What would Artificial Intelligence find aesthetically pleasing?
The burning question of generative art and its audience
In the cultural sphere, technology is often misunderstood as merely a delivery medium for fixed content (one need to think only of the progression in information transmission from the telegraph, to the telephone, to the television). "Generative art" turns this idea on its head - wherein not only is the content not fixed in advance, but the technology is the medium for that variable and manipulated content.
The most commonly cited definition of "generative art" is from Philip Galanter:
"'Generative art' is an art practice where the artist creates a process, such as a set of natural language rules, a computer program, a machine, or other mechanism, which is then set into motion with some degree of autonomy contributing to or resulting in a completed work of art." *1
While the field of generative art has expanded to include definitions including those of Noam Chomsky's "generative grammar" linguistic theory and Slavoj Zizek's notion of a "generative matrix", *2 the consensus is that underlying the generation process are rules, decided by the originator and effected by the computer, or input from the user. Therefore, it is the technology, effected by the user, that configures the content, then creates it, (renders it) and, then lastly, in most cases, delivers it, as art. In this narrowest sense, a "batch" command in Photoshop could be viewed as a generative art process.
As generative art is made with some adherence to a system of rules, which the artist/programmer sets, the resulting variable and manipulated content and structure, while surprising to its recipient, is not necessarily unpredictable. Baring the increased complexity brought about through the use of random parameters, if you know how the rules work in essence (if you understand the code underneath the interface) you can guess at, and in the more simple cases, predict the outcome.
That said, when the process is executed on the Web, or within a computer network, or with collaborators beyond the artist-programmer, a new, deeper layer of process and questioning becomes evident.
What this essay, inspired by Cornelia Sollfrank's net.art generator (since 1999), seeks to reveal is this deeper layer: the aesthetic experience of the recipient of generative art - the curator, the viewer, the audience, the user, the "interactor", the participant. Considering the role of the computer network in the generative art process, this essay also questions how the net.art generator can now be read within the more recent contexts of generative art, code art, or software art, and how it first read when it was first made, in the context of early net.art, or art made using computer networks and the World Wide Web.
A brief history of net.art aesthetics
Let us begin, for the sake of argument, with a definition developed by IBM to measure signal "logical depth" versus information unpredictability or noise:
"The value of a message appears to reside not in its information (its absolutely unpredictable parts), nor in its obvious redundancy (verbatim repetitions, unequal digit frequencies), but rather in what may be called its buried redundancy - parts predictable only with difficulty, things the receiver could in principle have figured out without being told, but only at considerable cost in money, time, or computation. In other words, the value of a message is the amount of mathematical or other work plausibly done by its originator, which the receiver is saved from having to repeat." *4
This particular assigning of value is inherently one of usability (and I use the term "usability" not in the common Jakob Nielsen sense of accessibility concerning web-design, but in the more general aesthetic sense of usefulness - a kind of signal-to-noise ratio concerning the work's message). It is a very different assignation of value to that normally ascribed to art works. The assignation of value in art has been built up over centuries of philosophical study of sense perception balanced with an understanding of how our critical faculties make judgements; this field of inquiry - aesthetics - is predominantly based on static art forms, and for contemporary examinations of media art, hinges on the ruptures between representational visual art forms and signifying written art forms such as poetry. *5 It bears remembering that both Kant and Hegel felt that if you could find a use-value for the art you were admiring, you hadn't made a true judgement of the work's aesthetics. To this end, much contemporary media art made using technological computational devices deliberately plays with the use-value of the technology it appropriates, as for instance electronic music plays with the grey area between signal and noise. This doesn't cancel out the above definition's usefulness to us here however, for the definition includes a belief which maintains a stronghold in art criticism and aesthetics, namely ascribing value to the amount of work the originator of the art has done in encoding their message, and to the surprising (and perhaps ironic) way in which the work's message has been encoded. *6
net.art (with the dot) as a field of art production can be said to have begun in 1996 with a meeting of like-minded cultural producers (artists, activists and computer programmers) through the online mailing list nettime. *7 It is difficult and mostly useless to generalise their art output in 1996 and 1997, given the ironic stance of the artists in question in relation to their own labelling as a group. However, crucial to their early interventions into the Internet and the World Wide Web was their understanding of the importance of the artist as author, as originator, and - something that seems significant now with hindsight - the importance they placed on how they personally knew each other, how they found one another's work. Josephine Bosma quotes Olia Lianina, one of the net.artists, as saying:
"net.art for me means early art on the web... not a thousand hits a day because you are listed in a net art category somewhere, but feedback of people who found your web site by chance." *8
In this sense, keeping one's tongue in one's cheek, it could be argued that net.art is what Internet art was before the museum and the institutions of the art world got wind of it.
