Artificial Creativity

Artificial Creativity:
Three artists who make art that makes art
George Fifield
First published New England, Vol. 15 No. 2, February/March 1994
© 1994 George Fifield

One of the truly revolutionary artworks in the city of Boston resides not at any of the traditional fine art or alternative spaces, but at the Computer Museum, where on the sixth floor a Vax computer connected to a pen plotter, is running Aaron, a program by Harold Cohen. Cohen is one of a number of computer artists from many fields including the visual arts, video, music and literature, who are twisting the questions posed by artificial intelligence research to explore the limits of the creative act and the role of the artist within it. They are making art which by making art itself, functions as a model for the structure of creativity.

Aaron draws original figurative images, each drawing unique in it’s composition, yet representative of Aaron’s individual style. The subject matter of the Computer Museum’s version of Aaron is of human figures existing in an eden-like jungle. In some of the drawings the figures dance, sometimes they stare out and sometimes peep about recognizable images of grasses, palms and trees. It takes Aaron about 50 minutes to do a drawing but as long as the plotter is supplied with ink and paper, it will continue to produce these original artworks 24 hours a day.

Aaron creates by replicating the artist’s understanding of the world. It follows a branching series of rules: all figures have a head, two arms which are connected to the shoulders; gravity exists, therefore feet (not heads) touch the ground; if one object is in front of another, it obscures part of it. Since the early seventies, Cohen has been codifing these rules, adding to them enough randomness at the right points so that each drawing is different. Over the years, Aaron has progressed from a generator of simple abstract shapes to its present complex images. The program has been built up, rule sequence by rule sequence, as Cohen set more and more sophisticated goals for his creation. An early inspiration were Native American petroglyphs. Could a program be written that generated them? At one point, Cohen made a study of toddlers scribbles. Each step has added a refinement to Aaron’s style. Recent Aaron drawings, even more sophisticated than the Computer Museum version, look to be recognizable portraits, albeit of imaginary people. Cohen explains that he has enjoyed the experience “watching the cognitive development of an entity”

Next step is a painting machine. Cohen says that Aaron is already a quite a respectable colorist. But software needs to be created that can translate the Red, Green, Blue information of the monitor to additive paint information and then a mechanism which physically mixes the paint and dabs it on the canvas. “Art needs a physical presence, not the seductive glow of phosphor on a screen.” Cohen says.

Karl Sims is a computer/video artist, who creates art using the paradigm of natural selection, as life has created different species. He is the artist in residence at Thinking Machines, Inc. Located in Cambridge, they are the developers and manufacturers of a super computer known as The Connection Machine.

Sims’ program starts by displaying sixteen simple pictures generated by sixteen random equations. The viewer picks the picture he likes the most. The program then makes sixteen random changes or additions to that equation, generating sixteen more pictures. Changes might involve color, shape, texture or a number of other parameters. Some are almost the same and some differ greatly but all are clearly descendants from their genetically successful predecessor. The next “favorite” is then selected and another generation is created. The viewer acts as natural selection applying an aesthetic filter to generations of random mutations. The images effortlessly grow more and more complex. In only five generations, the viewer has chosen one out of over a million possible paths. Sims explains “People do the interesting work and computers do the mathematical work, the boring work.” His goal: “To make complexity without having to understand it.”

The idea for the program came when Sims was working on his computer animated video Panspermia. He needed a way to design many different images of plants and wrote a program to mathematically create plants using twenty different parameters. At first, he sat modifying the parameters to see what resulted, but this was so time consuming that he wrote another program to randomly generate different parameters. He would then choose which results he liked best and the machine would then modify those equations to generate new generations. Sims then used this process to create whole images for his animation Primordial Dance in 1991. Then mixing this process with scanned images and morphing techniques he made Liquid Selves , with music composed by Peter Gabriel and John Paul Jones, for the Memory Palace at the World’s Fair in Spain in 1992.

In the spring of 1993, he presented the process itself as an artwork. A Connection Machine was installed in the Centre Georges Pompidou, in Paris, connected to sixteen monitors. Each monitor had a footpad in front of it so the image choice could be made by simply stepping in front of the favorite. From March 5th to May 2nd, the installation generated 1.5 million pictures responding to 204,123 choices by museum goers. This averages out to 22,000 images a day. All of them were stored as equations which Sims can recall and examine. The experience is empowering. Within a few quick generations, the viewer finds they are playing with images of great personal, emotional significance, yet except for simple choosing no work was done.

Marcel Duchamp described the creative act as an equation which included the artist, the art object and the spectator. Sims has effectively removed the artist, leaving the art object and the audience conspiring together to make art. Not that this process comes easy. The audience exhibited behavior ranging from shyness to arguments. “People aren’t used to making choices like that in an art gallery setting.” He says.

Neil Leonard has programed his Macintosh computer to play, in real time, improvisational jazz with himself and his sax. This interaction of man, musical instrument and machine is startling and fun. At times Leonard’s computer follows his lead, then suddenly it jumps out ahead building on a passage that Leonard introduced with fresh variations. Leonard, an accomplished composer and saxophonist, has explored many directions with his music. He played in 1992 with Ibrahima Camara at the Boston Globe Jazz Festival and last year he preformed solo at the Audio Art Festival in Krakow, Poland. In November, in a concert presented by the Space at the ICA, Leonard, along with Juan Blanco and George Lewis explored the furthest reaches of musical interactivity with computers. “Computer music,” he says, “is the most exciting area, because, unlike jazz, people’s preconceptions haven’t been formed yet.”

He calls his computer programs automatic performers and they generates their music according to Chaos theory. Mathematical concepts of chaos differ from randomness; chaotic behavior is seen in the motion of planets and is used in the study of business cycles. Leonard has translated these chaotic algorithyms into musical forms. In his studio, Leonard uses the computer as a tool to write musical compositions, which he then polishes and performs. In concert however, the computer becomes a collaborator. The Macintosh listens to the music from Leonard’s sax and to the degree it can categorize it, modifies the music it is generating on the synthizier. His goal, Leonard says, is to “make (the computer) surprise me more. I would like to let the machine set the mood of the evening.”

All three men are quick to limit the comparison between the creativity of their programs and human artists. Cohen says “creativity only takes place when a entity can modify its own performance … can change its mind, which Aaron can’t.” But the end product in each case exhibits surprising inventiveness that wasn’t placed there by the artist. All three are working from different models: Cohen’s program has rules and branching decision trees; Leonard’s work clearly demonstrates the aesthetic properties of chaotic systems in music; Sims’ images suggests how much randomness and natural selection are a part of creativity.

Sims further explains “if we accept randomness as part of our creativity let’s make the machine take over that part. It can generate random complexity far more efficiently and faster then we can.” In Sims’ work, all the viewer provides is the choice. One future direction in artificial creativity is for computers to provide the choice as well. But if the computer’s adeptness at complexity, either random, chaotic or preordaned, creates an object, image or composition that fufills an aesthetic need in us, does it matter? If the art succeeds as art, who cares how the artist achieved it? Whether the artist was meat or machine? As Neil Leonard cautions “because the machine can be anything, we have to think carefully what we want it to be.”

George Fifield is a video artist and the video curator of the Space, Boston’s alternative public art organization. For three years he has presented a monthly video art program. This January the Space is moving the program to the Institute of Contemporary Art, where it will be called VideoSpace at the ICA. Last January, he co-curated The Computer Is Not Sorry, an exhibition of computer installation art.