2 A Taxonomy of Computer Art
Margaret A. Boden and Ernest Edmonds
2.1 Introduction
Since the late-1950s, an ever-diversifying set of novel art practices has arisen that is still little known or discussed in aesthetics and art theory. (For examples of these practices, see Krueger 1991; Wilson 2002; Candy and Edmonds 2002; Whitelaw 2004; Woolf 2004; Popper 2007.) As Jon McCormack, one of the artists concerned, has put it, “Much of the innovation today is not achieved within the precious bubble of fine art, but by those who work in the industries of popular culture—computer graphics, film, music videos, games, robotics and the Internet” (McCormack 2003, 5).
The “precious bubble” of fine art is a (shifting) socially accepted norm. But artists often work outside the norm of their day as famously illustrated by Marcel Duchamp and his ready-mades or John Cage’s use of silence. And sometimes the bubble eventually expands so as to engulf the previously maverick efforts. The Impressionists, for instance, no longer have any need for a Salon des Refuses. They do not even need a salon: their images assail us every day on calendars and chocolate boxes. Whether the innovations mentioned by McCormack will one day be included in the expanding bubble remains to be seen. Their fate, in this regard, depends partly on how people—both curators and the general public—respond to the controversial aesthetic and philosophical questions raised in section 2.4.
The novel approaches involved here are closely interrelated, both theoretically and methodologically. So much so, indeed, that they are often all lumped together under a label, such as computer art, electronic art, process art, or generative art. This chapter aims to clarify how they can be distinguished.
2.2 Origins and Interrelations
From the theoretical point of view, this new art originated in cybernetics and general systems theory. The young painter Roy Ascott, later to be highly influential in the field, identified the novel activity as “a cybernetic vision” in 1966 (Ascott 2003b, chap. 3; see also Mason 2008, chap. 4). And the exceptionally creative cybernetician Gordon Pask was a key influence. Besides producing or imagining some of the first artworks of this general type (in the 1950s), he provided much of the theoretical impetus that inspired the more philosophically minded artists in the field (Boden 2006, 4.v.e; Mason 2008, chap. 2).
Very soon, the “cybernetic vision” was bolstered by ideas about structure and process drawn from computer science. Ernest Edmonds, for instance, turned from paintbrush and easel to the computer in the 1960s; he thought he could produce more interesting art in that way (see Boden and Edmonds 2009, sect. iii). At much the same time, music and visual art was produced that reflected the computational theories of mind of artificial intelligence (AI). Indeed, Harold Cohen, a renowned abstract painter in 1960s London, deserted his previous working practices largely because he felt that doing computer art would help him understand his creative processes better (McCorduck 1991; Boden 2004, 150–166, 314–315).
Over the last twenty years, this artistic field has been inspired also by ideas about emergence, evolution, embodiment, and self-organization. These concepts are borrowed from cognitive science and in particular from artificial life (A-Life). However, the theoretical roots (and the pioneering experiments) of A-Life reach back to midcentury cybernetics and automata theory (Boden 2006, 4.v.e, 15.iv–v). In short, the theoretical wheel has turned full circle.
The methodological wheel, meanwhile, has climbed an ascending spiral. Art practices outside the fine-art bubble are grounded in technologies for communication and information processing whose power and variety have burgeoned over the last half century. Often, this means that the customary lone artist is replaced by a team, some of whose members may be computer scientists or tele-engineers.
Most of them rely heavily on digital computing and in particular on methods drawn from AI and A-Life. Specifically, they have employed symbolic and connectionist computation and, more recently, cellular automata, L-systems (automatically branching structures that botanists use to study plant form and physiology), and evolutionary programming too. This is an ascending spiral, not a linear ascent, because two of those recent methods were foreseen (by John von Neumann) in 1950s cybernetics, and all three had been mathematically defined by the 1960s (Boden 2006, 15.v–vi). But none could be fruitfully explored, by artists or by scientists, until powerful computers became available much later.
The resulting artworks are highly diverse. They include music, sonics, the visual arts, video art, multimedia installations, virtual reality, kinetic sculpture, robotics, performance art, and text. And whereas some of these outside-the-bubble activities place ink or paint onto a surface, others involve desktop visual display units (VDUs) or room-scale video projection. Yet others eschew the virtuality of cyberspace, constructing moving physical machines instead.
The labels attached to these new art forms vary and have not yet settled down into a generally accepted taxonomy. The names preferred by the artists involved include generative art, computer art, digital art, computational art, process-based art, electronic art, software art, technological art, and telematics. All these terms are commonly used to denote the entire field—and although distinctions are sometimes drawn, they are often treated as synonyms.
With respect to the labels computer art and generative art, that was true right from the start. These terms have been used in tandem and more or less interchangeably since the very earliest days. The first exhibition of computer art, held in Stuttgart in February 1965, was called “Generative Computergraphik” (Nake 2005). It showed the work of Georg Nees, who wrote the first doctoral thesis on computer art, giving it the same title as the exhibition (Nees 1969). That thesis was widely consulted by the small but growing community, harnessing the words “generative” and “computer” together in its readers’ minds.
Their near equivalence was reinforced in November 1965, when an exhibition (again in Stuttgart) included both Nees’s work and the early computer graphics of Frieder Nake. Both men applied the term “generative” to their work and used it to identify art that was produced from a computer program and, hence, at least in part produced automatically. Others who were pioneering the activities outside McCormack’s bubble also adopted the term. For example, when Manfred Mohr started producing drawings with a computer program in 1968 he termed it “generative art” (Mohr 1986). And the philosopher Max Bense—who had composed the manifesto for the original Stuttgart exhibition of 1965—was writing about “generative aesthetics” (Nake 1998).
The label generative art is still current within the relevant artistic community. Since 1998 a series of conferences have been held in Milan with that title (see http://
With the recent appearance of art using methods drawn from A-Life (for examples, see Whitelaw 2004; Tofts 2003; 2005, 80–103; Popper 2007, 118–129), the label generative art, as used in the community concerned, has acquired biological overtones. In biology, “generative” is common in discussions of morphological development, of growth in plants and animals, and in references to reproduction. One or both of those meanings are sometimes explicitly stressed by self-styled generative artists whose work focuses on emergence, self-organization, or evolution. McCormack himself is one such example (e.g., Dorin and McCormack 2001; McCormack, Dorin, and Innocent 2004). Even so, the formal mathematical sense remains a core aspect of the label’s meaning.
Despite the continuing assimilation of the terms “generative” and “computer” art, one should not assume that these are exactly the same thing. According to the taxonomy given in section 2.3, not all generative art involves computers. To the contrary, the generative processes involved vary widely in type—some of which were already used by artists hundreds of years ago.
Even in the 1960s, however, alternative tags for this general area were already being offered, and more have been suggested since then. An influential discussion by the art historian Jack Burnham (1968), for instance, identified the new work, overall, as process art—a label that is still in use fifty years later.
Since Burnham’s discussion, the processes involved in process art have diversified hugely. As a result, and in addition to varying labels for the entire extrabubble field, there are today many names for subfields. These more discriminating categories include interactive art, evolutionary art, video art, media (and new-media and multimedia) art, holographic art, laser art, virtual art, cyborg art, robotic art, telerobotics, and net art. Again, however, the extension of these labels (the scope of the subfields) is not always clear.
It is partly for that reason that “a satisfactory critical framework of new forms in art technology has yet to be developed” (Candy and Edmonds 2002, 266). The distinctions made in this chapter should help toward such a framework.
There is a caveat, however. The definitions given in section 2.3—for instance, of computer art, generative art, evolutionary art, robotic art, and interactive art—use words that are already being used by the artists in question. Indeed, this analysis may help readers interpret these artists’ discussions of their own work. But the aim is not to offer a report of common usage: such usage is not consistent. Rather, this new taxonomy is intended as a theoretical tool to highlight certain distinctions that have aesthetic or philosophical interest. As we see in section 2.4, judgments concerning creativity, authorial responsibility, agency, autonomy, authenticity, and (sometimes) ontology are even more problematic outside the precious bubble than inside it.
2.3 The Taxonomy
Fourteen types of art are distinguished in the taxonomy given here. They are Ele-art, C-art, D-art, CA-art, G-art, CG-art, Evo-art, R-art, Net-art, I-art, CI-art, VR-art, LC-art, and 3DP-art.
Some of these activities, having been located within the classification, are then ignored. In other words, certain types of computer art are identified in the taxonomy but not further discussed in this chapter or elsewhere in this book. Most of the attention is paid to the forms of CG-art, because these raise the most interesting philosophical issues.
This taxonomy is a decidedly non-Linnaean structure. For one thing, the definitions given in this chapter, like most definitions, admit borderline cases and even anomalous counterexamples. And for another, there is no neat and tidy hierarchy of genus and species within which these twelve types can be located. Although there are some part-whole relations here, there are also untidy overlaps.
One type of overlap concerns links with more traditional, or familiar, categories of art. Most cases of such art do not fall under this new classification at all. But some of the classifier concepts—namely, G-art, I-art, Evo-art, and R-art—cover artworks both inside and outside McCormack’s “precious bubble.” Admittedly, those that lie inside the bubble are relatively maverick examples, as we will see. Indeed, some of them (produced by the conceptual artists) were specifically intended to undermine the commonly accepted notion of fine art in terms of which the bubble is defined. In McCormack’s usage, however, all examples of noncomputer art are located inside the bubble.
A list summarizes the definitions at the end of this section.
Let us start with electronic art, or Ele-art. This wide concept covers (df.) any artwork whose production involves electrical engineering or electronic technology. So it ranges from simple analogue devices of the 1950s and 1960s such as Pask’s Musicolour and Colloquy (Pask 1971; Mallen 2005) and Edward Ihnatowicz’s kinetic sculpture SAM (Zivanovic 2005, 103)—all pioneering examples of interactive art, or I-art—to the highly sophisticated man-robot integrations embodied by the performance artist Stelarc (Smith 2005). And along the way it covers the whole of computer art and media art, including examples that exploit the advanced computational techniques of virtual reality.
Ele-art does not include art that is produced only by humans but published by electronic means. So it excludes Japan’s Keitai novels, for instance. These are stories published online for download to mobile phones. Some are so popular—more than two million downloads in their first week—that they are issued in book form by traditional publishers. In 2007, half of Japan’s top ten fiction best sellers originated in this way. By contrast, some of the Manga-art comics now being published for mobile phones do count as Ele-art—namely, those involving animated graphics or interactivity. In some cases, the phone buzzes when the reader comes to a tense moment in the action depicted on the screen. A Keitai novel could of course be adorned with such buzzes and would then count as a marginal example of Ele-art—only marginal because these buzzes are superficial, and their loss (if the text were later published in book form) would minimally alter the reader’s experience.
Unlike mechanical art, such as Leonardo da Vinci’s metal lion, who “after he had a while walked vp and downe, stoode still opening his breast, which was all full of Lillies and other flowers of diuers sortes” (Marr 2003, n66), electronic art could not appear until the mid-twentieth century. But as the previous paragraph implies, the technologies concerned have diversified richly since then. Accordingly, the highly inclusive label Ele-art is not very interesting for our purposes.
