techniques. To be able to describe an image and to understand its aesthetics, it is necessary to be able to analyze both its phenomenological and its technical aspects. According to Boehm, “in the sense of iconic valences, [the images of new media] are no images at all, but simulations.”⁶⁰ Yet this can also be said of painting, for instance, whose materialities of oil paint and linen canvas or wood “simulate” an icon rather than being one. We will here briefly review some of the technical conditions of twentieth-century image making in order to show how Boehm’s question of what an image is, and what it can be, has motivated not only painters and image makers like Kazimir Malevich, Piet Mondrian, Lucio Fontana, or Barnett Newman, but also artists exploring iconicity by means of electronic and digital techniques.

The French artist François Morellet opted for a “programmed” art production for the same reason that motivated many avant-garde artists, namely to reduce the mystique surrounding artistic creativity. Morellet had been trained as an artist and had worked nonfiguratively since the late 1940s. With strongly conceptual, graphic works like 3 double trames 0º, 30º, 60º (Three double grids 0º, 30º, 60º) (1960–1961), in which line grids are overlaid on the same panel at different angles, Morellet developed the principles of conceptual art into a form that, from the 1960s onward, could also be realized by computers. The triviality of the “machines” of graphic rules with which he produced his work was thus a deliberate choice in the tradition of concrete artists like Theo van Doesburg, Ludwig Hirschfeld-Mack, Josef Albers, and Anni Albers.⁶¹ For Morellet, the development of geometric order by means of the successive execution of a predetermined program was not an homage to a technological aesthetics, but a way of renewing the avant-garde concept of a nonexpressive, abstract art. The prospect of an automated production of images by machines implied the conscious subversion of one of the heroic moments of modern subjecthood: the creation of visual art, of an image.⁶²

Another example of the introduction of algorithmic principles into visual art, before and independent of the discrete logic of digital technology, is the formal method for generating abstract images that was developed around 1959 by the Hungarian-born artist Vera Molnar. Molnar had lived in Paris since 1947, working in the tradition of concrete art, painting and drawing images of geometric shapes in different sizes and colors. These pictures often came in series that varied the composition of specific shapes, their angles, or colors, with the aim of finding aesthetically interesting configurations. Retrospectively, she writes in 1991:

In order to realize my research series in a truly systematic manner, I deployed a technique which I called machine imaginaire [imaginary machine]. I imagined that I had a computer. I designed a program and, step by step, I realized simple, limited series which, however, were complete in themselves, leaving out not a single combination of forms. As soon as I could, I replaced the make-believe computer, the imagined machine, by a real computer.

This method helped Molnar “to distance myself from the inherited cultural baggage, to free myself from the influences of civilization that shape us and from the cultural ‘ready-mades’, and to find combinations of forms that had never been seen before.”⁶³

As regards the operation of the machine imaginaire, Molnar would first define the classes of objects to be used in a particular work: squares, rectangles, and shapes with variable angles and sizes, lines and their thickness, etc. Then she defined the rules, the program, according to which the configuration of the objects would be established and iterated. This program was then executed in meticulous drawings, Molnar making precise notes about the steps that she had already completed, in order to avoid omissions or repetitions. She did not always use the same drawing techniques for the execution of the program, but in general she only used black and white, after experiments with colors had proved overly complicated. The colors were added later, again based on a systematic procedure using color charts. In order to avoid specific unwanted cases—like, for instance, diagonals in consecutive rows touching each other, forming continuous lines—Molnar introduced additional rules that would exclude, in this case, certain angles (e.g., 45 and 135 degrees).

Molnar’s procedure was not designed to produce specific and unique graphics; it is a method for visualizing all possible iterations of a predefined program. From these instantiations, and based on her own, subjective sense of aesthetics, she then made a selection which would be realized in a more elaborate visual form, such as drawings or paintings. “First I thought that there would be one correct version, but then I understood that I am more interested in situations than in individual cases, in the same sense as I don’t believe in a single artwork.”⁶⁴ The criteria of the selection are subjective and intuitive: “In general, I have understood that I like things that are either close to total order, or close to total chaos. I still don’t so much like anything between those two states.”⁶⁵

When the pioneers of computer graphics, like the engineers-cum-artists Georg Nees, Frieder Nake, or Michael Noll, were able to use mainframe computers for the creation of the first graphic images in the early 1960s, with the challenges of writing a program in an appropriate programming language that could be executed by a computer and whose result could be printed or drawn by a plotter, the complexity both of the technical apparatus and of the potential visual output expanded rapidly.⁶⁶

In the broader context of our discussion of the machine in art, a document by the German mathematician and early computer artist Frieder Nake is of particular interest, because it shows the prevalence of the conception of the “machine” as a specific, subjective category. Nake’s “Production of Drawings, Sound Sequences, and Texts with Electronic Computing Systems,” published by the Deutsches Rechenzentrum (German Computing Center) in Darmstadt in 1966, provides “remarks on the programming of computer graphics” with the aim of instructing art schools in the potentials of the new technological means.⁶⁷ Illustrated with his own early computer-graphic experiments, Nake’s text spells out in simple terms the different components that are used in the production of such graphics: a program set up to run on the computing system, the computing system which executes the program, and the transfer of the information output from this execution via punched tape to a drawing machine which executes the drawing. The running of this main program requires, writes Nake, a number of subprograms which will, for instance, provide pseudo-random numbers, or direct the punching of the punched tape. This assemblage can be complemented by programs “that determine certain aesthetic measures during the computation process, and accordingly influence the progress of the drawing. This production must be considered in the context of the artistic theory that M. Bense and his collaborators have developed following G. D. Birkhoff.”⁶⁸

This last remark points to the concept of information aesthetics, which grew out of attempts to find a rational, mathematical basis for aesthetic judgments, or, as some of its proponents hoped, an objective description of aesthetic quality. Around 1930, the American mathematician George D. Birkhoff had published texts on the statistical “aesthetic measure” which he derived from a structural analysis and quantification of complexity and order in artworks. This principle was adopted in the late 1950s by Max Bense in Germany and by Abraham Moles in France, both of whom, independently of each other, started using the term “information aesthetics.” Bense continued Birkhoff’s reception-aesthetic hunt for an objectification of aesthetic judgment—nurturing in some early German computer artists the illusion that an application of the quantitative criteria of information and redundancy would lead to artistically interesting results—while Moles proposed a production-aesthetic theory that was instructive for artists working with information systems and that, unlike the Bense school, took into account the unpredictability and openness of material and form.⁶⁹

In Nake’s description, the technical components of computer, computer program, algorithm, and drawing apparatus are not addressed as “machines.” Instead, Nake diligently speaks of a “digital electronic computation system (or installation)” (Rechenanlage), a “program” and its “subprograms” (Unterprogramme), and of the Graphomat Zuse Z 64 as a “fully automatic drawing apparatus” (Zeichengerät). The word “machine” is only used in two instances: once with reference to the drawing apparatus (as “the punch-tape-controlled drawing machine”), and once, more conspicuously, with regard to the computer:

The first and most important task is the setting up of the program. It should enable the production of a whole class of drawings (“aesthetic objects,” according to Bense). It should be able to go through all the variations of a specific schema. This should be seen in analogy to the approach of the artist who pursues a theme in all directions by means of his “intuition.” By intuition we can understand the selection of possibilities from a certain repertory. This intuition is simulated in the computing machine by the automatic selection of pseudo-random numbers.⁷⁰

The slippage—for once calling the “Rechenanlage” a “Rechenmaschine”—occurs precisely when Nake makes a direct comparison between the practice of the human artist and the simulation of this practice by means of the computer. This moment of encounter—the computer eye to eye with the intuiting artist—is what brings forth the technical “as machine.” And correspondingly, in the example mentioned before, the Graphomat Z64 is a drawing “apparatus” so long as it is connected to the other devices in the system, but for a moment it becomes a “machine” when we see it draw and compete with a human hand. By analogy, Molnar’s machine imaginaire is a “machine” because it externalizes and rationalizes a procedure in the course of which the human brain is turned into the computer which processes an algorithm and thus controls the drawing apparatus of the hand with a pencil on paper.