In 1997 and early 1998, during the hey-day of the semi-secret, face-to-face meetings of the net.artists of nettime, 7-11 and other mailing lists, much of net.art's aesthetics were most easily judged by those on the outside on two levels: 1) 2D screenal graphic design, or 2) concept. Paramount was the way the artist conditioned (or didn't) the manner in which a viewer navigated through the work (graphically or conceptually) - whether JODI's curious and deliberately frustrating hell.com or Olia Lialina's poetic text and image narrative my boyfriend came back from the war. The way a viewer navigated through the work, based on the decisions made in reference to the above levels, was paramount. To those closer to the inside of this avant-garde was an importance on the political stance and potential for change quoted by the personalities behind the works; the Internet itself was developing to be a major tool in the shifts around geo-political thinking.
female extension and net.art generator
This was certainly the case in 1997 when female extension was made. Curator and art historian Inke Arns describes the project thusly:
"In the scope of the female extension project in 1997, Cornelia Sollfrank generated individual Net-art projects for 289 virtual artists with the aid of a computer program (PERL-script) that collected and automatically re-assembled HTML material at random on the WWW." *9
female extension therefore seems to be a bridging work, between the early net.art crowd of works that were political in spirit and recycled in content, and a latter emerging pocket of Net art (without the dot) practice which sought to directly engage with the institution and the questions around authorship, originality and exhibition that the art system engenders.
When it came to the museum trying to get to grips with net.art (and Net art generally), curators turned to their tools of aesthetic judgement. They sought out works which generated interest on both levels mentioned above - for instance, Amy Alexander's multicultural recycler project which collaged screenshots of webcam images taken from websites the world over. Curators drew upon their art historical understanding and appreciation of media such as printmaking and photography, as well as their curatorial knowledge of other "networked" art movements such as conceptualism and mail art. Rather than get bogged down in learning programming, they focussed instead on how the work was received, its surface value if you like. To be blunt, the success of early net artworks was judged primarily on interface aesthetics and secondarily on their manipulation of the structure of the delivery mechanism the technology was understood to be in place for.
Clearly, female extension was a success on both levels. From the perspective of the curators and the jury it was predominantly a success on the first level - hundreds of works with interesting interface aesthetics were accepted into the competition by the museum (nevermind that as websites, these "artworks" didn't go anywhere or do anything). Yet, from the perspective of the originator of the works, the artist Cornelia Sollfrank and her team of programmers, female extension was a resounding success in its unexpected use of the possibilities of the technology. Here, on this second level, the true value of the work can be seen to be exactly, as the definition of logical depth states, in its "buried redundancy", and in its double entendre of that - the redundancy only being immediately notable and visible to its creator and, then later after the unveiling of the generator, to those who didn't check the code first.
Here we had a work of art (in many ways a performance), driven by a computer network, which challenged aesthetic judgement of net.art beyond what had been theorised to date. How do we account for this deeper level of assignation of value to this work on this second level of the work's concept?
One way is to investigate the machine behind the smart artist, the net.art generator. Further developed in 1999, the net.art generator is actually three generators, the work of four computer programmers commissioned by Sollfrank. The resulting works are more text or image-based depending upon the bias of their searches and the different search engines they pursue. Cornelia Sollfrank commented that she felt her net.art generator shifted the focus of Net art work:
"... from fixed content (like Olia Lialina's stories, for instance) to a more process-oriented, and at least pseudo-interactive, model of distributed authorship and a more or less unforeseeable outcome. In comparison to what the first generation of net.artists had done, this seemed to be revolutionary to me." *10
Thus the work of the generators not only, as all net.art did, comment on originality versus reproduction, materiality versus ephemerality, and the place of art in an online browsing space, it also more dramatically questioned, through its automation, authorship and artistic individual genius. Through its coded use of search engines, the resulting artworks can also be seen as comments on the temporal conditions of the net in 1999, more so than most of the net.art that came before.