Surprisingly, perhaps, neither is the concept of computer art, or C-art: (df.) art in whose productive process computers are involved. This concept is apt for general art-historical purposes, because it covers every example that anyone might want to call computer art—including many that are commonly given other labels. It is less useful for us here, however, for two reasons.
First, it includes analogue as well as digital computers. Some of the earliest C-art work combined digital methods with specially built analogue devices. Ihnatowicz’s giraffe-like kinetic sculpture Senster is a case in point (Zivanovic 2005; Mason 2008, chap. 5). As for analogue computers, these were used in the early days in, for example, visual arts such as Ben Laposky’s work of the 1950s (Laposky 1969) and electronic music at the same time, famously encouraged by the invention of the Moog synthesizer (Pinch and Trocco 2002).
Today, a few computer artists sometimes employ analogue processes— namely, electrochemical reactions like those pioneered by Pask. Some of their work, including Pask-inspired “sculptures” by Richard Brown (2001, 2006) and Andy Webster (2006), featured in the 2007 Edinburgh exhibition Gordon Pask and His Maverick Machines. In addition, a video on this theme titled “Tuning Pask’s Ear” has been shown in several European art galleries (Webster and Bird 2002). But analogue computers are another matter—and are very rarely used by artists today. Because of the huge flexibility that is afforded by the general-purpose nature of digital computers, it is those machines that underlie most C-art. Indeed, to speak of computer art is typically to assume that digital computers are being used.
In other words, computer art is tacitly classed as digital art, or D-art. D-art (df.) uses digital electronic technology of some sort. It includes not only artworks generated by computers but also digitally manipulable (but human-produced) music and video. Common usage sometimes treats “digital art” and “computer art” as near synonyms. In this taxonomy, however, they are analytically distinct with most but not quite all, C-art being included within D-art. If the word “electronic” were removed from our definition, the nineteenth-century Pointillists would count as D-artists; their pictures were composed not of continuous brushstrokes or color washes but of myriad individual spots of paint.
D-art is more wide ranging than may appear at first sight. For instance, some C-artists use visual software that is intuitively analogue and so relatively natural to work with. One example is continuous vector mapping, used instead of pixel editing (Leggett 2000). But they are relying on methods and hardware that are digital at base. In fact, most people who today say that they are using an analogue method (i.e., an analogue virtual machine) would actually be working on a digital computer, used to simulate an analogue computer.
Similarly, most neural networks, or connectionist systems, whether used by cognitive scientists or by computer artists, are simulated on von Neumann machines. That is true, for instance, of Richard Brown’s interactive Mimetic Starfish, a millennial version of Senster that was exhibited around the world and was described by The Times (2000) as “the best bit of the entire [Millennium] dome.” The starfish was built by engineering visual imagery, not metal: it is a purely virtual creature (an image projected onto a marble table) that responds in extraordinarily lifelike ways to human movements. And it is generated by a self-equilibrating connectionist system (programmed and built by Igor Alexander). In short, digital technology reaches further than one might think.
The second reason the definition of C-art given is too catholic for our purposes is that it includes cases in which the computer functions merely as a tool under the close direction of the artist, rather like an extra paintbrush or a sharper chisel. Artists in the relevant community sometimes speak of this as computer-aided or computer-assisted art, contrasting it with what they call computational art—where the computer is more of a participant, or partner, in the art making (e.g., P. Brown 2003, 1). Let us call this CA-art, wherein (df.) the computer is used as an aid (in principle, nonessential) in the art-making process.
Consider video art and music videos, for instance. These popular outside-the-bubble activities qualify as CA-art in this sense. The human-originated images and music are digitally stored and (usually) manipulated or transformed by the artist, using the computer as a tool. Other cases of CA-art include someone’s doing a line drawing by hand on the computer screen and then calling on a coloring program such as Photoshop to produce a limited edition of identical prints—or a unique image. This is an upmarket form of paint by numbers, wherein the hues for each area are chosen by the individual artist. Yet other examples include visual collages composed from image libraries and computer music that is so called because it uses electronic synthesizers and virtual instruments.
In practice, the computer aid may be necessary for the art making. It is impossible, for instance, to alter video images in certain ways except by using a computer. Similarly, some visual effects delivered by Photoshop could not have been produced by using oils, water colors, or gouache. The renowned artist David Hockney accordingly described Photoshop as “a fantastic medium” (Hockney 2009). And synthesized computer music exploits sounds that had never been heard before synthesizers were developed. Nevertheless, the computer is not essential in principle. The relevant visual or sonic effects are specifically sought by the human artist and might conceivably have been produced in some other way. Much as a species with iron-hard fingernails would not need chisels, so our vocal cords (or wood, metal, or cats’ sinews) might have been able to produce the sounds produced by synthesizers.
The subclass of C-art that is of interest in the present context is the type in which the computer is not used as a tool to effect some preexisting idea in the artist’s mind but is in a sense (just what sense is explored in section 2.4) partly responsible for coming up with the idea itself. In other words, the computer art that is most relevant here is a form of generative art, or G-art. In G-art, (df.) the artwork is generated, at least in part, by a process not under the artist’s direct control.
This is a very broad definition. It does not specify the minimal size of the part. It does not lay down just what sort of generative process is in question. It does not say what counts as being outside the artist’s direct control. And it is silent on whether the processes concerned were deliberately molded by the artist before losing direct control. In short, this definition of G-art is largely intuitive. In general, it picks out cases of art making in which personal control is deliberately diminished, or even wholly relinquished, and relatively impersonal processes take over.
Those impersonal processes vary greatly. They may be physical, psychological, sociocultural, biological, or abstract (formal). And if abstract, they may or may not be implemented in a computer.
For example, in the dice music written by Haydn and Mozart, the exact order of the precomposed phrases was decided by throwing a die. Although a human threw the die voluntarily, he or she could not influence, still less determine, just how it fell. That was due to purely physical forces.
Such forces also influenced some visual generative art that predated computers. One clear example is Kenneth Martin, whose 1949 abstract painting used basic geometrical figures (squares, circles, diagrams) and rules of proportion (Martin 1951). Later, his Chance and Order and Chance, Order, Change series combined rule-driven generation with random choice. Although the basic forms were laid down by the rules that Martin had deliberately devised, chance physical events—such as picking a number out of a hat—determined the actual course of the work.
As for generative literature, this too may involve chance events dependent on physical processes. The various versions of Bryan Johnson’s (1969) novel The Unfortunates are produced in this way. The novel was published as twenty-seven separate sections in a box: all but the first and last were to be read in a random order, decided by shuffling or dice throwing. Many other examples of interactive stories, partly narrated by the reader, have been produced since then (Montfort 2003). Most of these depend not on physical processes but on deliberate voluntary choices by the reader-author; however, some are partly generated by dice throwing and the like.
Arguably, G-art produced by physical forces can be found inside McCormack’s bubble, too. Given the phrase “at least in part,” used earlier to describe a computer’s share in Nees’s and Nake’s work, one might say that Jackson Pollock’s paintings exemplified G-art grounded in physics. Although he certainly was not throwing (still less choosing) paint at random, he did not have direct control over the individual splashes as he would have had over marks made with a paintbrush.
Even more control was lost, or rather deliberately sacrificed, when Hans Haacke, in the 1960s, began to exploit and even to highlight the physical behavior of water, vapor, and ice; of waves; and of weather conditions. He wanted to make “something which experiences, reacts to its environment, changes, is nonstable …, always looks different, the shape of which cannot be predicted precisely” (Lippard 1997, 38, 64–65). He saw these works not as art objects but as “ ‘systems’ of interdependent processes” that evolve without the viewer’s interaction or “empathy” so that the viewer is a mere “witness.” A few years later, Jan Dibbets placed eighty sticks in the sea, a few inches below the surface, and watched them oscillate in the water from fifty feet above: “That,” he said, “was the work” (Lippard 1997, 59).
The Surrealists of the 1920s, by contrast, had exploited psychological processes—but their work counts as G-art because these were of a relatively impersonal kind. Inspired by Freud, they engaged in automatic writing and painted while in trance states to prioritize the unconscious mind, which Andre Breton declared to be “by far the most important part [of our mental world].” Indeed, Surrealism was defined by Breton as “pure psychic automatism [sic] by which one proposes to express … the actual functioning of thought, in the absence of any control exerted by reason, exempt from all aesthetic or moral preoccupations” (Breton 1969). The unconscious thought was taking place in a person’s mind, to be sure, but voluntary choice and personal “preoccupations” (i.e., the reality principle and ego ideals) were not directing it.
The conceptual artist Sol LeWitt was also recommending G-art when he said that art should be designed by some formulaic rule. The crucial idea, he said, “becomes a machine that makes the art,” in which “all of the planning and decisions are made beforehand and the execution is a perfunctory affair” (LeWitt 1967, 824). Once the plan has been chosen, “The artist’s will is secondary to the [art-making] process he initiates from idea to completion” (LeWitt 1969, item 7; italics added). He even added that the artist’s “wilfulness may only be ego.” That art-making process was nevertheless psychological, in the sense that the implications of his abstract rules were discovered not by computers but by conscious reasoning.
Sociocultural processes—in the form of the US postal system—produced Douglas Huebler’s artwork called 42nd Parallel. Items were posted from fourteen towns spanning 3,040 miles on latitude 42° north, all sent to Truro, Massachusetts. The work, according to Huebler, was not the conception in his mind, or the posted items, or even the acts of posting. Rather, it was the widespread pattern of activity within the US postal system. But, he said, the work was “brought into its complete existence” through documents: the certified postal receipts (for sender and for receiver), and a map marked with ink to show the geographical relations of the fifteen towns. Its nature as G-art is evident in his remarks that “an inevitable destiny is set in motion by the specific process selected to form such a work, freeing it from further decisions on my part,” and “I like the idea that even as I eat, sleep, or play, the work is moving towards its completion” (Lippard 1997, 62).
The artist Hubert Duprat turned to biology for constructing a work of art. He put dragonfly larvae into an aquarium containing not pebbles and pondweed but tiny flakes of gold and a few small pearls, opals, and sapphires and left them to sculpt opulent little protective cases, held together by caddis silk (Duprat and Besson 1998). Some thirty years earlier, Haacke too had turned to biology. He experimented with the growth of grass and the hatching of chickens (as well as with water and weather) to make something “natural,” which “lives in time and makes the ‘spectator’ experience time” (Lippard 1997, 38). Mavericks though they were, both Duprat and Haacke were working inside McCormack’s bubble.