The two concepts of aesthetic practice represented by Molnar and Nake differ in an important way. Nake’s speculative proposal—inspired by the wish for a high score on Birkhoff’s scale of information aesthetics—to implement an additional program “that determines certain aesthetic measures during the computation process, and accordingly influences the progress of the drawing”—such a programmed correction would have been unacceptable for Molnar, who, both in the period of the machine imaginaire and after she started using digital computer systems from 1968 onward, slavishly stuck with the execution of the predetermined program. Molnar insists that she never cheated, for instance by interrupting the plotter during the drawing process when it had reached a stage where the drawing seemed ready. She always let the plotter take its course and complete the program. If she was not content with the result, she might modify the computer program and have it run again, but she would never manipulate the process during execution, as suggested by Nake’s aesthetic correction. Instead, she would at times be surprised by the results of programming mistakes, like the graphic effects of a wrongly placed comma in the written code.⁷¹

This is the moment of the machine, the experience of a technical action that may rationally be understood as fully programmed, yet that is felt to have an uncontrollable logic of its own. It differs from the concrete paintings of Theo van Doesburg whose shapes were developed on a rectangular grid but remained under the control of the artist’s sense of balance and compositions, and it is more singular than the serial images by Ludwig Hirschfeld-Mack who, starting in his years at the Bauhaus, algorithmically drew all the possible variations of certain graphic configurations—much in the methodical manner that Vera Molnar would perfect with the machine imaginaire. Instead of the mystique of apparent machinic volition, Hirschfeld-Mack’s graphs invite comparison and selectivity, but they do not offer the unique moment of creation that the drawing computer would perform in the execution of a specific software program.

In the course of the 1990s, computer software gained increasing critical attention both for its social relevance in structuring communication infrastructures and forms of computer-based agency and for its aesthetic possibilities.⁷² With the growing availability of affordable computers and the emergence of the World Wide Web, artists experimented with the configuration of networks and explored the potential for expanding the limits of how software can be employed in creative processes. Artistic browser software projects, like I/O/D’s Webstalker (1998), Netochka Nezvanova’s Nebula.M81 (1999), and JODI’s Wrong Browser (2000), and data analysis and visualization projects by artists like the Radical Software Group’s Carnivore (since 2001) or Minitasking (2002) by Schoenerwissen, indicate the typical relation between a critical articulation of information infrastructures and the visualization of machine operations.⁷³ The dynamic network environments in which projects like these operate make them akin to an aesthetics of systems rather than one of machines. In the same way as the autopoietic dimension of generative software projects points beyond the principle of the machine and uses this term, if at all, only in the metaphorical, metonymic sense in which authors like Maturana, Guattari, and others have employed it, such network-based projects presuppose an ecological understanding of technology which, as I will discuss more fully in chapter 6, transgresses the twentieth-century concept of the machine.

This concept still informs works that address the question of the iconicity of algorithmic processes, like French computer artist Antoine Schmitt’s Vexation 1 (2000, figure 4.8). This is a computer-based work in which, in the confines of a white-lined rectangle on the black surface of a display screen, a small white circle moves from one side of the rectangle to the other in a fixed sequence—bottom to top, then left, right, and back to the bottom—sounding a different tone whenever one of the sides of the rectangle is hit. While this rule, and thus the sequence of the four tones, is fixed, the path on which the little “ball” travels from one side to the next is calculated in real time and can take on a variety of speeds and trajectories, suggesting phases of protraction and hesitation as well as moments of excitement and speed. The erratic movement of the white ball and the arrhythmic repetition of the sound sequence are indicative of a randomness which is clearly programmed and based on tables of random numbers, yet somehow appears as subjective and intentional in its performative realization. Rationally, the viewer may think this system’s visual output is a kinetic grandchild of Morellet’s tilted grids. Yet, as one observes the white ball in Schmitt’s Vexation 1 hesitating, quivering, and then exploding into cheerful bouncing in the next moment, its strange dynamic makes it difficult not to project traits of subjectivity onto the quite obviously programmed artifact.⁷⁴

Figure 4.8 Antoine Schmitt, *Vexation 1* (2000), software program, sequence of screenshots. Copyright © the artist.

In contrast to such a formalist aesthetic approach, the Dutch-Belgian artist duo JODI have developed a more deconstructive approach to software-based images, in the Dadaist tradition of kineticism. Since the early 1990s, JODI have explored the different layers of encoding and decoding, of representation and of communication, of the interaction between algorithmic and iconic protocols that determine the screen-based interfaces of computer games, Internet services, and various client software.⁷⁵ Their work Untitled Game (1996–2001) is based on four levels of the popular computer game Quake. The game engine software has been manipulated and the visual representations of the different levels have been stripped of all details so that they are now barely recognizable, composed only of schematic op-art-style black-and-white surfaces, or in one case of a totally indistinct whiteness (figure 4.9). However, the virtual spatial structure and the different traps and enemies that inhabit the game are still there, and the abstracted game levels can still be played. JODI thus offer a commentary on the idea of smooth interactivity which is normally associated with such virtual environments. The radical pictorial abstraction gestures not only toward the conventions of pictorial realism, but also toward modernism’s false assumption about the inherent loftiness of emptiness as pure form.

Figure 4.9 JODI, *untitled game* (1999), software, 12 game modifications of *Quake*, screen image. Copyright © the artists.

Rather than ask whether such visual spaces constitute “genuine iconic manifestations” (Boehm), the question for a contemporary notion of images must be how the algorithmic and interactive conditions of such visual spaces are changing concepts of visuality, and of iconicity. In the late 1990s, the US artist John F. Simon created Every Icon (1997), a work that shows a 32-by-32-pixel grid in which each cell can be either black or white. The work pledges to display all possible configurations of the pixels, i.e., “every icon” that can be represented in such a limited, 32-by-32-pixel matrix (figure 4.10). The speed of the calculation is dependent on the clock speed of the computer on which the program runs, but even on a fast computer, going through all possible configurations would take several million years. On one level, Every Icon is a parody of the algorithmic processing done by Vera Molnar and shows how the computer, faced with such a simple and overwhelming task, can itself become a machine imaginaire. And on another level, while the piece is certainly not an “image,” it addresses the mediality of images firmly in the tradition of artists asking this very question throughout modernism.

Figure 4.10 John F. Simon, *Every Icon* (1997). Image courtesy of numeral.com.

Physiological and Technical Embodiments of Images

Even long before the invention of cinematography at the end of the nineteenth century, artists have been aware of the medialities and the technical conditions of images they can create. The different levels of technicality that we analyzed in the previous section are stacked onto one another: digital techniques require electronic circuits to be processed, and electronic techniques have to be coupled to physical, at times mechanical interfaces in order to be perceivable. A project of “digital art,” like Schmitt’s Vexation 1 or Simon’s Every Icon, requires the electronic circuit boards on the computer that runs its software, and the material supports, the physical electrodes or liquid crystals, in the displays that show its visual representations.