net.art and the museum In 1997 when Cornelia Sollfrank created female extension and even in 1999 when net.art generator was released, the relationship between Net art (both net.art and Internet art) and the museum was substantially different than it is now; museums were just getting to grips with this new medium for making art. *11 Technology historian Friedrich Kittler, speculating on the digital museum writes:
"The haven that, according to Valery, a museum offers for paintings and statues would be called 'storage' in the cold terminology of computerese. For of the three functions of a Universal Discrete Machine (storage, transfer and processing of input data) two functions, transfer and processing, are omitted in a museum. Nothing must be changed in things that are preserved (...) The museum is a hybrid medium that - historically or opportunistically, but in any case unsystematically - coupled its elementary storage functions to other media of processing and transferring. The rift between art and technology, as inflicted by the classical museum, remains untouched by modern forms of presentation [...] Their mutual indifference may have been possible and feasible so long as the transfer media dominated the other technological media, while the museum continued to specialize in storage. But in the context of an information system that tendentially abolishes the principle of transfer because it reduces all transfer to the processing and computing of data, mere storage per se appears somewhat dysfunctional." *12
One significant change since female extension is that museums have realised that they must be more than simply storage specialists if they want to maintain engagement with their audiences and if they want to maintain their unique roles as specialists of art. This means that in addition to hosting competitions for great web-based art, museums also commission new work in all new media, starting most prominently with net art. *13 But while museums have changed for the better in this sense (artists now can see them as patrons of new work rather than just as conservators of older work), the Internet has in some ways changed for the worse. As Frederic Madre has stated, the Internet "is not artist-friendly", *14 and this is all the more true in our "post-dot-com-boom era". State-sponsored museums - which purport to hold things for the public good - are rarely able to defend or protect artists and their work when they clash with the more commercial or governmental aspects of working online. *15 female extension should be commended for showing that the sanctuary the museum was proffering (its exhibition-as-storage function) was just as flimsy and subject to corruption and breakdown as any other commercial enterprise.
The format of competitions and open-call submission policies still exists in nearly every exhibition opportunity and venue for new media art. This is in part because curators, with the technical and financial constraints of their institutional roles, still have trouble finding out what is out there in the gloriously unregulated space of the Internet. Whatever filters are in place are quickly knocked down by new forms of art which stretch the boundary definitions of the filtering systems. *16 Walker Art Center curator Steve Dietz used to joke that net-artists JODI had received significantly more hits to their site than "the combined museum sites of all the curators who have done a studio visit with them." *17
Yet as the storage function of museums appears to be taking a back seat to the institutions attempts to engage in transfer and processing (i.e. commissioning the making of new art, and educating the public about the process behind it), the notion of the exhibition still depends on predominantly static works of art. Florian Cramer writes:
"net.art has shown that nothing goes unless you produce commodity objects for exhibition spaces. (Or, if your art is immaterial or process-oriented, that you produce commodifiable objects as, to borrow from semiotics, indexical replacements of those processes, like Fluxus artists did with their 'boxes' and multiple objects.)" *18
The difficulty, of course, is that most new media, Net art and generative art has to struggle to appear static.
The context of the Net itself has of course changed in the last half decade also; the push of automatic content generation is that much stronger and more ubiquitous. Think for instance of the "amazon recommends" feature of the online book shopping experience. The placement of cookies, the sniffers that track your movements across a website and rank your choices and take note of your "eyeball drag time" all work to generate a tailored surfing experience for you, the consumer. Just like with television, in certain sectors we are becoming the products of the Internet, made malleable to suit the needs of the business economy, our choices insidiously being made by someone other than ourselves.