Others have even exploited physical and biological degeneration to produce their G-art. The environmental sculptor Andy Goldsworthy sometimes highlights effects caused by undirected physical change, in his gradually melting ice sculptures, for example. And in Gustav Metzger’s “auto-destructive” art (notorious for an overnight gallery cleaner innocently throwing Metzger’s bag of rotting rubbish into the dustbin), the point of the exercise is to remind us of the deterioration that awaits all human constructions—and human beings, too (Metzger 1959, 59; 1965). He was thinking not only of biological decay but also of the terrible destructive power of the Cold War arms race. The artwork is usually assembled by a human artist (or sometimes, Metzger said, by machines in a factory). But it attains its final form, and its significance, through the natural processes of damage and decay.
Such inside-the-bubble (albeit unorthodox) cases, however, are not what the new artists normally have in mind when they refer to generative art. Their phraseology is borrowed from mathematics and computer science, with which the maverick artists just named were not concerned. These disciplines see generative systems as sets of abstract rules that can produce indefinitely many structures or formulae of a given type and that—given the Church-Turing thesis (Boden 2006, 4.i.c)—can in principle be implemented in a computer. The G-art community outside the bubble put this principle into practice. That is, their art making rests on processes generated by formal rules carried out by computers—as opposed to physical, biological, or psychological processes or abstractions personally discovered by conscious thought.
In other words, the instances of G-art that most concern us here are those that are also instances of C-art. They are computer-generated art, or CG-art.
A very strict definition of CG-art would insist that (df.) the artwork results from a computer program being left to run by itself, with zero interference from the human artist. The artist (or a more computer-literate collaborator) writes the program but does not interact with it during its execution. In effect, the artist can go out for lunch while the program is left to do its own thing.
Such cases do exist. Cohen’s AARON program (described later) is one well-known example. Nevertheless, that definition is so strict that it may be highly misleading. Most people working in, or commenting on, generative art allow a compromise in the core concept so as to include interactive art. That is such a prominent subclass of what is called generative art that, even though the taxonomy given here does not aim to capture common usage, it would be highly anomalous to exclude it.
To be sure, the definition of CG-art does cover most interactive art, because it insists on zero interference from the human artist rather than from any human being, whether artist or audience. However, it would be very easy for readers to elide that distinction, which in any case makes a questionable assumption about authorial responsibility (discussed in section 2.4). Moreover, the overly strict definition of CG-art excludes those cases (whether inside the bubble or outside it) wherein artists rely on their intuitive judgment to make selections during an artwork’s evolution.
It is preferable, therefore, to define CG-art, less tidily, as art wherein the artwork results from some computer program being left to run by itself, with minimal or zero interference from a human being. “Minimal,” of course, is open to interpretation. It necessitates careful attention to just what interference goes on and by whom in any particular case.
Most of what people call computer art is CG-art, in this sense. Indeed, the phrases “computer art” and “generative art” are often regarded as synonyms. Notice, however, that in our terminology not all C-art is CG-art. CA-art is not, because the computer is used as a tool subject to the artist’s hands-on control and is of no more philosophical interest than a paintbrush or a chisel.
Admittedly, the distinction between CA-art and CG-art is not always so clear-cut as in the CA-art examples mentioned earlier. We see in chapter 6, for instance, that a program for simulating various painting materials and styles can be run either more or less autonomously (Colton, Valstar, and Pantic 2008). More, and it counts as CG-art; less, and it is better seen as CA-art—although even so, it relies heavily on CG-processes. In general, CA-art may involve AI agents: programs called on by the human artist to aid in specific ways during the production of an artwork (Boden 1994; 2006, 13.iii.d). The more that the ongoing direction is assumed by the computerized agents, not by the human being, the closer the project is to CG-art.
CG-art is intriguing on two counts. First, the generality and potential complexity of computer programs means that the possible space of CG-artworks is huge, indeed infinite. Moreover, most of the structures in that space will be images or music that the unaided human mind could not have generated or even imagined as the artists themselves admit. CG-literature can be ignored here: unless it is heavily interactive, it is much less successful, because the relevant knowledge of language and of the world is too rich to be implemented in computers.
A skeptic might object that much the same is true of a trumpet or a cello: not even the most skilled stage impressionists could mimic these instruments plausibly. In short, human artists often need help from machines. Trumpets, computers—what is the difference? Well, one important difference has just been mentioned: the generality of digital computers. In principle, these machines can (and do) offer us an entire symphony orchestra and an infinite set of visual images and sculptural forms—indeed, an infinite range of virtual worlds. McCormack (2003, 7) goes so far as to compare this infinite space of possibilities, way beyond our comprehension, with the Kantian sublime.
The second point is even more pertinent. Whereas there is no interesting sense in which a trumpet or a cello can be left to do its own thing, a computer certainly can. And it is part of the definition of CG-art that this happens. As we see in section 2.4, this aspect of CG-art raises some tricky problems concerning concepts such as autonomy, agency, creativity, authenticity, and authorial responsibility.
Especially well known cases of CG-art are the successive versions of AARON. This is a drawing-and-coloring program developed over the last forty years by the one-time abstract painter Cohen (1995, 2007) and whose works are exhibited at venues all around the world—including the Tate Gallery. It has shown clear progression along aesthetic dimensions. Indeed, Cohen (personal communication) describes the 2006 version as a “world-class” colorist, whereas he himself is merely a “first-rate” colorist: “I wouldn’t have had the courage to use those colours,” he has said. Earlier versions of AARON mixed liquid dyes and used painting blocks of five different sizes to place them on paper; a later version printed out colors instead of using liquids, but these colors too were mixed at the program’s behest. And the current implementation of AARON shows the development of the images in real time, projected onto an electronic screen.
An almost equally famous example is Emmy, a computer musician developed over much the same period by the composer David Cope (2001, 2006). This generates music in the styles of many renowned composers and very convincingly, too (see Hofstadter 2001, 38–39). Cope has recently abandoned it, because of the prejudiced reactions of audiences (see section 2.4).
Both of those programs were based on methods drawn from what the philosopher John Haugeland (1985) dubbed GOFAI, or good old-fashioned AI (see Boden 2006, chaps. 10 and 13). However, the mature Emmy also uses connectionist AI (Boden 2006, chap. 12). More recent methods for constructing CG-artworks, as remarked in section 2.2, include cellular automata, L-systems, and evolutionary programming, all widely used in A-Life research (Boden 2006, chap. 15).
Cellular automata are systems made up of many computational units, each following a small set (usually the same set) of simple rules. In such systems, surprising patterns can emerge from the simple, anodyne base. Further variations ensue if another level of rules is added to the system. Examples in CG-art include the tessellated visual constructs within Paul Brown’s Sandlines and Infinite Permutations and his other works (Whitelaw 2004, 148–153; Tofts 2005, 85–86; and see http://
L-systems are automatically branching structures, used by botanists to study plant form and physiology (Lindenmayer 1968; Prusinkiewicz and Lindenmayer 1990; Prusinkiewicz 2004). In the hands of CG-artists, they have led to McCormack’s Turbulence installation (McCormack 2004; Tofts 2005, 80ff.). This generates images of unnatural yet lifelike vegetation growing in front of one’s eyes and in response to one’s actions (thus qualifying as interactive art). Another example is the use of a “swarm grammar,” based on L-systems generating structures in (simulated) 3-D space, comparable to the decentralized yet organized constructions of social insects such as termites (Jacob and von Mammen 2007).
As for evolutionary programming, this has given rise to an important subclass of CG-art: evolutionary art, or Evo-art. Examples include Karl Sims’s Genetic Images and Galapagos (Sims 1991, 2007), plus many others (Whitelaw 2004, chap. 2). In Evo-art, the artwork is not produced by a computer program that has remained unchanged since being written by the artist. Rather, the artwork is (df.) evolved by processes of random variation and selective reproduction that affect the art-generating program itself.
Evo-art relies on programs that include self-modifying processes called genetic algorithms. To begin, a population of near-identical artworks—or to be more precise, the miniprograms that generate them—is produced by the computer. There can be any number, nine or sixteen or even more. In aesthetic terms, these first-generation artworks are boring at best and chaotic at worst. Next, each of these first-generation programs is altered (mutated) in one or more ways by the computer, at random. Usually, the alterations are very slight. Now, a selective procedure—the fitness function (decided by the artist-programmer)—is applied to choose the most promising candidate for breeding the next generation. And this process goes on repeatedly, perhaps for hundreds of generations. Provided that the mutations allowed are not too fundamental (see section 2.4), what ensues is a gradual evolutionary progress toward the type of structure favored by the artist.
Occasionally, the fitness function is fully automatic, being applied by the computer itself. If so, there may be scores, or even hundreds, of siblings in a given generation. This is a prime example of the computer’s being left to do its own thing. More usually, the selection is made by the artist or a gallery visitor, for instance, using intuitive, and often unverbalized, criteria. In such cases, the population size rarely rises above sixteen because people cannot take in more than a limited number of patterns at once. In other words, and for reasons touched on in section 2.4, there is usually no programmed fitness function. In such cases, the Evo-art also counts as interactive art, or I-art.
One might argue that the suggested definition of Evo-art is faulty on the grounds that evolutionary art need not involve a computer. It is certainly true that the very earliest G-art works of the sculptor William Latham, who later became famous as a computer artist, were drawings generated by repeated dice throwing and hand sketching. At that time, he had no idea that computers might be able to do the job for him—and do it better (Todd and Latham 1992, 2–6). But that is highly unusual: virtually all art that is produced by an iterative process of random variation plus selection is computer based. Indeed, there may be no noncomputerized examples besides early Latham. Someone might suggest Claude Monet’s water lily series, but although these showed gradual improvement by way of small changes, those changes were far from random. Even Richard Dawkins’s simple biomorphs, which were hugely seminal for Evo-artists, were computer generated (Dawkins 1986, 55–74). It is therefore acceptable to define Evo-art as a subclass of CG-art, even though this excludes the early Latham efforts.
Another subclass of CG-art is robot art, or R-art, which is (df.) the construction of robots, for artistic purposes, that are physical machines capable of autonomous movement or communication. This (happily!) is not the place for attempting to define “artistic purposes.” As for “autonomous,” the word may be understood intuitively here. At some point, however, it should be considered carefully, not least because the concept of autonomy is closely connected also with agency and creativity (see section 2.4).
This definition covers all cases of C-art wherein robots are involved. However, robots may be constructed for artistic purposes when the focus of aesthetic interest is not—as is usual—on the robots themselves but on drawings done by them (see chapter 5). A narrower definition that excludes such maverick cases would be that R-art is the construction of robots that are regarded as objects of aesthetic interest and that are physical machines capable of autonomous movement or communication.
Clearly, not all R-art is Ele-art. Indeed, R-art covers examples built many centuries ago. A few of these ancient machines could move as a whole from one place to another—such as Leonardo’s mechanical lion that “walked vp and downe” the room or Daedalus’s mercury-filled Venus that (according to Aristotle’s De Anima) had to be tethered to prevent it from running away. Most, however, could move only their body parts—like the moving statues of nymphs and satyrs in the grotto fountains at Saint Germain, which enthused Rene Descartes as a young man (Boden 2006, 2.ii.d), or Jacques de Vaucanson’s mechanical flute player (Boden 2006, 2.iv).