Another level of mediality required in visual art is the physiological apparatus of human visual perception. Although physiologically grounded visual phenomena like the afterimage have been known since antiquity, it was not till the nineteenth century that they were systematically researched and described. In his seminal study Techniques of the Observer (1990), American art historian Jonathan Crary discussed the parallel developments of this research and of the media techniques that made it possible to study the respective phenomena, and that turned some of these phenomena into entertaining spectacles.⁷⁶ Crucially, the researchers of the nineteenth century discovered the physiologically determined latency and thus the temporality of seeing, and hence the autonomy of the act of seeing from the seen object. The eye and the human perceptual (optical and neurological) apparatus were discovered to be part of a continuum that, by means of media techniques like stereoscopy or stroboscopy and under specifically controlled spatial and lighting conditions, could be made to perceive images as moving or three-dimensional.

Like all images, these perceptions are effects of the respective media apparatus, which necessarily depends on the physiology of visual perception as one of its conditioning frameworks, an insight that has also been reflected in the work of painters since the nineteenth century—the paintings of Paul Cézanne are the classic illustration of this trope.⁷⁷ The “representation of perception” and of the mediality of human perception was epitomized by Marcel Duchamp’s Rotoreliefs of the 1920s and 1930s. Duchamp produced six double-sided disks 20 centimeters in diameter, titled for instance Montgolfière, Gold Fish, or Chinese Lampion, which were to be rotated at a speed of 40–60 rpm by a gramophone-like motorized apparatus. When spun like that, it appeared to the human eye as though parts of the designs were making additional or contrary movements, or as though certain elements were three-dimensional.⁷⁸

The Rotoreliefs form part of Duchamp’s critique of a retinal art which reproduces visual experiences. Instead, he sought to enable a new type of experience that directly speaks to the mind. Yet at the same time the spinning disks can be understood as “hyper-retinal,” in that they allow us to observe at least part of the processes through which the human visual apparatus creates the impression of movement and spatial depth from specific graphic and dynamic configurations.⁷⁹ What appears as an image in the spinning Rotoreliefs is a phenomenon which uniquely and exclusively appears to the human eye, or rather to the human mind mediated by the eye and the neurological processes of visual perception. This phenomenon is latent in the design printed on the disk, and will be realized only under the conditions of the motor spinning the disk at a particular speed, and in the presence of the perceiving eye.⁸⁰

Research into the functioning of human vision led, around 1830, to the discovery of the stroboscopic effect: based on the latency of the afterimage, images that appear to the human eye in discrete form will be perceived as a continuous, moving sequence, if the differences between the individual images are small enough. The phenakistiscope and the zoetrope were the first devices that made use of this effect to simulate movement through the consecutive exposure of discrete still images.

Unlike Duchamp’s spinning disks, whose visual potential lies in specific graphic configurations that can be activated by exposing a continuous movement to the human eye, the zoetrope is based on discrete and discontinuous images which are fused by the latency of human visual perception. For this effect to work, one factor plays a crucial role: the speed at which the images change, or the number of image states per second that are necessary to create the illusion of continuous movement. This gives a first understanding of the notion of “resolution,” i.e., to what degree a movement has to be resolved into individual steps in order to appear continuous when presented in the zoetrope. Sixteen to eighteen images per second are sufficient to create the illusion of movement. The technical apparatus and the image sequence have to be matched and “synchronized” in order to simulate a movement that appears “natural” to the human eye.

In the nineteenth century, zoetropes and similar cinematographic apparatuses were mainly produced as toys and for entertainment. The chronophotographs of Eadweard Muybridge lent themselves to the reproduction of movement in such apparatuses, though these chronophotographs had been produced for the very purpose of studying the movement of humans and animals in discrete steps, rather than “blurring” them again into continuous movements.

The introduction of image telegraphy and of television made it necessary to consider the aspect of resolution not only with regard to the segmentation of movement in time (temporal resolution, image frequency) but also with regard to the density of visual information on the image surface (pictorial resolution). In order to transmit an image by means of electrical or electronic signals, the image has to be split up into a matrix of image points (pixels) that are individually encoded into discrete signals, transmitted, and then decoded for display on a monitor with an equivalent pixel matrix. The higher the number of pixels and the greater the detail of the representation, the higher also the requirement of information that has to be coded and transmitted.

Here again the human perceptual faculties play an important role, since the level of accuracy of the image representation depends on the accuracy of human vision and the degree to which adjoining elements of the image are perceived as separate or seem to merge. Thus, in a fine matrix, different shades of gray can be constructed from different mixtures of white and black pixels, and on a classical RGB color TV monitor, colors and shades are constructed from more or less bright pixels of red, green, and blue.⁸¹

The American artist Jim Campbell has explored the aesthetics of the pixel-based image and the phenomena of image resolution, of blurring and clarity of the image, in his Low Resolution video installations. In the work Church on Fifth Avenue (2001), a matrix of 32 by 24 red LEDs is used as a monitor to display a slow-motion video of a generic street scene of people walking by on a wide pavement and cars driving on a street in the background (figure 4.11, plate 6). A sheet of diffusing Plexiglas is placed in front of the display at an angle, so that on the left side the Plexiglas is very close to the LEDs and barely diffuses their light, whereas at the right edge the Plexi sheet is around 10 centimeters in front of the LEDs, which here appear as highly blurred. While the LEDs at the left edge can be clearly distinguished as individual points of light, to the right there are only more or less bright areas of red light. The video image of the street scene is best discernible in the middle area where the diffused refraction on the Plexi sheet merges the light from the individual LEDs into continuous shapes and figures. Toward the left side, the LED pixels, at more than 2 centimeters distance from each other, do not form a continuous image but appear as points of light that brighten up and dim down, without this sequence of visual impulses becoming legible as iconic information. And toward the right side, there is a continuous yet indistinct, myopic blur.⁸² Due to the movement in the video image of pedestrians and cars from left to right and from right to left, the viewer can follow individual figures passing gradually from one state of iconicity to another, moving from the low-resolution abstraction through a finely diffused legibility into the nebulous blur on the right of the screen, or in the opposite direction.

Figure 4.11 Jim Campbell, *Church on Fifth Avenue* (2001), installation view. Copyright © the artist.

The question of image resolution is a crucial factor in the context of the digital coding and transmission of image data files, where, due to the technical limitations of transmission bandwidth, a balance has to be found between image accuracy, as perceived by the human eye, and data volume. Image compression standards for still and video images seek to optimize this balance on the basis of “perceptual coding” that, as in the JPEG standard, explicitly takes the characteristics of human visual perception into account to mathematically model the optimal level of loss of detail for specific display environments.⁸³

The visual aesthetics of JPEG compression and the questions of photographic indexicality and veracity have been highlighted in the jpegs photo series by the German photographer Thomas Ruff, for which the artist took JPEG-coded photographs from the Internet, enlarged them to a monumental size, and printed them as sharp and fine-grained photographs, so that the tiled structure of the JPEG images becomes blatantly visible.⁸⁴

What unites these examples of artworks, from Duchamp’s Rotoreliefs, through Campbell’s low-resolution video installations, to Ruff’s jpeg photos, is the fact that they not only articulate aspects of technical and human physiological dispositions—which is something that the visual media technologies on which they are founded already do—but they all reapproach and reinvent the medium which they deploy and make its particular aesthetics available for further scrutiny. They are media for human perception. Insofar as these works explore the articulation of human vision and technical media, they also highlight the entanglement of human perception in the technical construction of images, and vice versa.