To this context of life in the networked age, generative art has taken form as an understood field of production. Tethered nearby are software art, code art, and the open source movement. The contemporary field of Internet art includes numerous artist-designed alternative web-browsers. To counteract the increasingly limited content and the ever-increasing restrictions on access to that content online, movements such as copy left have sought to bring art made for computer networks into the gift economy. This in itself rubs against traditional definitions of aesthetics, which seek to assign one author and one static product to each value judgement. net.art generator names its programmers and it is only now, four years after its inception, that the field is right for its originator, the artist Cornelia Sollfrank, and her collaborators to jointly author the script and release it under a "general public license" for others to use. If we concede that the notion of an "artist-programmer" is, in terms of art history, an emerging genre/label, then we must concede also that an understanding of art's aesthetics can no longer be judged in a static space free from the changes in the attedant socio-economic framework which have altered our notions of authorship and originality.
"In generative art, the artist-programmer and machine work in partnership to disrupt tired old mythologies of creativity - emphasising that art conforms to formal structures, and that computers might be particularly useful for manipulating these structures." *19
Which brings us to the question of the reception and judging of generative art by the audience or user. As Geoff Cox, Alex McLean and Adrian Ward point out in their essay, The Aesthetics of Generative Art, "the aesthetic value of code [arguably the basis of generative art] lies in its execution, not simply its written form" (correctly noting that a study of the latter would be one of poetics, not aesthetics).
Thus, to our first definition of where in the transmission of information "value" lies (in buried redundancy), we have to add to an understanding of the work value of the originator, an understanding of the action - the making manifest, the execution - of that work.
But, by what or whom is that action executed? The artist? The technology? The audience? The user? Furthermore, where do we draw the boundary between original authorship and joint collaboration?
These questions put to generative art have different answers depending on the context in which the work is made and seen - from the commercial sector (software development as art) to the more traditional art sector (other forms of net art, other digital practices, music). My interest here is in what answers are available when considering what happens when generative art acts up against the context of the exhibition - the museum or gallery institution? What is the audience experience?
exhibiting generative art
This makes the exhibition Generator recently seen across England all the more interesting. *20 It included, on the digital art side, in addition to Cornelia Sollfrank's net.art generator: Ade Ward's Auto-Illustrator 1.1 (a reconfigured new version of the famous Adobe Illustrator software); Joanna Walsh's oulibot generative text works made from 3D mathematical models; Alex McLean's forkbomb (a perl script that crashes its host computer, and visualises the process); Netochka Nezvanova's genograph live cell photographs; and on the non-digital but nevertheless generative art side the work of conceptual artists such as Yoko Ono, Sol Lewitt and Angus Fairhurst.
As the exhibition has toured it has changed with each visitor's experience and input. The branches of Yoko Ono's trees are laden with more wishes, the printouts of the digital works grow in number and complexity as the generating programs work themselves out, Sollfrank's net.art generator database grows each day. Only the drawings, photographs and linear video works remain the same from venue to venue. Under the tough conditions of exhibition and viewer interaction these works show their wear and tear - computer operating systems crash and freeze up, printers run out of paper or ink, networks fail and are reconfigured.
While the publicity for the exhibition states the projects exhibited are "self-generating" and that all work "is produced 'live', in real time, with some elements continuing indefinitely" *21 this is entirely conditional on the presence of an audience member for many of the works - and not just the digital ones. Neither Ono's nor Sollfrank's works would be complete without the audience member's participation. The familiar argument that the art is in the eye of the beholder is brought to its logical and technical conclusion here when, as Cox et al have shown, "the aesthetic value" of the work "lies in its execution". While the execution routine might well be coded into the work, it won't actually happen in the case of some of the works until an "interactor" or participant puts pen to paper or finger to keyboard.
Even since this exhibition began its tour, new digital networked art projects have emerged which expand the idea of generative art outwards, shying away from the conceptual aesthetic of the code in favour of a romance with the image. Lev Manovich's Soft Cinema project, in his words, creates fictional stories as a series of short movies:
"While the voiceover narrating the stories was edited beforehand, everything else is constructed by software in real time, including what appears on the screen, where, and in which sequence. The decisions are based partly on a system of rules, and are partly random. In other words, Soft Cinema can be thought of as a semi-automatic VJ (Video Jockey) -or more precisely, a FJ (Film Jockey)." *22
Even more interesting is Thomson & Craighead's Short films about flying which generates on-the-fly internet-based movies; it grabs moving pictures from airport webcams, intersperses them with old-fashioned silent film inter-titles generated from the net by a text-bot, and layers on top a soundtrack randomly chosen from a digital radio station. This project strikes me as akin to net.art generator in that it is creating new work from the enormous amount of visual and textual content already online. It is reconfiguring information, data, images and sounds into new aesthetic experiences for the viewer, in this case letting the program do the choosing, without necessitating input from a participant.