Electronic R-art is highly varied (Whitelaw 2004, chap. 4). It includes Ihnatowicz’s eerily hypnotic Senster, Stelarc’s thought-provoking man-robot hybrids, and Ken Goldberg’s early 1990s TeleGarden—wherein living plants are watered by a robot that is controlled by Everyman via the Internet. (For an archive of photos and videos, see http://
The TeleGarden was an early example of net art, or Net-art, in which (df.) the artwork is generated on the Internet, by multiple human interactions with the computer and indirectly with each other. So a Net-artwork is not just a computer-generated artwork that happens to be put onto the web by its human artist, not even if it can then be modified by other people. Rather, it is one for whose very existence the Internet is an essential condition. As Thor Magnusson has put it (personal communication), the immaterial property of interconnectedness is almost an artistic material, as clay is.
Artwork that was comparable, to some extent, existed forty years ago (Edmonds 1975). In 1971, Edmonds exhibited Communications Game, an electromechanical system of switches and lights controlled by six people sitting at individual stations linked by wires into three networks. And in 1972, he exhibited Rover, a rotating sphere driven, and partly lit, by three people operating joysticks. The aim was to enable an experience of communication and cooperation to arise, despite the absence of any direct communication among participants. These examples do not fit our definition of Net-art, however. Because no computer was involved, they were not even cases of C-art. Moreover, the Internet was not involved either, so the human participants were few compared with the numbers involved in Net-art today. But one might call them early prototypes, or anyway precursors. Indeed, Edmonds had also experimented with a few more participants and networks (unsuccessfully: people were overwhelmed by the larger number of signals), and he already had in mind the possibility of ARPAnet connections allowing for remote, and much more numerous, interactions.
Ten years later, Ascott conceived a work (La Plissure du Texte) that used the ARTEX computer network as the basis for the creation of a worldwide distributed narrative, or “collective global fairy tale” (Ascott 1983). The participants (or small participant groups) were each responsible for improvising the actions of a different character in the story: witch, princess, beast, wise old man, and so on. They were drawn from eleven cities around the world, and the system was continuously online for twelve days at the Musee d’Art Moderne’s ELECTRA 1983 exhibition in Paris. The versions at different locations actually vary slightly, although they should be identical. (For a Toronto version, see http://
This was much nearer to Net-art than Edmonds’s 1970s examples had been. But true Net-art—generated by the interactions between hundreds or even thousands of people—is significantly different again. The word “multiple” in the definition needs to be interpreted generously. Although Net-artworks are sometimes confined to a relatively small or closed group, the general spirit of the Net-art enterprise encourages not only extensive interconnection but also nearly unlimited openness.
The pioneering TeleGarden remains unusual because only relatively few examples of Net-art involve robotics. Most Net-art is literary, visual, or musical (Bolter 1991; Becker 1986, 1995; Ascott 2003a; Greene 2004). The rarity of robotic Net-art is hardly surprising, because a robotic installation requires constant on-the-spot monitoring if only to receive the friendly attentions of an oil can. Goldberg’s garden, for instance, was kept in operation and exhibition for nearly a decade by a team based in Austria’s Ars Electronica museum (see http://
Whether it should really be called Goldberg’s garden is of course debatable (see section 2.4). To be sure, it was his idea in the first place. But as with all Net-art (by definition), its detailed nature at any time has depended on the individual choices of many other human beings. And many, here, really does mean many. The number of participants-artists involved in the TeleGarden had already reached nine thousand by the end of its first year online and mushroomed massively thereafter.
Huge numbers apply also to some of the nonrobotic examples. The literary instances—of which Ascott’s global fairy tale was a forerunner—are multiauthor hypertexts. They include narratives composed by many hundreds of participants, offshoots of game-playing MUDs and MOOs (Montfort 2003). Their possibility was glimpsed long ago by Vannevar Bush, whose prescient “As We May Think” (1945), originally written as early as 1937, foresaw not only hypertext but search engines such as Google, too (Bolter 1991; Boden 2006, 10.i.h).
In the cases of R-art previously mentioned, only one robot is involved. Sometimes, however, groups of interacting (distributed) robots are constructed. Such groups usually employ the techniques of situated robotics, wherein the machines respond directly to specific environmental cues, including here the behavior of other robots (Boden 2006, 13.iii.b and iii.d). Occasionally, they exploit self-organizing techniques whereby the system gradually reaches an equilibrium state. Futuristic though they may seem, both these methodologies were first used by midcentury cyberneticians: Grey Walter and Ross Ashby, respectively (Boden 2006, 4.viii). One example of the latter type is Jane Prophet’s Net Work installation (Bird, d’Inverno, and Prophet 2007). One might think of this as a high-tech version of Dibbets’s oscillating sticks. But instead of eighty isolated sticks placed below the surface of the sea, Net Work consists of 2,500 floating, intercommunicating buoys, each of which is color emitting and wave sensitive. More accurately, it will consist of 2,500 such buoys: a miniature version has been tested on the Thames, but it is planned to surround the pier at Herne Bay.
Such mutually interacting robot groups do not count as interactive art in the definition of I-art given later unless they are also capable of interacting with the human audience. Net Work does have that capability: the audience can affect it by shining torchlight on the buoys or by remote control over the Internet. Other examples of interactive (and interacting) robot groups include Kenneth Rinaldo’s works in what he calls ecotechnology. His R-art (and I-art) installation called The Flock comprises three wire-and-vine robotic arms suspended from the ceiling that interact with each other and with the moving and speaking human audience. Similarly, his Autopoiesis has fifteen robot wire-frame arms distributed around the room that sense the observer’s movements and communicate with each other so as to coordinate their behavior in various ways.
This brings us to the tenth category: interactive art, or I-art. In this genre, the human audience, which need not include the artist, is not a passive observer but an active participant. Audiences are never wholly passive, of course, because art appreciation involves active psychological processes. Indeed, Duchamp (1957) went so far as to say, “The creative act is not performed by the artist alone; the spectator brings the work in contact with the external world by deciphering and interpreting its inner qualification and thus adds his contribution to the creative act.” Even for Duchamp, however, the spectator’s contribution concerns only the work’s “inner” qualification (its role, he said, is “to determine [its] weight on the aesthetic scale”). The work’s perceptible nature—or, many would say, the artwork itself—does not change as a result. In interactive art, by contrast, it does.
In I-art, then, (df.) the form or content of the artwork is significantly affected by the behavior of the audience. And in CI-art (i.e., the computer-based varieties), (df.) the form or content of some CG-artwork is significantly affected by the behavior of the audience. Again, I am speaking intuitively here: worries about just what counts as the artwork are left to the next section. The word “significantly” is needed, even though it is a hostage to interpretative fortune, so as to exclude performance art, because performance is usually subtly affected by audience reception. As for the word “behavior,” this must be interpreted with generosity. In CI-art it covers voluntary actions (such as waving, walking, touching the computer screen, or choosing a plot line within a story), largely automatic yet controllable actions (such as the direction of eye gaze), and involuntary movements (such as breathing). It even includes arcane physical factors such as the radiation of body heat.
Occasionally, the interaction involves not the audience but the physical environment: aspects of the weather, for example, or wave movements. Some installations in city squares respond to the ambient temperature and rainfall, a gentle drizzle causing changes different from those seen in a downpour; others focus on the changing patterns caused by waves on the sea. Strictly speaking, such cases fall outside CI-art as it is defined here, unless—which is usually the case—they also involve interaction with the human audience.
CI-art is generative art by definition. But it is not generative in our strictest sense as AARON is. Although the artist can go to lunch and leave the program to do its own thing, the audience cannot. However, it qualifies as CG-art in the broader sense because the artist has handed over control of the final form of the artwork to the computer, interacting with some other human being. The degree of control attributable to the audience varies: they may not realize that they are affecting the artwork or (if they do) just what behavior leads to just which changes. We see later that this variability is an important dimension in the aesthetics of CI-art.
I-art is not an entirely recent phenomenon; remember Haydn’s dice music, for instance. But it became prominent in the mid-twentieth century. This was often justified in political terms: I-art was seen as offering valuable human-human communication, in societies in which the sense of community had been diluted (Bishop 2006). It was made possible largely by cyberneticians such as Pask applying their theory of communicative feedback to art and by the new electronic technology developed in World War II.
That is not to say that all these I-art efforts were examples of Ele-art. Many artists, indeed, eschewed such technology for (countercultural) ideological reasons: it was too strongly linked with the military-industrial complex. Even Ascott’s first I-art had nary an electron in sight; it consisted of canvases with items or images on them that could be continually moved around by hand, so that the viewer of the resulting collages was their maker too (Mason 2008, 54–58). SAM and the Senster were early examples of I-art that did use electronics. But as we have seen, they did not involve computers.
Today’s I-art, however, is overwhelmingly computer based. That is because the generality of digital computers enables them, in principle, to support an infinite variety of human-computer interactions.
The types of interaction explored in CI-art are already widely diverse—hence the inclusiveness of the term “behavior” in the definition. The by-now countless examples range from interactive CD-ROMs viewed on a desktop and altered (for instance) by touching the screen (Leggett and Michael 1996), to room-sized video or virtual reality (VR) installations—such as Christa Sommerer and Laurent Mignonneau’s Trans Plant. In this case, a jungle gradually appears on the walls as the audience moves around the enclosure: grass grows when the viewer walks, and trees and bushes when the viewer stands still; the plants’ size, color, and shape depend on the size and body attitudes of the human being; and the color density changes as the person’s body moves slightly backward or forward. Trans Plant is driven by the viewer’s movements, but some CI-artworks are modified also, or instead, by the sound of human voices or footsteps. This is reminiscent of the Senster—but these computer-generated changes are much more varied and complex than those that could be engineered in the 1960s by Ihnatowicz.
Sometimes, the relevant interactions involve online access to the Internet. This is true of Net-art in general, of course, wherein the artwork itself exists only by means of the Internet. But it is also true when the artwork being shown or generated on the walls of a gallery is enhanced by the automatic incorporation of items that happen to be present on the Internet at that particular moment. One example is The Living Room, another installation built by Sommerer and Mignonneau. Unlike Trans Plant, this CI-artwork does not undergo changes that depend systematically on what the viewer is doing. Instead, random images and sounds, picked up from the Internet, appear in the room as a result of the viewer’s movements and speech.
It is usual, as in that example, for the change in the CI-artwork (whether systematic or not) to be nearly simultaneous with the observer’s triggering activity. In much of Edmonds’s recent CI-art, however (some of which was presented the 2007 Washington commemoration of the ColorField painters), the effects of the viewer’s behavior are delayed in time. In his most recent pieces, delayed changes in the software, introduced by more than one artist and also by members of the public, interact to modify the final artwork. An example, involving interactions between his Shaping Form and Sean Clark’s Transformations, is described in chapter 1.