The cascade of physiological, physical, electronic, and digital conditions of image making described here points to the fact that there is no basic form of iconicity, and that images always depend on highly specific and complex material conditions. The media that convey images never exist in a pure and definitive form. The artworks discussed here illustrate the diversity of critical and creative approaches to those material conditions, and point to the aesthetic “surplus” which any image medium carries. Our particular interest lies in those aspects of the technical conditions of image making where the respective apparatus takes on, or seems to take on, a level of autonomy, appearing not as a mere ancillary if idiosyncratic medium, but as a machinelike agent that determines by itself the production and formation of the images that we see. Thus, the Rotoreliefs by Marcel Duchamp—and the works by Campbell and Ruff do this by analogy—display for the critical viewer both the somewhat magical phenomenon of shapes and movements appearing in the spinning disks which cannot be seen on the still disks, and the equally strange awareness that one’s visual perception is a process directly coupled to the technical medium in front of the eyes, and able to bring about phenomena, optical or rather visual illusions, that only exist under the media-technical conditions of this particular human-machine dispositive.

The media-technical and physiological conditions of image making and perceptions are affordances that imply the possibilities of understanding and acting upon an image.⁸⁵ The inscription of the ways of seeing into an image dispositive—whether it is geared to a human or a machine as recipient—requires a knowledge, or an intuition, about the deployment and decoding of the perceptual offers that an image makes. A particular aspect of the aesthetics of an artwork can be to counteract the expectations inscribed in an image or an interface, and to offer a counterintuitive if not counterfactual impression that requires a readaptation of the intuitions that usually guide the response to such affordances.

Such a challenge to the expectations of the behavior of images is proposed by Julien Maire’s installation Memory Cone (2009), which also levels the boundary between image production and image perception by building the need for interaction by the exhibition visitor into the setup (figures 4.12, 4.13; plates 7, 8).

Figure 4.12 Julien Maire, *Memory Cone* (2009), installation view, MUHKA, Antwerp. Copyright © the artist.

Figure 4.13 Julien Maire, *Memory Cone* (2009), projection detail. Copyright © the artist.

On entering the exhibition space, the visitor encounters a table with a black tablecloth. There is a video camera on a small tripod pointed at the black cloth, and, on a separate table and connected to the camera, a video projector with unusual extensions that projects onto a white screen. White strips of paper can be placed by the visitor on the black cloth; as they are moved into the field of view of the video camera, the projector displays equivalent parts of what look like generic family images. Thus, rectangular or square fragments of an image can be explored by moving around the white paper, or by placing additional paper strips which will enlarge the zone of visibility of the image. Gaps between the light papers, and shadows thrown by one piece of paper on another, show up as empty black, as a void. This void can be restored, and the image removed entirely, by removing all paper strips with a swipe of the hand.

The production of the projected image remains enigmatic, for although the entire apparatus appears in front of our eyes, it is difficult to understand where the image is actually generated. The video camera is only pointed at the tabletop, recording the differences between black and white areas. Part of the extended video projector is a straight-tray slide projector with slides that are consecutively projected. A special set of optical lenses reduces the image from this slide projector and directs it at a mirror device that is part of the digital light-processing (DLP) technology used in contemporary video projectors. This digital micromirror device (DMD) carries hundreds of thousands of small mirrors on its ca. 2-by-3.5-centimeter surface. Each of these small mirrors can be tilted electronically to produce multiple shades of gray, depending on the speed at which it is tilted back and forth. In a video projector using the DLP technology, the DMD together with a special color wheel technology can produce color video images of high-definition quality, using a white light source.

Julien Maire repurposes this technology in the Memory Cone installation: his DLP video projector regularly converts the live image taken by the video camera on the tabletop into the projection of an, in this case, dark gray surface image. When a white strip is moved into the field of view of the camera, the respective micro mirrors on the DMD are tilted at a frequency that normally produces the equivalent of brightness. The color wheel of the projector is bypassed, so the image of the tabletop, and of the hands arranging the white paper strips, is without technical color manipulation. The light source, however, is not a regular LED, but the slide projector whose lens projects the slide image onto the DMD. The light that gets projected onto the screen is the light from the slide projector bulb, filtered by the slide, so that all the colors, the bright and the dark areas of the slide, determine the composition of the image triggered by the white strips. This effect can also be observed when a hand reaches into the field of view of the camera: the visitor’s hand then becomes the light trigger and defines, with its shape and areas of brightness and dark, the shape of the parts of the slide image reflected onto the screen. While the slide image is reversed by the DMD mirrors, the shaping video image is, under normal circumstances, a straight, analogical presentation of the camera image—although, of course, the digital settings in either the camera or the projector could easily be changed to get a reversed version of the camera image.

There is a chain of material, visual, and electronic events, each of which can be addressed as an image or as an image source, and each of which is taken, processed, and passed on to other parts of the technical assemblage. The two main image sources are coupled optically (the slide projector) and electronically (the video camera) to produce one projected image that is an optical reflection of the slide image, with the light configuration being shaped by the electronic transmission of the camera image onto the DMD.

Maire’s title for the installation explicitly relates to French philosopher Henri Bergson’s concept of the “memory cone.” It counterposes the expanse of all our memories, dispersed across the recesses of our mind—represented by the broad circular base of the “cone”—with what we remember at a specific moment, literally pinpointed by the tip of the cone. This relationship has been discussed by film and media historian Edwin Carels, who has also pointed to the technical aspects of duration and memory, terms which are crucial for both Maire and Bergson.⁸⁶ Yet Maire does more than offer a translation of Bergson’s metaphor into an optical process. The technical contraption he has built constructs a setting and a form of relation between images and actions that is unique and allows for an aesthetic experience that cannot be had otherwise. In that sense, Maire’s Memory Cone surpasses Bergson’s conception and offers its own model experience for the way we interact with our own memories, as well as with technically mediated images. This model is made possible by the specific technical construction, and it inherits the conceptual hybridity of the technical, visual, and interactive elements of the installation. In the same way that Bergson took cinematography as a metaphor for the mechanism of human thought, Maire takes the philosophical concept of the memory cone as a metaphor for the mechanisms of image construction.

Memory Cone deals with images that are not static or durable but transient image events, and thus raises the question of what and where the image actually occurs. The installation makes it possible to distinguish the image in its different aspects as hybridized and accumulated states: what we see on the wall is an analog reflection of the light that is projected by the slide projector onto the DMD mirror device; the micromirrors are positioned according to the images taken continuously by the video camera of the tabletop, and electronically processed according to the DLP protocols. The physical interaction of the visitor with the white strips is electronically mediated by the video camera and translated into the superfast physical movements of the micro mirrors on the DMD chip. What becomes visible of the photographic (positive) image projected by the slide projector is technically conditioned by the electronically controlled configuration of the DMD mirrors, and visually conditioned by the placement of the white strips by the visitor.

There is neither analog continuity nor static durability in electronically transmitted images like those of Memory Cone: while the moving image in a traditional film is constructed from a projection sequence of 24 images per second, and the PAL television standard generates 25 images per second on the cathode ray tube, the micro mirrors in the Memory Cone installation wiggle thousands of times per second, originally to create different shades of gray, whereas here they reflect a slide image at a rate at which the human eye perceives as a continuous image what is really a high-frequency flicker. So even though the slide projector produces a continuous analog image, the DMD mirrors “digitize” this analog projection into discrete on/off light events which are, however, too fast to be distinguishable by the latent human visual perception.