This is an especially seductive tactic for it gets around the "usability" debate mentioned above which, in the field of aesthetics is moot when it comes to judgments of quality or beauty, but which in the world of new media, and particularly web design, has come to have greater significance. Dead-end questions of usability plagued a great deal of early net.art (I think again of the work of Jodi) and continue to circle the creation of software as art. Here the aesthetics of the art are clear: beautiful if not sublime images and movies to watch. As Steve Dietz has said:
"[The] argument regarding usability is a slippery slope one - you would never say Jasper Johns should make his painting more user-friendly. [One should] be careful about using an instrumentalist notion of interface in relation to art online. 'It is hard to get around.' 'So what? What does that have to do with art?'" *23
This emergent field of generative art made using existing webcontent thus also produces a further distinction between what could be called screen-aesthetics and database-aesthetics, where in one case changing the interface dramatically changes the reading of the work and in the other the interface might indeed be the context and the content of the work itself. *24
Which brings us back around to the question of the experience of the viewer. In these more recent cases of "generative art", it doesn't matter to the aesthetics of the work, even to the buried redundancy of the message (the information transmission) within the work if the execution of the work is self-generated by the code, by the technology or by the user. What matters most is the context in which the generative artwork is run and seen.
Interestingly, Cornelia Sollfrank decided for the third venue of the Generator exhibition to print out large-format images of the webpage-art generated by the net.art generator to hang on the gallery's walls. From my position as a curator, I see this as a concession to the audience experience; a gesture to the commodity status of art objects within the framework of an exhibition, a nod to the traditional formats for viewing art. For the artist, these print-outs were a provocation to the expectations of the audience of the show, a way of asking them what kind of fixity they expected from an inherently process-based art work. Clearly, for an understanding of the net.art generator as generative art itself, this shouldn't have been necessary, pleasurable and curiously absurd but not necessary. This begs the question, how might the work appear or be received in conditions other than an exhibition format? How does the work read online compared to how it reads within the context of this group show? Or, as Joanna Walsh has asked of her generative text poetry, what if the reader were an artificial intelligence engine, another generative computer programme?
By making the code for the net.art generators available, Cornelia Sollfrank will now be able to encourage an audience of users/interactants beyond the machine itself. Her artistic collaborators will soon include those who know the rules the code creates and obeys, and who can change those rules at will. Might the results "disrupt" more than the structures of the exhibition-submission-competition-display? Might they further manipulate art historical notions of authorship and creativity? I expect so; stay tuned to find out.
Where is the art happening here?
A few thoughts on the net.art generator with respect to Sarah Cook
Blithely, signs generate, codes communicate, bits process and the chains of symbols steer towards each other, working away at each
other. The artist/author and viewer/reader are amazed at all the goings on, but the programmer simply smiles a knowing smile.
The codes are deep and productive. And what about us? An extra knot in the chains of signs, not at the beginning or at the end, but yet another passage generator in the endless line of self-processing signs, whether of language or imagery, or simply digital information. Another location of the execution and performance of binary commands.
But also - and here's where questions arise - a place of performance. Performance for whom or what? The value of a code is its logical depth. How do we ascertain it? By measuring the amount of work required to decode it and its embedding in a context in which it only reveals meaning. It must be at this point that we can begin to tell the difference between the execution and the performance for something, in this case, us.
For something or someone - that is, for an entity that asks, "What for?" At this moment, a depth appears that is made explicit not so much through work as through enjoyment and pleasure. A depth that operates more reflexively than recursively - that is, a subject calling itself into question rather than merely feeding back into the process the result of an algorithmic operation as the initial value for what follows.
Art can't help it: For all the sublimity and unapproachability for us as supposed administrators of meaning or as sovereign-creative subjects, it remains beautiful nonetheless. It isn't work that's the final criterium of its value, but rather, desire and pleasure in the mental activity engaged in the attempt to fathom what's happening there, and above all, what's happening to me, placing me in meaningful connection with the artistic event (or whatever sort of experience it might be). In this way, as pleasure taken in one's own mental stimulation, art is a "premonition of good" (Kant), a possible way to experience freedom in the free play of imagination and intellect which - once aroused - never quite approaches literal satisfaction. The experience of art, when it opens up such space for play, allows one to perceive, think and act differently than one does within a previously limited framework.