Partly because of the lesser likelihood that the viewer will realize—and be able to control—what is going on, Edmonds speaks of “influence” rather than “interaction” in these cases. As we see in section 2.4, whether mere influence can be aesthetically satisfying is controversial even outside the precious bubble.
Certainly, mere influence, as against instantaneous interaction, would not be enough for the twelfth category, virtual reality art, or VR-art. VR-art is the most advanced version of CI-art (for examples, see Popper 2007, chaps. 4–6). Already foreseen in the mid-1960s by Ivan Sutherland (1965, 2, 506–508, 582–583), it was not technologically possible until the late 1980s (Boden 2006, 13.vi).
In VR-art, interaction leads to illusion—of an especially compelling kind. In other words, (df.) the observer is immersed in a computer-generated virtual world, experiencing it and responding to it as if it were real. One cannot pretend that this definition is clear: just what is it for someone to experience and respond “as if it were real”? Some relevant issues are covered in section 2.4. Meanwhile, let us continue to rely on an intuitive understanding of such language.
Someone might want to argue that VR-art was initiated centuries ago. Pseudorealistic mimetic worlds have been depicted in various forms of trompe l’oeil (including realistic panoramas) for many centuries and even appeared in some of the wall paintings and architecture of classical Rome. But there is a crucial difference between the relevant aesthetics in times ancient and modern. As Oliver Grau (2003, 16) has pointed out, the “moment of aesthetic pleasure” in trompe l’oeil comes when viewers consciously realize that they are not experiencing reality. In VR-art, by contrast, the enjoyment lies in feeling as though one is really inhabiting, and manipulating, an alternative world. The longer the awareness of its unreality can be delayed, the better. In other words, the experience of past forms of mimetic art was based only on illusion, not on immersion. Although one can say that the viewers were invited or deceived into responding to the art as if it were real, that “as if” was much less richly textured, much less sustained, and therefore much less persuasive than it is now.
Cinema is a halfway house (Grau 2003, chap. 4). It often elicits an emotional or narrative immersion in the filmgoer and sometimes—using special screens and techniques—leads to near-veridical experiences of inhabiting the cinematic world. These tend to exploit our reflex responses to visual images that suggest falling or imminent collision: so roller-coasters, white-water rafting, and tigers leaping toward us out of the screen are familiar favorites. But there is little psychological subtlety in this inhabitation and no detailed interaction with the alternate world, still less any physical manipulation of it.
In general, VR-art aims to make the participants (often called immersants) feel as though they are personally present in the cyberworld concerned. Normally, this world is visual or audiovisual, being presented on a visual display unit or a personal head-set, or projected onto the walls or floor of a real-world room. McCormack’s Universal Zoologies VR-artwork is an exception: its images and sounds are projected onto two large talking heads in an attempt to provide a realistic illusion of human conversation (Tofts 2005, 81–82). But sometimes VR-art leads also to convincing experiences of touch, pressure, and motion by providing the observer with special gloves and other equipment (Boden 2006, 13.vi). Sometimes, too, the observer experiences utterly unreal capacities, such as being able to fly or to activate highly unnatural causal chains within the virtual world.
Even when the viewer is not presented with such shockingly unfamiliar experiences as those, something about the virtual world will be perceptibly unlike the real world. And this is deliberate. VR-artists are not aiming to achieve a fully detailed mimesis: what would be the point of that? Rather, they use near mimesis to cast some aesthetically or conceptually interesting light on our usual experiences and assumptions.
Detailed mimesis may be appropriate for other purposes, of course. For instance, a VR brain used in training neurosurgeons provides nicely realistic sensations of touch and vision when the trainee’s virtualized surgical tool prods, pinches, or cuts different parts of it (Wang et al. 2007). Given that brains are very soft (compared with hearts, for example), the visual or haptic information at issue here is highly complex. So each of the different types of “touching” (prodding, pinching, cutting) has its own distinctive material and perceptual effects. Such highly realistic effects would be appropriate in an artistic VR work only if they prompted thoughts about matters beyond the practical exigencies of brain surgery—the physical vulnerability of all human flesh, perhaps, and (by extension) of human plans.
The thirteenth category, LC-art, covers cases of live coding, wherein (df.) the art object is generated on the fly by code that is written in real time by the artist.
Most LC-art is music, but occasionally images or texts are generated instead. The reason is that producing an interesting or attractive artwork is only part of the LC-artist’s goal. And appreciating the art object for its own sake is only part of what the LC-audience is intended to do. In addition, they are supposed to read the code (which is displayed in real time, while it is being written) and appreciate how it informs the resulting artwork. Why did that instruction lead to that observable feature? And what sort of music is likely to result from that new piece of code? It is even more difficult to do this when both code and result are presented visually than when one is visual and the other auditory. So music is the dominant medium in LC-art. Indeed, very often live coders do not place that much emphasis on the exact manner and detail of the visual presentation of the code, hence indicating indirectly that the music is primary.
Even in the case of music, however, only a highly computer-literate audience can achieve the desired experience. It follows that the problems of art appreciation discussed in chapter 4 are especially great for LC-art. Moreover, it takes a special sort of coding to enable the audience to do so. Some LC-artists use existing programming languages, which may already be familiar to the audience. But others design novel programming languages, so that the successively appearing lines of code can be immediately seen to imply particular musical results.
LC-art is a type of CG-art because it involves programming by the artist and the generation of the LC-artwork by the computer. But it is similar to CA-art in one way: the human artist has direct control over what the program is doing at every moment. It therefore raises fewer philosophical questions than other categories of CG-art as CA-art in general does too (see section 2.4).
The final category is 3DP-art, which involves (df.) the construction of artworks by means of 3-D printing. Using this new technology, objects are built by laying down successive levels of some hard-setting fluid. This computer-controlled process is being used to produce toys, jewelry, artificial organs (e.g., ears), and even guns. But it is also beginning to be used by artists (as yet, only very few).
In principle, 3DP-art could be generative in the sense required for CG-art. For instance, it is suggested in chapter 6 that some future version of a program like The Painting Fool might employ 3-D printing to build brushstrokes in the chosen style, having appropriate 3-D texture as well as the requisite colors.
A team of artists and engineers in the Netherlands is already copying a Rembrandt self-portrait and his painting The Jewish Bride, as well as Van Gogh’s Flowers in Blue Vase, all in three dimensions, not just two. However, that project counts as CA-art rather than CG-art: the desired results were initiated by Rembrandt and Van Gogh, and the team is reproducing only those. Similarly, many 3DP-artists who aim to produce their own, novel artworks are using the technology in much the same spirit as Hockney uses Photoshop. In theory, they could make the works without the technology, but in practice the facility that it offers makes the construction of the sculptures realistic. See, for example, the exhibition Beyond the Buzz, shown at the Minneapolis College of Art and Design in 2015 (http://
Achieving truly generative 3DP-art involves added challenges. These challenges do not involve concerns for computation, as such, but concentrate on relationships between the virtual and the physical and the ways that different physical objects can relate to one another through their shared virtual digital models. One of the few examples of attempts to meet such challenges can be found in a 2010 exhibition displayed in Auckland, New Zealand (Hybrids at the Media and Interdisciplinary Arts Centre; http://
In sum, the fourteen definitions in this taxonomy are as follows:
- Ele-art involves electrical engineering or electronic technology.
- C-art uses computers as part of the art-making process.
- D-art uses digital electronic technology of some sort.
- CA-art uses the computer as an aid (in principle, nonessential) in the art-making process.
- G-artworks are generated, at least in part, by a process not under the artist’s direct control.
- CG-art is produced by leaving a computer program to run by itself, with minimal or zero interference from a human being. The stricter definition of CG-art (art produced by a program left to run by itself, with zero interference from the human artist) was deliberately rejected.
- Evo-art is evolved by processes of random variation and selective reproduction that affect the art-generating program itself.
- R-art is the construction of robots for artistic purposes, in which robots are physical machines capable of autonomous movement or communication. C-art that uses robots but has an artistic focus not on the robots themselves is excluded.
- Net-art is the generation of the artwork on the Internet, by multiple human interactions with the computer and indirectly with each other.
- In I-art, the form or content of the artwork is significantly affected by the behavior of the audience or sometimes by purely physical causes.
- In CI-art, the form or content of some CG-artwork is significantly affected by the behavior of the audience.
- In VR-art, the observer is immersed in a computer-generated virtual world, experiencing it and responding to it as if it were real.
- LC-art is generated on the fly by code that is written in real time by the artist.
- 3DP-art is the construction of artworks by means of 3-D printing.
2.4 Questions for Philosophical Aesthetics
Aesthetic and philosophical problems arise with respect to CG-art in general and others with respect only to particular varieties of it. None of these can be explored at length here. (For fuller discussions, see the other chapters in this volume and also Boden 1999; 2004; 2006, 13.iii.d–e and 16.viii.c; 2007a; 2007b; 2010a; 2010b; Cornock and Edmonds 1973; Costello and Edmonds 2007; Edmonds 2006, 2007; Muller, Edmonds, and Connell 2006.) Instead, here we briefly discuss the wide range of puzzles that attend the consideration of generative art.
One obvious question can be put like this: Is it really the case that a computer can ever do its own thing? Or is it always doing the programmer’s (artist’s) thing, however indirectly?
To answer that question seriously requires both general philosophical argument and attention to specific aspects of particular CG-art examples in the underlying program and in the observable artwork. That sort of attention is not appropriate in cases of G-art that are not computer based. The physical, psychological, or biological processes in which they are grounded are not specifiable in detail—not even by scientists, let alone by artists. Computer programs, in contrast, are so specifiable.
That is why one can make sensible comparisons between the extent to which different CG-art programs are or are not under the artist’s direct control and the extent to which, and the points at which, they are subject to interference from a human being. However, one can do this only if one knows something about how the program or installation works (see chapter 4). Merely observing, or even participating in, the resultant artwork is not enough.
Whether it appears to participants that the program or installation is independent, or autonomous, is of course another question, one that may not be easy to answer in practice.
Even the programmer may be misled, here. In the earlier paper on which this one is based (Boden and Edmonds 2009), Edmonds pointed out that step-by-step programming (i.e., writing a program as a sequence of explicit instructions) feels more directive than rule-based programming (i.e., defining a set of constraints—for instance, that Z should always be bigger than Y or X must never equal W—that the computer must follow, without stating how this will be effected by the machine). In the latter case, one might say that the artist leaves the computer to do its own thing without knowing just what it is that the computer will be doing. The computer-art community usually regards it as important that the artwork is generated from a set of rules, or constraints, rather than from a step-by-step algorithm. But this is more a matter of taste than anything else. Even when programmers have written explicit step-by-step code, they do not necessarily—or even usually—know the outcome. If they did, there would be no bugs (except those due to typing mistakes and punctuation errors). Despite the difference between the feel of the two programming approaches, there is no distinction at the most fundamental level: in all but the very simplest cases, both types of program are unpredictable by their programmer. Rule-driven systems merely appear to have a greater degree of autonomy relative to the conscious decisions of the human artist.