The specific aesthetics of the image—both its particular visual appearance and the technical and pragmatic apparatus through which it is generated and formed—hinges on a dispositive that can be described in analogy to the apparatus theory of cinema. In the early 1970s, film theoreticians like Jean-Louis Baudry and Jean-Louis Comolli applied Louis Althusser’s critique of “ideological state apparatuses” to film and cinema, describing the related notions of reality and subjectivity as ideological constructs. At the time, apparatus theory exposed the image-producing apparatus of cinema to a form of formal and political critique that was soon also applied to the critical analysis of photography. This critique turned against representational realism and the supposed transparency of the technology, and against the illusion of a medium that would transmit information without loss or deformation, and revived the intent of the critical avant-garde of the 1920s, of Vertov, Brecht, and others, to reveal the mechanisms of aesthetic representation by inscribing the construction of the technical and social dispositive into the artwork.⁸⁷

Even though Julien Maire’s installations formally follow such a deconstructivist aesthetics, they maintain a certain level of what we might call obscure transparency: despite the technical mechanisms lying bare and open, the aesthetic experience of his works is characterized by a sense of wonder rather than by rational comprehension. Even though we may understand the functioning of the slide-reflecting DMD, the apparition of the image content in the projection, which corresponds to the position and shape of the paper strips on the table, remains miraculous.⁸⁸

Maire’s aesthetic program is, accordingly, not one of enlightenment but of expression and invention. He says that he does not strive for the straightforward production of images, but wants “to rethink all the machinery that is producing the image.”⁸⁹ For him, the technical system of the image production is an artistic medium that he deploys, appropriates, and modifies in order to be able “to show what I want; … the system has to fit with the image that I want to produce.”⁹⁰ The art historical context within which this work is taking place is that of the technically produced image. Through his work, Maire is “looking for a new type of expression that shows what is missing in the classic recorded image,”⁹¹ seeking to extend the expressive range of such technically mediated images, “like a painter.” In this effort, one of his strategic interests is to counter the tendency toward ever greater speed of transmission and ever higher resolution. In contrast, he wants to explore slowness and the aesthetic potentials of low-resolution images.

Maire’s work dissolves the oppositions between static and moving image, between image, medium, and body or production, interaction, and reception, and makes it impossible to consider what an image is without reflecting on the technical and perceptual dispositives that construct it.

The Aesthetics of the Vision Machine

The image machine operates on multiple levels—conceptual, material, technical, corporeal—and requires these different levels to be put to work. The American artist Sol LeWitt famously spoke of the machinic aspect of conceptual art: “When an artist uses a conceptual form of art, it means that all of the planning and decisions are made beforehand and the execution is a perfunctory affair. The idea becomes a machine that makes the art. This kind of art is not theoretical or illustrative of theories; it is intuitive, it is involved with all types of mental processes and it is purposeless.”⁹² With respect to his Wall Drawings (1969–1973), for which LeWitt only wrote the instructions to be interpreted and executed by others, he explained:

The artist conceives and plans the wall drawing. It is realized by draftsmen (the artist can act as his own draftsman). The plan is interpreted by the draftsman. There are decisions which the draftsman makes, within the plan, as part of the plan. Each individual being unique, given the same instructions, would carry them out differently. … The draftsman may make errors in following the plan without compromising the plan. All wall drawings contain errors. They are part of the work.⁹³

The machinic quality of the Wall Drawings is thus not dependent on a specific technology, but on the method of execution.

The notion of the image machine proposed here affords two forms of subjectivation—a subjectivation of the human recipient of the images, and a subjectivation of the apparatus that produces the images. By way of concluding this chapter, I want to briefly gloss a number of works that provide models for these different forms of subjectivation rooted in the materials, in the techniques, and in the bodies involved.

Herwig Weiser’s installation Lucid Phantom Messenger (2010) is a sculptural screen and transparent container filled with chemical liquids and electrodes that cause the chemicals to crystallize into exuberant and colorful plantlike shapes (figure 4.14, plate 9). The chemicals and the electrochemical processes are similar to those used in liquid crystal display (LCD) screens. But while those processes are highly controlled and timed to allow for precise visual representations, here these substances are made to coagulate and interact freely and in a lifelike, polymorphous process. The materials which are, as Matthew Fuller writes, “familiar by name to the consumer of hi-tech goods, but which are always somehow inert, territorialized into certain kinds of highly fixed behaviours, start to exhibit another kind of dynamic, catalysis, stabilizations, floods and blobs, percolation, the drool of gels, tendrils grown of crystals and polymers globulating under the influence of electrical charges and ultrasonic shivers.”⁹⁴

Figure 4.14 Herwig Weiser, *Lucid Phantom Messenger* (2010). Copyright © the artist.

Weiser’s is an image machine turned inward, without any hint at anthropomorphism. We observe a more or less pure, physical (chemical) process that seems to elicit a virtuality deeply engrained in the material, a process that is allowed the freedom of afunctionality, the machine operating and generating its images through the workings of the material.

Where such image techniques are not left to themselves but put to use with the purpose of representation, the recipient of the image is deliberately implicated. Since Petrarch’s account of his ascent to Mont Ventoux in the fourteenth century, the concept of landscape has been closely tied to the emergence of the modern subject that reflects on itself as it perceives the world, and as it perceives itself in opposition to nature. Landscape painting thus often implies the longing for an ideal or utopian return—most poignantly expressed in the romantic period and its celebration both of the sublime and of the individual.⁹⁵ It is therefore consequential that artists like Wolfgang Staehle, Ingo Günther, and David Rokeby have used the landscape motif to reflect on the mechanisms of subjectivation in machine vision systems.

In a series of projection works, the New York-based German artist Wolfgang Staehle has presented live-video images of land- and cityscapes. The first, in 1999, was an image of the Empire State Building in Manhattan. This choice was a response to Andy Warhol’s 1964 film Empire, whose eight-hour take was limited by the available stock material, and was instantly historical, whereas Staehle’s images of the building were video stills taken and updated every eight seconds and transferred electronically to the exhibition venue. Some of the following works in the series presented images from cities like New York, Berlin, Rome, while others, like Eastpoint (2004), showed landscapes and explicitly referred to historical landscape painting, especially to the Hudson River School of the nineteenth century, by the carefully chosen framing of the particular view that the projected images showed (figure 4.15). Staehle’s projection presented the same landscape in a radically different time frame, given that the image was not one of an arrested historical moment but a present view, as transient as time itself. The question about the landscape view as an event space was strangely twisted when, during an exhibition at Postmasters Gallery in September 2001, the camera pointed at the skyline of Lower Manhattan from the Brooklyn riverside, with ships and docks in the foreground and the Manhattan Bridge at the center of the horizontal image composition, also caught the 9/11 attack on the World Trade Center. The indifferent gaze of the camera was not agitated by the event, as it was not agitated by the Hudson River flowing past. The abstraction and disembodiment of the machine gaze affected not only the scene it observed but also that of the exhibition viewer. As in the classical view of Eastpoint, the simultaneity of eventscape and view imply a decoupling in time as well as in space.⁹⁶

Figure 4.15 Wolfgang Staehle, *Eastpoint* (2004), screen image. Courtesy of the artist, © VG Bild-Kunst, Bonn 2015.