Codes, even codes of communication, are chains of signs - perforce. No reaction is left unperformed; the chain cannot be broken as long as the electricity flows and the program doesn't crash. But art - and this is Kant's still-relevant thesis dating back to the 18th century - begins where its expression in language dissolves or fails altogether, where the complete translation of an experience into representational signs cannot be made. Where, with regard to an event, there remains at least some speechlessness and the event raises a question that remains unanswered. The fatality of code is not that it operates with signs and its framework of rules - we do, too. Its bad luck has much more to do with the circumstance that it has not yet arrived at the pleasure of existentially calling itself into question.
*1 Quoted on the eu-gene mailing list. *2 Chomsky's definition can be found in Syntactic Structures 1965, Zizek's can be found in Mapping Ideology 1997. *3 Florian Cramer writes: "Digital net.art (mostly) is digital code, software art (mostly) is digital code, with the only difference that (a) focuses on networks and (b) focuses on algorithms/executables. In any case, the art can be created by a solitary genius-star-artist or an anonymous collective [think of all the underground software available in the internet], it can be a clearly defined/limited piece of work [like a website, an ASCII movie... etc. in net.art or a desktop user application in software art] or it can be something where it is difficult to determine what the "work" actually is." - CREAM, posting April 19, 2002 *4 Charles H. Bennett, Logical Depth and Physical Complexity in Rolf Herken, ed., The Universal Turing Machine: A Half-Century Survey, Oxford University Press, 1988, p.230) *5 See David Rodowick's book Reading the Figural, or Philosophy After the New Media. *6 For a dated but nevertheless brilliant and provocative discussion of the necessity of irony in contemporary art versus the lack of it in most digital new media art, see Lev Manovich's essay, The Death of Computer Art, 1996. *7 For a short and sweet history of net.art, I refer you to a text written by Josephine Bosma The Dot on a Velvet Pillow for an exhibition of net.art at the Oslo Museum of Contemporary Art, curated by Per Platou in March 2003. Available from http://www.student.uib.no/~stud2081/utstilling/bosma.htm *8 Josephine Bosma, quoting correspondence with Olia Lialina, ibid. *9 Inke Arns, in: Update 2.0, Current Media Art from Germany, Goethe Institut, ZKM | Zentrum für Kunst und Medientechnologie Karlsruhe,2000. *10 From an E-Mail from the artist to the author, February 2003. *11 Cornelia Sollfrank interviewed by Tilla Telemann, "Hackers are Artists - ans Some Artists are Hackers", http://artwarez.org/femext/content/interviewEN.html *12 Friedrich Kittler, Museums on the Digital Frontier, in The End(s) of the Museum, Fondacion Antoni Tapies, 1995 *13 Want a list? The Walker, The Tate, Dia, MOMA, the Whitney, SFMoMA, ... some to coincide with new exhibitions such as biennials, others in a programme of their own. *14 In a presentation at De Balie on net art and art criticism, January 2002. *15 This is itself is a fascinating topic, think, for instance of the censorship imposed by museums' technology company sponsorships, or of the Tate's commissioning of Heath Bunting's information dispersal project, Border Crossing (2002). *16 The debate around rhizome mostly centres on this issue. *17 Steve Dietz, in his essay From Signal to Noise available from http://www.walkerart.org/gallery9/dietz *18 in a posting to CREAM, April 19, 2002 *19 Cox, Mc Lean, Ward, The Aesthetics of Generative Code, 2000, http://www.generative.net/papers/aesthetics *20 Curated by Tom Trevor and Geoff Cox it toured to Spacex Gallery in Exeter, at the Liverpool Biennial, and Firstsite gallery in Colchester, GB. *21 Press release that accompanies the exhibition. *22 Lev Manovich, project description for Soft Cinema, 2002. *23 From a conversation concerning the curation of a forthcoming exhibition about databases Pretty Good Access, Regina, Canada, December 2001. *24 The number of artist's projects run from the software carnivore are a good case in point.