Autonomy, of course, is a concept that is closely connected with art making in general. John Ruskin, for example, made much of it in his theory of aesthetics (see chapter 7). But does it ever make sense to ascribe autonomy to a computer?
Some sorts of programs can indeed have autonomy and of several different types (Boden 2007b). But only one of these is similar to the sort of human autonomy that we normally associate with art. Is that ever found in CG-art?
It is certainly-art, in which the human artist’s moment-to-moment coding is responsible for the artwork that results. But what about the other categories of CG-art? Irrespective of the artist’s phenomenology while writing the program or of the participants’ phenomenology when experiencing it, do some of these have more autonomy than others? What of Evo-art, for instance: does the self-modification involved and the automatic selection (in cases in which that happens) mean that evolutionary programs are more autonomous than, say, AARON? With respect to AARON, can we ascribe at least a minimal level of autonomy to the computer, given that Cohen has no hands-on control over what picture will be drawn or how?
Insofar as a program is doing its own thing, does it take on the authorial responsibility? Let us ignore the fact that authorial responsibility is often unclear here anyway, because most CG-art is produced by a team, not a lone artist.
For instance, did AARON generate those magnificent “world-class” colored drawings or did Cohen? He admits, after all, that he himself “wouldn’t have had the courage to use those colours.” On the other hand, he says he is happy that there will be more of “his” original artworks appearing well after his death (Cohen 2002a). Is he right, or is he deluded? The answer will depend not only on one’s philosophy of the self but also on one’s views as to whether any computer program can be seen as an author or artist. Douglas Hofstadter, for example, interprets the self in such a way that he would be content to ascribe the posthumous works to Cohen himself, and he would even deny that they are in the fullest sense posthumous (Hofstadter 2007).
However, if he was emotionally moved by them, he would also resist ascribing authorship to the computer (Hofstadter 2001). In the end Harold Cohen comes down on his, rather than the computer’s, side when it comes to taking the credit for the creativity: “I think of autonomy now in terms of that weakly defined but strongly felt future state that manifests itself in the criteria that direct creative behavior. And if I take some satisfaction in being able to formulate it in those terms, I also recognize that the criteria and the creativity in question are mine, not the program’s” (Cohen 2002b).
Does Evo-art leave more room, or less, for human authorship than AARON does? That is, does the artist’s choice of fitness function suffice to give him or her the authorial credit for whatever artwork emerges after many generations of random (i.e., undirected) change? Is the credit greater, or less, if instead of relying on a programmed fitness function, the artist does the selecting by hand?
One reason for the Evo-artist’s choosing to do the selection by hand is to produce only works in his or her own style. This is also the reason the mutations that are allowed are usually very slight. An artistic style is a sustained pattern of activity, lasting over time (Boden 2010c). In Evo-art that allows radical mutations (and that does not ration them to once every two thousandth generation, for instance), no pattern can be sustained for long—not even if the human artist is trying to shape the results by making the fitness selection at each stage. On the contrary, huge structural changes can occur in a single generation (see Sims 1991). This leads to fascinated amazement on the part of the gallery audience. Nevertheless, Evo artists do not normally allow such mutations. They prefer to explore a stylistic space that, despite many surprising variations, remains recognizable as theirs to someone with an experienced eye. In other words, they are primarily engaged in exploratory creativity, not transformational creativity (Boden 2004).
There are some exceptions. The CG-artist Paul Brown has embarked upon an Evo-art project whose aim is to evolve robots that will make aesthetically acceptable drawings that do not carry Brown’s general style or personal signature (Bird et al. 2006). Thus far, his hope has not been satisfied. And perhaps that is not surprising; Brown, after all, was setting the fitness functions at all stages of the work. It is not clear whether it is in principle possible for his artistic mark to be lost as this project proceeds or whether the robots might be able to develop a personal signature of their own (Boden 2010c).
This example raises questions also about the relation between CG-art and embodiment. Many philosophers of mind discount AI and A-Life in general (as models of mind or life) for being concerned with virtual, bodiless, systems. However, Brown’s R- and Evo-art creatures are robots moving in the real world and are therefore subject to physical forces. It is known that truly fundamental changes—that is, new types of sensory receptors—can evolve in robots as a result of unsuspected physical contingencies (Bird and Layzell 2002). Compare the biological evolution, not of a primitive eye into a better eye, but of a light sensor in which no such sensor existed before. In principle, then, a fundamentally new style might develop in this way, whereas (arguably) that could not happen in a purely virtual, programmed system. (This question is discussed further in chapter 5.)
Similar puzzles about authorial responsibility arise in CI-art in general, of which hand-selected Evo-art is a special case. Just where, in the man-machine system concerned, is the true author?
That worry affects all I-art, of course, but is there any extra difficulty where CI-art is concerned? For present purposes, let us ignore Duchamp’s suggestion that all art is multiauthored. And what difference, if any, does it make if—as sometimes happens (see chapter 11)—the audience provides feedback during the construction of the CI-work, so that its final form depends not only on the decisions of the artist but also on the reactions of the audiences who encountered it in its prototype stage? Perhaps the distinction between decisions and reactions is crucial here, debarring the audience from earning any extra authorial credit in such cases?
To speak of a worry here, however, is perhaps to counteract what CI-artists are trying to do. Despite its sturdy roots in cybernetics and computer technology, CI-art has attracted favorable notice from postmodernists precisely because of the ambiguity of authorship involved. The pioneering Ascott (2003a, 2003b), in particular, has always seen the value of CI-art as its democratizing ability to engage the viewer or participant as creator. In his words, “creativity is shared, authorship is distributed” (1990, 238). If authorship is deliberately distributed, then to worry about its locus (about ascribing the status of author) is to miss the point.
For all the heady talk of creative participation, some CI-art is fairly limiting (Kahn 1996). That is so, for instance, when the possible changes in the artwork are explicitly preset by the artist, as opposed to their emerging from the program’s doing its own thing. The limitation is especially great when the participant chooses from a menu of selections.
Another way of putting questions about authorial responsibility is to ask where the creativity lies. But what, exactly, do we mean by creativity?
It certainly involves agency—which is why considerations of autonomy and authorial responsibility are inevitable. But what is agency? The interacting arms and floating buoys identified as examples of R-art are typically described by the artists and technicians concerned as agents—a word borrowed from AI and A-Life research on distributed cognition. But does that research misuse the concept? Even if it does, does it include agents of interestingly different types (Boden 2006, 13.iii.d–e), some of which are more deserving of the name than others? If so, should we at least reserve the term—and the ascription of creativity—for those cases of CG-art in which the agents involved are of the more plausible variety? Again, such questions cannot be answered without careful attention to the details of the programs and communications involved.
It is commonly assumed that creativity—and art, too—involves unpredictability. But what is its source? Is it merely lack of computational power on the part of human minds? We have seen that CG-art, like complex programs in general, is indeed unpredictable for that reason. But CI-art and Evo-art are unpredictable for other reasons as well: CI-art, because the artist cannot predict the sequence of interactions that will take place, even if able to predict what would happen at a given moment if that audience movement were to occur; and Evo-art, because of the many random changes to the program and because of the choices made at successive generations by the artist. Does the unpredictability of traditional art have any deeper source? And if so, is this something that cannot be ascribed to, or even simulated in, computers? Answering these questions requires one to distinguish different types of unpredictability very carefully (see Boden 2004, chap. 9).
Another set of questions concerns ontology, or what counts as the artwork. How can we identify the artwork when an artist’s computer program generates countless unique images or musical compositions, none of which have been seen or heard by the artist? Is each image or musical composition produced by AARON or Emmy an artwork or is the artwork the program that generates it? In Evo-art, does one and the same artwork exist at differing levels of sophistication at different generations? Or does every generation produce a new artwork or, perhaps, a new population of (sibling) artworks?
What counts as the artwork when the uniqueness is due not only to a richly generative computer program but also to the contingent (and ephemeral) behavior of a participatory human audience? Perhaps the familiar concept of artwork is well suited only to the unchanging artifacts that form the overwhelming majority of the cases inside McCormack’s bubble?
Traditional artists can fully comprehend the painting or sculpture that they have executed so carefully (although whether this applies to the G-art dimension of Pollock’s paintings is questionable), but CI-artists cannot fully know the CI-artwork that they constructed with equal care. This is not merely a matter of the unpredictability of detail: in sufficiently complex cases, it is not even clear that they can recognize the general potential of their own work. With regard to CI-art, then, perhaps we should speak not of the artwork but of the art system, which comprises the artist, the program, the technological installation (and its observable results), and the behavior of the human audience? And perhaps, if the concept of the artwork falls, then that of the artist or author falls too?
Or maybe we should think of each occurrence of CI-art as a performance and the program or installation as the score? If so, then philosophical discussion of what counts as an artwork in music is pertinent (e.g., Goodman 1968). But the performance is more like a jazz improvisation than the playing of classical music, because it can vary considerably from one occasion to another. Even if the form of each particular human-computer interaction can be completely determined by the artist (which is not so, for instance, when the computer’s response can be modified by its memory of the history of previous interactions), the sequence of such events cannot.
Yet another problematic area concerns aesthetic evaluation. Are entirely novel aesthetic considerations relevant for CG-art in general or for some subclass of it? And are some aesthetic criteria, normally regarded as essential, utterly out of place in CG-art—authenticity, for instance?
The devotees of CI-art, in fact, do not use the familiar (inside-the-bubble) criteria to judge different interactive installations. Or insofar as they do, these are secondary to other considerations (Boden 2010a). The criteria they see as most appropriate concern not the nature of the resulting artwork (the beauty of the image projected on the wall, for example, or the melody and harmoniousness of the accompanying sounds), but the nature of the interaction itself. There is general agreement on that point. But there is significant disagreement on just what type of interaction is the most aesthetically valuable.
Some CI-artists, especially those engaged in VR-art, stress the disturbing sense of unreality involved and the participant’s new take on everyday experience that ensues. Many value the participant’s conscious control of the artwork, others aim to highlight their sense of personal embodiment, and yet others stress the audience’s disconcerting experience of unpredictability triggered by their own actions. All those criteria concern the participant’s experience, but difficulties arise if one asks how that experience can be discerned, or logged, by anyone other than the individual participant. As remarked in section 2.3, if the observers can never come to realize that they are affecting what happens, then the “I” in “CI-art” might better be thought of as “influence,” not “interaction.”
Some especially jagged philosophical rocks lie in wait for VR-artists. The concept of virtual reality has been defined in various ways (Steuer 1992). Most, like the definition of VR-art given in section 2.3, refer to the participant’s experience of being immersed in a real world and reacting accordingly. This notion seems to be intuitively intelligible, especially if one has actually encountered a VR installation. But just what it means, whether in psychological or philosophical terms, is very difficult to say.