Both the indifference of the gaze and the spatiotemporal decoupling are exacerbated when the perspective is shifted from anthropomorphic horizontality into the increasingly normalized verticality of images taken by satellites, airplanes, or drones. The formerly mimetic abstraction of a territory into a cartographic map is transformed in these images into the visualization of a terrain which becomes an interface for operational planning, a virtual territory that can be directly acted upon. The history of this vertical view begins in the 1850s when the French photographer Nadar took the first photographs of a battlefield from a hot air balloon, followed by pigeons and airplanes equipped with photo cameras at the beginning of the twentieth century. The military use of the images requires the introduction of aerial reconnaissance units and training in reading and interpreting such photographs, followed by the development of camouflage techniques for hiding infrastructure and weaponry from the equally medialized gaze of the enemy. An arms race of hiding on the ground and detecting from the air ensues.⁹⁷

Yet the history of the vertical, forensic gaze has not only a military but also an ecological strand to it. Measuring stretches of land, geological and biological exploration, the search for minerals and fossil resources, the control of agricultural activities, and the archaeology of human life all make use of images taken from above.⁹⁸ The expansion of the machinic gaze soon transgresses the limits of optical photography and the spectrum of human visibility to include radiographic, seismographic, and thermographic sensors, turning the surface under observation into a sounding, moving, and radiating terrain, porous, transparent, and immaterial.

One of the first artworks that explicitly addressed the iconography and aesthetics of satellite images was the installation K4 (C3I [Command Control Communication and Intelligence]) (1987) which the German-American artist Ingo Günther showed at documenta 8 in Kassel.⁹⁹ The walls and floor of the installation space are covered with a light gray marble (figure 4.16). A video projector is placed in the dark brown metal-covered ceiling and projects downward onto a massive, table-size block of marble placed at the center of the space. The video projection is the only source of light. It shows a montage of satellite images and camera pans across aerial photographs and maps, giving the impression of a sequence of strategically relevant images and diagrams that are being reviewed, analyzed, and worked upon in a military command room. There is a certain theatricality in the demonstration of the different types of images, and the occasional human arms and hands that can be seen moving across the cartographic material as it is being filmed. Yet what seems crucial here is the nonintentional and mechanical production of the images, which is not determined by the individual perspective of a human photographer but by the indifferent gaze of the vision machine. This indifference pinpoints the politics of these images and forms the basis of their potential operationality, here represented by the filmed inscription of electronic images with lines and vectors, and the gesturing hands.

Figure 4.16 Ingo Günther, *K4 (C3I [Command Control Communication and Intelligence])* (1987), installation view, Skulpturenmuseum Marl. Courtesy of the artist.

In contrast to the static camera position in Staehle’s real-time projections, and to the flyover aesthetics in Ingo Günther’s K4, a third example of such a machinic landscape is the gallery installation of David Rokeby’s Machine for Taking Time. It shows a video projection in which we see a specific landscape—in one instance it is the park of Oakville Galleries (2001–2004), in another the cityscape of Montréal (Boulevard Saint-Laurent, 2007), seen from the rooftop of one of its buildings (figure 4.17). The camera pans and zooms in slow, regular movements, from left to right, zooms in, moves further, zooms out, pans to the left, etc. In addition to this movement in space, the objects that we see also change their appearance, the video transitioning, as it seems, through different light situations and seasonal phases, with bright sunlight, cloudy skies, rain, and snow.

Figure 4.17 David Rokeby, *Machine for Taking Time (Boul. Saint-Laurent)* (2007). Photomontage by David Rokeby. Copyright © the artist.

In the course of an entire year, a digital video camera placed in a fixed position was programmed to run a particular course of directional and optical movements every day, during which course a fixed number of 1,079 still images were taken daily.¹⁰⁰ Thus an image was taken from each of the 1,079 camera positions on 365 consecutive days, with the respective light and weather conditions determining the appearance of the scene on any one of those days. For the video installation, images from this database of 365 times 1,079 images are algorithmically stitched together in real time, the resulting images in the projection moving through the “stack” not only as the camera eye moved in space, but also into the “depth” of the different temporal layers, creating the impression of a synchronicity of the seasons in the continuous space of a particular landscape setting. While this changing of the seasons is easily comprehensible for a human viewer, the repetitive precision of the movement and the ease of sliding forward and backward in time are eerie and inhuman. Rokeby’s “film” is composed of algorithmically animated still images that are calculated in real time, so the speed of movement that we see in the video is determined by the processing speed of the computer system rendering the transitions between the different data sets.

More than in the images taken from a fixed viewpoint, we are made aware that what we see is determined by the camera eye. We project our gaze into the camera eye whose rigid movements are apparently determined by a robot. As in the other examples discussed here, the camera and the recording system are not visible—only the resulting images are presented, while we have to infer the technique that led to their construction. We can imagine Rokeby’s Machine for Taking Time as an operational image system that monitors an urban or natural environment and then orders and controls specific (robotic or human) actions. The video in the projection would then be a monitoring image that is only relevant for a human observer, who is, in the face of Rokeby’s installation, given the clear sense that he or she might be only a secondary audience of the represented field of operations.

At the heart of these machine landscape dispositives is the display of a gaze that is not human but embodied by a technical device. This is an experience that was already diagnosed by Dziga Vertov, and was reiterated by Gilles Deleuze when he said of the cinematic machine of subjectivation, “The sole cinematographic consciousness is not us, the spectator, nor the hero; it is the camera, the camera [which is only] sometimes human.”¹⁰¹

A reinsertion of the human body into such a machinic assemblage of disembodiment was realized in 1984 by the Canadian artist David Tomas in the installation performance Behind the Eye Lies the Hand of William Henry Fox Talbot (figure 4.18, plate 10). It predates, by a few years, Jonathan Crary’s analysis that the construction of the modern observer-subject by apparatuses like the phenakistiscope is threefold: “The very physical position required of the observer by the phenakistiscope bespeaks a confounding of three modes: an individual body that is at once a spectator, a subject of empirical research and observation, and an element of machine production.”¹⁰²

Figure 4.18 David Tomas, *Behind the Eye Lies the Hand of William Henry Fox Talbot* (1984), installation view, S. L. Simpson Gallery, Toronto. Photo by Alex Neumann. Copyright © the artist.

In the complex yet minimalist spatial arrangement of mirrors, photo and video cameras, monitors, and furniture objects for Behind the Eye, Tomas sought to construct a version of the photographic process that would make it possible to experience the metaphorical layering of visual representation.¹⁰³ Importantly, his body was part of the installation, sitting at a drawing board, looking through a vintage 1840s camera lucida and “performing” Fox Talbot’s “pencil of nature” by making drawings that exemplified the symbolic transitions between the different media involved. The drawing human body of the artist becomes a part of the machinic dispositif, where it is not a privileged site of image creation or perception but a technical element contributing to an overall image-producing structure.

Tomas performs what Lorraine Daston and Peter Galison have described as the contested paradigm of “mechanical objectivity”; what is here an ethical and epistemological question—as it is in the examples that Daston and Galison discuss for “objective” scientific representation in the nineteenth century—becomes a question of technical functionality in operational images: these latter machinic images are sufficiently objective if the computer can act on them, when the manufacturing robot hits the screw head, and when the self-guided missile hits its designated target.¹⁰⁴

By placing the seeing human eye at the center of the installation yet degrading it to an optical device, Tomas pinpointed, in 1984, the question of the sightless gaze that Virilio would raise in his text on the Vision Machine a few years later: “Digital optics is indeed a rational metaphor for intoxication, statistical intoxication, that is: a blurring of perception that affects the real as much as the figurative, as though our society were sinking into the darkness of a voluntary blindness, its will to digital power finally contaminating the horizon of sight as well as knowledge.”¹⁰⁵ This sightless gaze is characterized by the disavowal (Farocki) of complexity and by the perfection of a statistical analysis of visual data that requires no looking and interpreting subject, but a system which, like Alphonse Bertillon’s identification signalétique of the 1880s, is characterized by the mechanized reading of coded fragments that requires neither bodies nor subjects for taking decisions and acting upon them.

Notes

4. In Taken, 100 or even 200 previously taken images of individual visitors are occasionally displayed in a screen-filling grid, but the normal mode for this left side of the double projection is the large-scale representation and “labeling” of individuals present at that moment. Rokeby had also explored the aspect of individualizing surveillance systems in works like Guardian Angel (2001), Watched and Measured (2000), and Border Patrol (the latter together with the New York artist Paul Garrin, 1995). The computer-generated designations of exhibition visitors is reminiscent of a function of Ihnatowicz’s 1973 interactive sculpture Bandit, a one-armed-bandit-style interface which gave verbal feedback on the interacting visitor’s reaction (see Catherine Mason, A Computer in the Art Room [Hindringham, UK: JJG, 2008], 94).

8. Mark B. N. Hansen, New Philosophy for New Media (Cambridge, MA: MIT Press, 2004), 10. There is a broad range of approaches in image theory that are often shaped by different theoretical and philosophical traditions, in which, not least, the different etymologies of the image-related vocabulary in different languages play an important role. For instance, whereas Gottfried Boehm conceptualizes the image as a phenomenon that achieves an autonomous status in the act of reception, both Belting and Mitchell study images in their social context; cf. the comparative discussion on Boehm, Mitchell, and Kittler in Gabriele Werner, “Nicht jede Wende zum Bildlichen meint das Gleiche. Zu Boehm, Mitchell—und Kittler,” in Bernhard J. Dotzler, ed., Bild / Kritik (Berlin: Kulturverlag Kadmos, 2010), 13–43.

14. Ibid.; the sentence from Paul Klee’s diary is quoted on p. 59. The topos of the image being alive and looking back was already pinpointed by Aby Warburg when he wrote the epigram “You are alive and you do me no harm.” See Horst Bredekamp, “‘Du lebst und thust mir nichts.’ Anmerkungen zur Aktualität Aby Warburgs,” in Horst Bredekamp et al., eds., Aby Warburg. Akten des internationalen Symposions Hamburg 1990 (Weinheim: VCH, 1991), 1–7. Cf. also Rosalind E. Krauss, The Optical Unconscious (Cambridge, MA: MIT Press, 1993), 98, on Clement Greenberg’s idea of the painting “gazing back at you”; and Georges Didi-Huberman, Ce que nous voyons, ce qui nous regarde (Paris: Editions de Minuit, 1992). Gottfried Boehm, “Die Wiederkehr der Bilder,” in Gottfried Boehm et al., eds., Was ist ein Bild (Munich: Fink, 1994), 33, reminds us that the ancient Greeks described the image as “zoon,” as “living.”

23. Virilio, The Vision Machine, 60. Neither Farocki nor Virilio make direct reference to the field of “operations research,” whereas Manuel De Landa, in his account of this area of strategic research into the analysis, preparation, and prediction of military operations, in War in the Age of Intelligent Machines (New York: Zone Books, 1991), esp. 194–203, speaks about “machine vision,” but not about “operational images.” On the context of military strategies, see also Jimena Canales, “Operational Art,” in van Tomme, Visibility Machines, 37–54. Canales reminds us of Jacques Lacan’s question of what images machines would make of the world, independently of humans (see Lacan, “A Materialist Definition of the Phenomenon of Consciousness,” in The Seminar of Jacques Lacan, ed. Jacques-Alain Miller, vol. 2, 1954–1955 [Cambridge: Cambridge University Press, 1988], 46–47), and insists that “operational images have to be seen” (39).

30. See Allan Sekula, “Steichen at War,” in Sekula, Photography against the Grain (Halifax: Nova Scotia College of Art and Design, 1984), 33–52. For an overview of computer vision systems on which the following summary is also based, see Lev Manovich, “Modern Surveillance Machines: Perspective, Radar, 3-D Graphics, and Computer Vision,” in Thomas Y. Levin, ed., CTRL SPACE: Rhetorics of Surveillance from Bentham to Big Brother (Cambridge, MA: MIT Press, 2002), 382–395, esp. the section “Computer Vision: Automation of Sight,” 390–393. An earlier version of this paper by Manovich (in which the section on computer vision is more or less identical) was published as “The Automation of Sight: From Photography to Computer Vision,” in Timothy Druckrey, ed., Electronic Culture: Technology and Visual Representation (New York: Aperture, 1996), 229–239. Cf. also E. R. Davies, Computer and Machine Vision: Theory, Algorithms, Practicalities (New York: Academic Press, 2012).

35. See Horst Bredekamp, Theorie des Bildakts. Frankfurter Adorno-Vorlesungen 2007 (Frankfurt am Main: Suhrkamp Verlag, 2010). Cf. Inge Hinterwaldner, “Zur Operativität eLABORierter Wissenschaftsbilder,” in Felix Mittelberger et al., eds., Maschinensehen. Feldforschung in den Räumen bildgebender Technologien (Leipzig: Spector Books, 2013), 51–61, who conceives of operationality as a form of mediation in scientific discourses and, following Bruno Latour, as the field of things that can be done with images. Nina Samuel, in “Images as Tools: On Visual Epistemic Practices in the Biological Sciences,” Studies in History and Philosophy of Biological and Biomedical Sciences 44 (2013): 225–236, understands “operational images” as “tools” that can be used either by human or by technical agents. A similar conception underscored the exhibition “Operational Images,” Centre de la Photographie Genève, March 2011.

36. See Sybille Krämer, “Operative Bildlichkeit. Von der ‘Grammatologie’ zu einer ‘Diagrammatologie’? Reflexionen über erkennendes Sehen,” in Martina Hessler and Dieter Mersch, eds., Logik des Bildlichen. Zur Kritik der ikonischen Vernunft (Bielefeld: transcript, 2009), 94–123. While Krämer explicitly excludes operational (or operationally efficacious) images from the field of operational pictoriality, the set of six characteristics she proposes for the latter might form a useful point of departure, or diagram, for analyzing the functionality and aesthetics of operational images. These six aspects are: planeness and simultaneity (Flächigkeit und Simultanität), directionality (Gerichtetheit), graphism (Graphismus), syntacticity (Syntaktizität), referentiality (Referenzialität), and operationality (Operativität). The latter implies, for Krämer, that phenomena of operational pictoriality are not only manageable and explorable, but that they are also generative and that they constitute their objects—something that is particularly the case in objects of knowledge and of theory in which there is an interlocking of visualization, operationalization, and generation.

61. Van Doesburg, in his manifesto “Art Concret” (1930), claimed that the concrete work of art should refer not to nature or emotions but only to itself: “The construction of the picture, as well as its elements, should be simple and controllable visually. Technique should be mechanical, that is to say exact, anti-impressionistic” (cited in Margit Rosen, ed., A Little-Known Story about a Movement, a Magazine, and the Computer’s Arrival in Art: New Tendencies and Bit International, 1961–1973 [Cambridge, MA: MIT Press, 2011], 44). A similar formal and conceptual aesthetic in the 1960s influenced artists like Agnes Martin, Sol LeWitt, and Gary Kuehn. In the field of sculpture, this practice points backward to Brancusi, Naum Gabo, and Le Corbusier—thus the references suggested by Lewis Mumford in Art and Technics (1952)—and forward to kinetic sculptures and environments by David Rokeby, like Cloud (2007) or Long Wave (2009).