It is not even clear that it is coherent (Boden 2006, 16.viii.c). Several leading philosophers of mind have addressed this hornet’s nest of questions in writing about the film The Matrix (see especially the papers by Hubert Dreyfus and Andy Clark on the Warner Brothers website, http://
As for authenticity, this is a tricky concept. For several reasons, of varying plausibility, someone might argue that it is not applicable to any instance of CG-art (Boden 2007a). And CG-artists have suffered as a result.
For example, Cope (2006, 364) has been continually disappointed by people’s failure to take his music seriously—not because they dislike it on hearing it (sometimes they refuse to hear it) but simply because it is computer generated. Even when they do praise it, he has found that they typically see it less as music than as computer output—a classification that compromises its authenticity. For instance, even though each Emmy composition is in fact unique, people know that the program could spew out indefinitely many more tomorrow. That human composers die, says Cope, has consequences for aesthetic valuation: someone’s oeuvre is valued in part because it is a unique set of works, now closed. As a result of this common reaction, Cope decided to destroy the database of dead composers’ music that he had built up over the last twenty-five years and used as a crucial source in Emmy’s functioning. Emmy’s successor will compose music only in Cope’s own style; whether audiences regard this as being significantly more authentic still remains to be seen.
Finally, what of the claims made by many CG-artists to be exploring the nature of life? It is clear from Rinaldo’s choice of the titles Autopoiesis and The Flock (plus the rest of his oeuvre; Whitelaw 2004, 109–116), for instance, that his R-art works are not intended as mere fairground toys but as meditations on significant aspects of life. He is not alone in this: many of the CG-artists who have been influenced by A-Life see their work in that way. Concepts such as emergence and self-organization, and of course evolution, crop up repeatedly in their writings and interviews—as does the key concept of life itself.
One may well agree that their work throws light on, or anyway reminds us of, important—and puzzling—properties of life. But one need not also agree with the claim sometimes made by these CG-artists that purely virtual life (or strong A-Life) is possible and that their work, or something similar, might even create it. Indeed, one should not accept those claims, because physical metabolism (which is much more than mere energy dependency) is essential for life but is denied to computers (Boden 1999).
Perhaps the most obvious philosophical question of all is this: Yes, it can help while away a Sunday afternoon, but is it art, really?
Whatever art may be (a topic on which gallons of ink have been spilled), many people feel that computers are the very antithesis of it. Indeed, some philosophers argue this position explicitly (e.g., O’Hear 1995). On their view, art involves the expression and communication of human experience, so that if we did decide that it is the computer that is generating the artwork, then it cannot be an art work after all—no matter how decorative or even beautiful it may be.
A closely related worry (addressed in chapter 6) concerns emotion in particular: if computers are not emotional creatures, then—on this view—they cannot generate anything that is properly termed art (Hofstadter 2001). Another common way of discrediting computer art in general is to argue that art involves creativity and that no computer—irrespective of its observable performance—can really be creative (for a discussion, see Boden 2004, chap. 11; 2014). Furthermore, a person’s aesthetic approval of an artwork is sometimes instantly renounced on discovering that it is, in fact, a CG-artwork, a reaction that led Cope to destroy the database on which Emmy’s—or should one rather say “his”?—CG-music rested.
Despite these (fairly common) views on the relation—or lack of it—between art and computers, there are undeniable continuities between CG-art and noncomputer art. Several of these were mentioned in section 2.3, which showed that some of the fourteen taxonomic categories include examples drawn from both inside and outside McCormack’s precious bubble. And those categories that apply only to CG-art cover many individual cases that are aesthetically related to traditional artworks. Even key aspects of Gothic art, seemingly a million miles away from CG-artworks, can be shared by these electronic instances (see chapter 7).
Moreover, the art world itself—however suspicious it may be of computers in general and however dismissive it may be of particular CG-art efforts—does sometimes award these new activities the coveted status of art. Sometimes this happens in a specialized corner of the art world: for instance, London’s Kinetica Gallery (opened in 2006) is devoted to interactive, robotic, and kinetic art. But some major traditional galleries clearly accept that traditional and CG-art are players in the same cultural ballpark.
Two such instances have been mentioned in this chapter: the Tate’s one-man show of Cohen’s AARON and the Washington exhibition featuring Edmonds’s work as a development of that of the ColorField painters (Mark Rothko and the like). The latter example is especially telling, precisely because it was not a show celebrating only CG-art. On the contrary, this sixtieth-anniversary exhibition was putting CG-art alongside the precious bubble—or even inside it. In reply to the skeptical challenge But is it art, really? what more persuasive answer could there be?
______________
This chapter is based on M. A. Boden and E. A. Edmonds, “What Is Generative Art?,” Digital Creativity 20 (1–2) (2009): 21–46. The research for that earlier paper was supported by AHRC Grant no. B/RG/AN8285/APN19307, Computational Intelligence, Creativity, and Cognition: A Multidisciplinary Investigation and also by the Australasian CRC for Interaction Design, established under the Australian Government’s Research Centres Programme.
References
Ascott, R. (1966/1967). “Behaviourist Art and the Cybernetic Vision.” Cybernetica 9:247–264; and 10 (1967): 25–56. Reprinted as chap. 3 in Ascott 2003b, 109–156.
Ascott, R. 1983. La Plissure du Texte. http://
Ascott, R. 1990. “Is There Love in the Telematic Embrace?” Art Journal (College Arts Association of America) 49:241–247. Reprinted as chap. 15 of Ascott 2003b.
Ascott, R. 2003a. “Technoetic Aesthetics: 100 Terms and Definitions for the Post-biological Era.” In Telematic Embrace: Visionary Theories of Art, Technology, and Consciousness, 375–382. London: University of California Press. First published in Japanese, trans. E. Fujihara, in R. Ascott, Art and Telematics: Toward the Construction of a New Aesthetics (Tokyo: NTT Publishing Co., 1998).
Ascott, R. 2003b. Telematic Embrace: Visionary Theories of Art, Technology, and Consciousness. London: University of California Press.
Becker, H. S. 1986. Doing Things Together: Selected Papers. Evanston, IL: Northwestern University Press.
Becker, H. S. 1995. “A New Art Form: Hypertext Fiction.” In How to Read a Hypertext, edited by M. Bernstein. Cambridge, MA: Eastgate Systems.
Bird, J., M. d’Inverno, and J. Prophet. 2007. “Net Work: An Interactive Artwork Designed Using an Interdisciplinary Performative Approach.” Digital Creativity 18 (1): 11–23.
Bird, J., and P. Layzell. 2002. “The Evolved Radio and Its Implications for Modelling the Evolution of Novel Sensors.” Proceedings of Congress on Evolutionary Computation, CEC-2002, 1836–1841.
Bird, J., D. Stokes, P. Husbands, P. Brown, and B. Bigge. 2006. “Towards Autonomous Artworks.” Leonardo Electronic Almanac. Available from jonba@sussex.ac.uk.
Bishop, C. 2006. Participation. Cambridge, MA: MIT Press.
Boden, M. A. 1994. “Agents and Creativity.” In “Agents,” edited by D. Riecken, 117–121. Special issue, Communications of the Association for Computing Machinery. Reprinted in Boden 2010b, chap. 8.
Boden, M. A. 1999. “Is Metabolism Necessary?” British Journal for the Philosophy of Science 50:231–248. Reprinted in Boden 2010b, chap. 12.
Boden, M. A. 2004. The Creative Mind: Myths and Mechanisms. 2nd ed. London: Routledge.
Boden, M. A. 2006. Mind as Machine: A History of Cognitive Science. 2 vols. Oxford: Clarendon/Oxford University Press.
Boden, M. A. 2007a. “Authenticity and Computer Art.” In “Creativity, Computation, and Cognition,” edited by P. Brown. Special issue, Digital Creativity 18 (1): 3–10. Reprinted in Boden 2010b, 193–209.
Boden, M. A. 2007b. “Stillness as Autonomy.” In Proceedings of Computers in Art and Design Education (CADE) Conference, Stillness (Perth), September 2007, edited by S. Worden, L. Green, and P. Thomas, 175–192. An expanded version is in Boden 2010b.
Boden, M. A. 2010a. “Aesthetics and Interactive Art.” In Creativity and Art: Three Roads to Surprise, 210–234. Oxford: Oxford University Press.
Boden, M. A. 2010b. Creativity and Art: Three Roads to Surprise. Oxford: Oxford University Press.
Boden, M. A. 2010c. “Personal Signatures in Art.” In Creativity and Art: Three Roads to Surprise, 92–124. Oxford: Oxford University Press.
Boden, M. A. 2014. “Creativity and Artificial Intelligence: A Contradiction in Terms?” In The Philosophy of Creativity: New Essays, edited by E. S. Paul and S. B. Kaufman, 224–246. Oxford: Oxford University Press.
Boden, M. A., and E. A. Edmonds. 2009. “What Is Generative Art?” Digital Creativity 20 (1–2): 21–46.
Bolter, J. D. 1991. Writing Space: The Computer, Hypertext, and the History of Writing. Hillsdale, NJ: Lawrence Erlbaum.
Breton, A. 1969. Manifestoes of Surrealism. Translated by R. Seaver and H. R. Lane. Ann Arbor: University of Michigan Press.
Brown, Paul. 2003. “Generative Computation and the Arts.” Special issue, Digital Creativity 14 (1): 1–2.
Brown, Richard. 2001. Biotica: Art, Emergence, and Artificial Life. London: RCA CRD Research.
Brown, Richard. 2006. “The Electrochemical Glass.” Strange Attractor Journal Three. https://
Burnham, J. 1968. Beyond Modern Sculpture: The Effects of Science and Technology on the Sculpture of This Century. London: Allen Lane.
Bush, V. 1945. “As We May Think.” Atlantic Monthly 176 (July): 101–108. Reprinted in Multimedia: From Wagner to Virtual Reality, edited by R. Packer and K. Jordan, 135–153. London: W. W. Norton, 2001.
Candy, L., and E. A. Edmonds, eds. 2002. Explorations in Art and Technology. New York: Springer.
Cohen, H. 1995. “The Further Exploits of AARON Painter.” In “Constructions of the Mind: Artificial Intelligence and the Humanities,” edited by S. Franchi and G. Guzeldere. Special issue, Stanford Humanities Review 4 (2): 141–160.
Cohen, H. 2002a. “A Million Millennial Medicis.” In Explorations in Art and Technology, edited by L. Candy and E. A. Edmonds, 91–104. New York: Springer.
Cohen, H. 2002b. “A Self-Defining Game for One Player: On the Nature of Creativity and the Possibility of Creative Computer Programs.” Leonardo 35 (1): 59–64.
Cohen, H. 2007. AARON’s World: 20 de Junho–28 de Julho 2007. Lisbon: Antonio Prates, Arte Contemporanea.
Colton, S., M. Valstar, and M. Pantic. 2008. “Emotionally Aware Automated Portrait Painting.” In Proceedings of the 3rd International Conference on Digital Interactive Media in Entertainment and Arts (DIMEA), edited by Sofia Tsekeridou, 304–311. New York: ACM Press.