63. Vera Molnar, “Meine Arbeit mit dem Computer,” in Horst Rave and Dagmar Weste, eds., Vera Molnar. Neun Quadrate (Bonn: gkg Gesellschaft für Kunst und Gestaltung, 1991), 8–9 (my translation from the German). On the “machine imaginaire,” see Vincent Baby, in Véra Molnar—une retrospective, 1942–2012 (Paris: Chauveau, 2012), 11–13. Molnar recounts that she had heard of computers in the 1950s, but what she could read at the time gave her the impression that it was impossible to get access to such a machine for her artworks—books like Edmund C. Berkeley’s Giant Brains, or Machines That Think (1949), or Louis Couffignal’s Les Machines à penser (1952), or the many articles in popular magazines like Time which, in January 1950, ran a title story on the Mark III computer and the “computer men” (Peter Galison, “The Ontology of the Enemy: Norbert Wiener and the Cybernetic Vision,” Critical Inquiry, no. 21 [1994]: 312). However, she came to see that a computer could help realize her desire to work strictly systematically. Molnar says that she cannot remember a specific moment when she had the idea for the machine imaginaire. Her earlier work had been equally systematic in its procedures, so we can assume that the “invention” of the machine imaginaire was for Molnar a moment of formalizing an existing method and coining a term for it, rather than the creation of a completely new procedure. (Vera Molnar, in conversation with the author, Paris, October 22, 2007.) The same principle was for instance used by Herman de Vries in the 1960s (see Rosen, A Little-Known Story, 33, and the handwritten document on p. 247) and has, if we follow theoretician Sybille Krämer, a much older history: “Long before the computer as a universal medium and programmable machine, we developed the computer ‘in us’, understood here as a machinelike handling, relieved of interpretation, of signs on paper.” Sybille Krämer, “Schriftbildlichkeit, oder: Über eine (fast) vergessene Dimension der Schrift,” in Horst Bredekamp and Sybille Krämer, eds., Bild—Schrift—Zahl (Munich: Fink, 2003), 171.

69. Cf. Barbara Büscher, H.-C. von Herrmann, and C. Hoffmann, eds., Ästhetik als Programm. Max Bense—Daten und Streuungen, special issue of Kaleidoskopien 5 (2004); and Günther Pfeiffer, Kunst und Kommunikation (Cologne: DuMont, 1972), 173–187. Cf. also Max Bense’s summary account in “Aesthetics and Programming” (1968), rpt. in Rosen, A Little-Known Story, 296–299, and Frieder Nake’s critical report, “There Should Be No Computer Art” (1971), reprinted in ibid., 466–468. The media historian Martin Warnke speaks of the “heroic failure” of Bense’s and Birkhoff’s attempt at a mathematically formalized aesthetics; see Martin Warnke, “Ästhetik des Digitalen,” in Hedy Graber and Dominik Landwehr, eds., Kultur digital—Begrife, Hintergründe, Beispiele (Basel: Christoph Merian, 2011).

71. Molnar, “Meine Arbeit mit dem Computer,” 8–9; Molnar claims that there is no rupture in her oeuvre, either by the introduction of the machine imaginaire around 1959 or by the use of the computer from 1968 onward. However, the works gained in complexity, and the computer seems to have helped her push the boundaries of disorder and chaos. See also Rosen, A Little-Known Story, 34–35, on the discussion among New Tendencies artists about the differences between human and computer execution of a drawing; an important point of debate (and of division between artists and engineers in the group) was the impression that the visual outcome was put into question and that the program gained priority over the resulting designs; A. Michael Noll claimed that “artistic originality in the sense of improbability can be simulated perfectly” (ibid., 35).

75. See Tilman Baumgärtel, Plug In, and Büro Friedrich, eds., Install.Exe/JODI (Basel: Christoph Merian, 2002). The “postdigital” concepts of “Glitch Aesthetics” and “New Aesthetic” derive from these early experiments of “net.artists”; art historically, they refer back not only to the experiments in electronic art from the 1960s and 1970s, but also to the constructivist exploration of faktura in the 1920s. The argument put forward by Tim Barker in his essay “Aesthetics of the Error” points in a similar direction, yet it is marred by a lack of precise definitions of the key terms of “machine” and “error.” What Barker calls “errors” are in fact functions and potentials of the computer (explored by artists such as JODI, Yann Le Guennec, or Cory Arcangel), not accidents or uncontrolled aberrations—which is why Barker’s references to such “predecessors” as Hans Arp, Bruno Munari, Robert Rauschenberg, or John Cage are also imprecise. Pias insists that “every program that runs is legitimate. … Hacking subverts the notions of correct or incorrect usage; it deconstructs, in a certain way, the ‘misuse’ itself by showing that a concept of technical function that is tied to a human intentionality of purposes does not make sense with regard to computers.” Claus Pias, Computer Spiel Welten (Zurich: diaphanes, 2010), 84.

76. See Jonathan Crary, Techniques of the Observer (Cambridge, MA: MIT Press, 1990). For a commentary on Crary’s theses, see Friedrich Kittler, Optical Media (London: Polity, 2010), 145–154. For a detailed study of the history of experimental research on visual perception, see Romana Karla Schuler, Seeing Motion: A History of Visual Perception in Art and Science (Berlin: De Gruyter, 2016). For studies of the different media apparatuses developed during the nineteenth century, see Bodo von Dewitz and Werner Nekes, eds., Ich sehe was, was du nicht siehst. Sehmaschinen und Bilderwelten. Die Sammlung Werner Nekes (Munich: Steidl, 2002); Erkki Huhtamo, Illusions in Motion: Media Archaeology of the Moving Panorama and Related Spectacles (Cambridge, MA: MIT Press, 2013). For an overview of a number of “vision machines” of the nineteenth and twentieth centuries, among others those of Friedrich Kiesler and Alfons Schilling, see Peter Weibel, “Contours of a History of the Theory and Art of Perception in Austria,” in Peter Weibel, ed., Beyond Art: A Third Culture (Vienna: Springer, 1996), 16–32.

97. See Virilio, The Vision Machine, for a reflection on the development of electronic decoy techniques for visual operations in electronic spaces, which followed such territorial methods that responded to photographic surveillance. On cartography, see Cartes et figures de la terre, exh. cat. (Paris: Centre Pompidou, 1980). On aerial reconnaissance, see Christoph Asendorf, Super-Constellation—Flugzeug und Raumrevolution: die Wirkung der Luftfahrt auf Kunst und Kultur der Moderne (Vienna: Springer, 1996). Artists like László Moholy-Nagy and György Kepes were involved in US aerial reconnaissance and camouflage efforts during the Second World War; see László Moholy-Nagy, Vision in Motion (Chicago: Paul Theobald, 1947). The belated analysis of aerial photographs of the Nazi concentration camp in Auschwitz forms an important motif in Harun Farocki’s film Images of the World (1988); cf. Farocki, “War Always Finds a Way,” 110–111, and Volker Pantenburg, Film als Theorie: Bildforschung bei Harun Farocki und Jean-Luc Godard (Bielefeld: transcript, 2006).

4 Image Machine

Machines that See