Cope, D. 2001. Virtual Music: Computer Synthesis of Musical Style. Cambridge, MA: MIT Press.
Cope, D. 2006. Computer Models of Musical Creativity. Cambridge, MA: MIT Press.
Cornock, S., and E. A. Edmonds. 1973. “The Creative Process Where the Artist Is Amplified or Superseded by the Computer.” Leonardo 6:11–16.
Costello, B., and E. A. Edmonds. 2007. Playful Pleasures: A User Evaluation of Three Artworks. Sydney, Australia: University of Technology, CCS Report 2007-1.
Dawkins, R. 1986. The Blind Watchmaker. Harlow: Longman.
Dorin, A., and J. McCormack. 2001. “First Iteration/Generative Systems.” Leonardo 34 (3): 335.
Duchamp, M. 1957. “The Creative Act.” In The Essential Writings of Marcel Duchamp, edited by M. Sanouillet and E. Peterson, 138–140. London: Thames and Hudson.
Duprat, H., and C. Besson. 1998. “The Wonderful Caddis Worm: Sculptural Work in Collaboration with Trichoptera.” Leonardo 31 (3): 173–182.
Edmonds, E. A. 1969. “Independence of Rose’s Axioms for M-Valued Implication.” Journal of Symbolic Logic 34:283–284.
Edmonds, E. A. 1975. “Art Systems for Interactions between Members of a Small Group of People.” Leonardo 8:225–227.
Edmonds, E. A. 2006. “New Directions in Interactive Art Collaboration.” CoDesign: International Journal of CoCreation in Design and the Arts 2 (4): 19–194.
Edmonds, E. A. 2007. “Reflections on the Nature of Interaction.” CoDesign: International Journal of CoCreation in Design and the Arts 3 (3): 139–145.
Eno, B. 1996. “Generative Music: Evolving Metaphors, in My Opinion, Is What Artists Do.” http://
Goodman, N. 1968. Languages of Art: An Approach to a Theory of Symbols. Indianapolis, IN: Bobbs-Merrill.
Greene, R. 2004. Internet Art. London: Thames and Hudson.
Gwilt, Ian. 2013. Data-Objects: Sharing the Attributes and Properties of Digital and Material Culture to Creatively Interpret Complex Information. In The Re-materialisation of the Art Object, edited by D. Harrion, 14–26. Hershey, PA: IGI Global.
Haugeland, J. 1985. Artificial Intelligence: The Very Idea. Cambridge, MA: MIT Press.
Hofstadter, D. R. 2001. “Staring Emmy Straight in the Eye—and Doing My Best Not to Flinch.” In Cope 2001, 33–82.
Hofstadter, D. R. 2007. I Am a Strange Loop. New York: Basic Books.
Jacob, C., and von Mammen, S. 2007. “Swarm Grammars: Growing Dynamic Structures in 3D Agent Spaces.” Digital Creativity 18:15–64.
Johnson, B. S. 1969. The Unfortunates. London: Panther.
Kahn, D. 1996. “What Now the Promise?” In Burning the Interface, edited by M. Leggett and L. Michael, 21–30. Sydney: Museum of Contemporary Art. CD-ROM.
Krueger, M. W. 1991. Artificial Reality. 2nd ed. Reading, MA: Addison-Wesley.
Laposky, B. 1969. “Oscillons: Electronic Abstractions.” Leonardo 2 (4): 345–354.
Leggett, M. 2000. “Thinking Imaging Software.” Photofile: Photomedia Journal 60:28–31.
Leggett, M., and L. Michael, eds. 1996. Burning the Interface. Sydney: Museum of Contemporary Art. CD-ROM.
LeWitt, S. 1967. “Paragraphs on Conceptual Art.” Artforum 5 (10): 79–83. Reprinted in Theories and Documents of Contemporary Art: A Sourcebook of Artists’ Writings, edited by K. Stiles and P. Selz, 822–826. London: University of California Press, 1996.
LeWitt, S. 1969. “Sentences on Conceptual Art.” Art-Language 1:11–13. Reprinted in Lippard 1997, 75–76; and in Stiles and Selz, 826–827.
Lindenmayer, A. 1968. “Mathematical Models for Cellular Interaction in Development, Parts I and II.” Journal of Theoretical Biology 18:280–315.
Lippard, L. R. 1997. Six Years: The Dematerialization of the Art Object from 1966 to 1972. London: University of California Press.
Mallen, G. 2005. “Reflections on Gordon Pask’s Adaptive Teaching Concepts and Their Relevance to Modern Knowledge Systems.” In Creativity and Cognition 2005: Proceedings of the Fifth Conference on Creativity and Cognition, edited by L. Candy, 86–91. New York: ACM Press.
Marr, A. 2003. “ ‘He Thought It the Deuil, Whereas Indeede It Was a Meere Mathematicall Inuention’: Understanding Automata in the Late Renaissance.” Paper presented at the Workshop on the History of AI/A-Life, Stanford University, CA, October. Also published as A. Marr, “Understanding Automata in the Late Renaissance.” Journal de la Renaissance 2 (2004): 205–222.
Martin, K. 1951. “Abstract Art.” In Broadsheet No. 1: Abstract Paintings, Sculptures, Mobiles. London: Lund Humphries. AIA exhibition catalog. Also in Nine Abstract Artists, Their Work and Their Theory: Robert Adams, Terry Frost, Adrian Heath, Anthony Hill, Roger Hilton, Kenneth Martin, Mary Martin, Victor Pasmore, William Scott, edited by L. Alloway. London: Tiranti, 1954.
Mason, C. 2008. A Computer in the Art Room: The Origins of British Computer Arts 1950–80. Hindringham: JJG Publishing.
McCorduck, P. 1991. AARON’s Code: Meta-art, Artificial Intelligence, and the Work of Harold Cohen. New York: W. H. Freeman.
McCormack, J. 2003. “Art and the Mirror of Nature.” In “Generative Computation and the Arts,” edited by Paul Brown. Special issue, Digital Creativity 14 (1): 3–22.
McCormack, J. 2004. Impossible Nature: The Art of John McCormack. Melbourne: Australian Centre for the Moving Image.
McCormack, J., A. Dorin, and T. Innocent. 2004. “Generative Design: A Paradigm for Design Research.” In Proceedings of Futureground, edited by J. Redmond, D. Durling, and A. de Bono, 156–164. Melbourne: Design Research Society.
Metzger, G. 1959. “Manifesto Auto-destructive Art.” In A Computer in the Art Room: The Origins of British Computer Arts 1950–80, edited by C. Mason, 59. Hindringham: JJG Publishing.
Metzger, G. 1965. Auto-destructive Art. London: Destruction/Creation.
Mohr, Manfred. 1986. “Divisibility III, Generative Works, Part III, Works from 1985–1986.” Galerie D+C Mueller-Roth, Stuttgart, Germany, December 6 1986–January 24, 1987.
Montfort, N. 2003. Twisty Little Passages: An Approach to Interactive Fiction. Cambridge, MA: MIT Press.
Muller, L., E. A. Edmonds, and M. Connell. 2006. “Living Laboratories for Interactive Art.” CoDesign: International Journal for CoCreation in Design and the Arts 2 (4): 195–207.
Nake, F. 1998. “Art in the Time of the Artificial.” Leonardo 31 (3): 163–164.
Nake, F. 2005. “Computer Art: A Personal Recollection.” In Creativity and Cognition 2005: Proceedings of the Fifth Conference on Creativity and Cognition, edited by L. Candy, 54–62. New York: ACM Press.
Nees, G. 1969. Generative Computergraphik. Berlin: Siemens AG.
O’Hear, A. 1995. “Art and Technology: An Old Tension.” In Philosophy and Technology, edited by R. Fellows, 143–158. Cambridge: Cambridge University Press.
Pask, G. 1971. “A Comment, a Case History, and a Plan.” In Cybernetics, Art, and Ideas, edited by J. Reichardt, 76–99. London: Studio Vista.
Pinch, T., and F. Trocco. 2002. Analog Days: The Invention and Impact of the Moog Synthesizer. Cambridge, MA: Harvard University Press.
Popper, F. 2007. From Technological to Virtual Art. Cambridge, MA: MIT Press.
Prusinkiewicz, P. 2004. “Art and Science for Life: Designing and Growing Virtual Plants with L-Systems.” In XXVI International Horticultural Congress: Nursery Crops; Development, Evaluation, Production and Use, edited by C. G. Davidson and T. Fernandez, 15–28. Toronto: International Society for Horticultural Science. CD-ROM.
Prusinkiewicz, P., and A. Lindenmayer. 1990. The Algorithmic Beauty of Plants. New York: Springer-Verlag.
Sims, K. 1991. “Artificial Evolution for Computer Graphics.” Computer Graphics 25 (4): 319–328. See also http://
Sims, K. 2007. “Galapagos Interactive Exhibit.” http://
Smith, M., ed. 2005. Stelarc: The Monograph. Cambridge, MA: MIT Press.
Steuer, J. 1992. “Defining Virtual Reality: Dimensions Determining Telepresence.” Journal of Communication 42 (4): 73–93.
Sutherland, I. E. 1965. “The Ultimate Display.” Proceedings of the International Federation of Information Processing Congress, edited by W. A. Kalenich, 506–583. Washington, DC: Spartan Books. Reprinted in R. Packer and K. Jordan, eds. Multimedia: From Wagner to Virtual Reality. New York: W. W. Norton, 2001.
The Times (London). January 8, 2000.
Todd, S. C., and W. Latham. 1992. Evolutionary Art and Computers. London: Academic Press.
Tofts, D. 2003. “Avatars of the Tortoise: Life, Longevity, and Simulation.” In “Generative Computation and the Arts,” edited by Paul Brown. Special issue, Digital Creativity 14 (1): 54–63.
Tofts, D. 2005. Interzone: Media Arts in Australia. Fisherman’s Bend, Victoria: Craftsman House.
Wang, P., A. A. Becker, I. A. Jones, A. T. Glover, S. D. Benford, C. M. Greenhalgh, and M. Vloeberghs. 2007. “Virtual Reality Simulation of Surgery with Haptic Feedback Based on the Boundary Element Method.” Computers and Structures 85 (7–8): 331–339.
Webster, A. 2006. Solar Stacking. (An electrochemical sculpture.)
Webster, A., and J. Bird. 2002. “Tuning Pask’s Ear.” Video (7:26).
Whitelaw, M. 2004. Metacreation: Art and Artificial Life. London: MIT Press.
Wilson, S. 2002. Information Arts: Intersections of Art, Science, and Technology. Cambridge, MA: MIT Press.
Woolf, S. 2004. “Expanded Media: Interactive and Generative Processes in New Media Art.” PhD diss., University of Sussex.
Zivanovic, A. 2005. “The Development of a Cybernetic Sculptor: Edward Ihnatowicz and the Senster.” In Creativity and Cognition 2005: Proceedings of the Fifth Conference on Creativity and Cognition, edited by L. Candy, 102–108. New York: ACM Press.