Evolution of Art by Aesthetic Selection

Aaron Davidson


      In recent years much debate in the world of art has been focused on the merits of art enhanced or created by advanced technology. Critics claim that technology does away with the creativity and skill of the artist. Advocates insist that technology frees them from the non-essential aspects of art instead enabling them to concentrate purely on their creations. In Part I I will discuss some general problems in relation to art and technology. Part II will focus on a special blending of art technology and nature. In Part III there is even an artistic movement which embraces technology as part of its overall philosophy.

Part I. Technology & Art

      As in any discussion of Aesthetic Philosophy, the argument hinges around the presented definition of “Art”. Before we can go any further, my use of the term must be clearly defined to prevent any misinterpretations. Art is expression and creation. It is the manipulation of the external world with the purpose of reflecting the will, emotion, or existence of the artist. The difference then between “good art” and “bad art” is the skill of the artist in his medium, or perhaps more subjectively, the consensus among the audience members. Many people will argue with this definition calling it too narrow or too general, or even completely off the mark. I am not an experienced aesthetic philosopher or artist, so this definition may be directly biased towards my ignorance. To my credit, I did examine aesthetic philosophy texts and have dabbled in art. Having never done this stuff professionally or academically, this makes me a bit of an aesthetic quack. However, I feel that nearly all things which we would define as art will fit into this definition, and likewise those things which we do not endow with the label of “art” will not be covered by it.

      The gut reaction to technology and art has usually been an almost vitalist claim that art has to be physically created by a human being. There is a view that art must be banged and molded out of clay or metal by the human touch, and layers of paint applied by delicate strokes of the artist’s hand. If a person is not “getting their hands dirty” and skillfully manipulating some physical or mental medium then they are not doing art. What is considered technology these days? Is a paintbrush not a lower form of technology? It is a tool used to manipulate our environment in a way we cannot physically do with our naked bodies. I often wonder that if the paintbrush was a recent invention, if there would be an outcry from the artistic community that paintings that are not finger painted are not works of art. New technologies are viewed with a critical eye. Often, a new technology will make an art form obsolete. It is a natural reaction to fear changes in one’s field.

      A common complaint made about people using technology to do art is that the technology can make it too easy. Talentless people can make mediocre abstract paintings by simply applying Photoshop filters to an image, for instance. It is true that this does happen, but give technology to talented artists and they will be able to produce incredible things with it. Technology does not raise the bar for only the talentless and shallow, but also the skillful and deep. Much technology does not have the above effect of making things easier for people. Typically technology is a daunting challenge for a person to master. Many projects which involve creating art with technology require a huge amount of skill and knowledge. Take the post-industrial robot wars of Mark Pauline for instance (Kelly, 1994). His fighting mechanical beasts made from “Obtainium” are unique works of art, and the shows in which they meet their doom, doubly so. It is safe to say that not many people have the talent to wield technology the way Pauline can. It is certainly not easy to do so.

      Technology, from the view of the art world, is simply a blanket term for tools which are new. In a few decades the computer will be as entrenched in art as the paintbrush is today. It will be just another tool at the service of the artist.

Part II. Aesthetic Selection

      One particular way of melding computer technology with art holds a unique set of properties about it that sets it apart from everything else. By emulating the process of natural selection on a computer, artistic creations can be evolved. Instead of natural selection, aesthetic selection is used as the criteria for survival of these “artificial artisans”.

      At its heart, evolution is an algorithmic process which leads to accumulations of design over time. Natural selection was Darwin’s explanation of how things can be designed without a designer. With the natural world selecting for beneficial traits, any population of randomly variant self-replicators would tend to statistically favor the members that have better traits than others. Iterating this blind process yields non teleological design. When humans interfere with nature (as we so often do) and make intentional choices over which things get to live and which things get to die and who gets to mate with whom, it is no longer termed Natural Selection but instead “Artificial Selection” or “Unnatural Selection”. The genetic algorithm, that which is key to the power of evolution, remains more or less unmodified. Only the criteria for selection has been changed. The process becomes goal oriented, or teleological. When the criteria behind artificial selection becomes aesthetic appeal, we will call it “Aesthetic Selection”.

      Aesthetic Selection has been around for hundreds, if not thousands of years. Humans have bred various animals such as dogs for hundreds of years to produce the numerous and sometimes bizarre forms of Canis familiaris we see today — everything from poodles to bulldogs.

      One of the first to present the idea of computerized Aesthetic Selection was biologist Richard Dawkins. Dawkins created a simple program on his Macintosh that could draw symmetrical stick figures from a set of eight parameters (or “genes”) which controlled branching factors, recursion, and so forth. By starting with a small population of random “BioMorphs” on the screen and selecting whichever stick figure appears most interesting, one can step through the BioMorph library of form. One can evolve BioMorphs through whatever selection criteria they desire. Dawkins noticed that some BioMorphs started to fold their branches back upon themselves and form bodies, making them look like insects and frogs. By continuously selecting for insect-like characters he could guide the evolution of his BioMorphs into deeper insectoid spaces. While the form-space is far smaller that the one in Borges’s Library of Babel2 , it is still a vast space in which one can easily become lost. He initially expected his program to draw only plant like shapes such as bushes and trees, but it was actually capable of much more than that. Trees, bushes, flowers, spiders, beetles, ants, butterflies, frogs, letters of the alphabet, rockets, and vases all were possibilities representing just a small fraction of the total BioMorph landscape.

      At the 1991 SIGGRAPH conference, Karl Sims unveiled his amazing “Artificial Evolution for Computer Graphics” paper and accompanying software. Running on a Thinking Machines Corporation Connection Machine (a massively parallel super computer), Sims software could draw full colour, high resolution images created from digital genomes. By mutating and mating these artificial genomes a user could evolve images of stunning complexity and of organic detail. The amazing thing about Sims' system was how rich a universe one could explore with what was a very simple program at heart. Sims system allowed the artist to explore the space of all coloured two-dimensional images.

      I was so impressed with the descriptions of Karl Sims genetic images that eventually I began to entertain the idea of writing my own system for exploring evolving art. In the summer of 1997 I had enough spare time to develop a quick and dirty working prototype that tested my ideas for the algorithms I would need to create such a program on a desktop PC. At the time, I had little idea at all about how Karl Sims' software actually worked under the hood, so I essentially had to reinvent the wheel. As it turns out, I ended up mimicking Sims system quite closely in many respects. It worked shockingly well for a prototype — I was able to evolve images after only a day or two of programming. My program worked with strings of numbers which represented the genotype of an image. These strings of numbers could be mutated (randomly change a few of the numbers, or add and delete numbers from the strings), and mated (mix two or more genomes together. In order to construct a phenotype (the resulting image) from the genotype, the program would translate the string of numbers into a machine code for a simple virtual computer. For every pixel in the image that needed rendering, the software would do the following:

  1. Place the X,Y coordinates of the current pixel into specific locations of the virtual machine's memory banks.
  2. Execute the genotype as a computer program on the virtual machine.
  3. When the genotype program terminates, read some specific values from the memory banks and use these values to create the colour of that pixel.

      In other words, at a high level, the genotype can be viewed as a function which maps the coordinates on a 2D-graph to a colour value.

      As it turned out, I would not return to the project until two years later when I acted as team leader in a six-person software engineering project for a computer science course I was taking at the time. Our three month project was to resurrect my prototype and create a finished, usable piece of quality engineered software. We called the program “BioGraphy” and produced a fully featured evolving art system. Users could save and load saved genotypes, adjust the parameters used in the genetic algorithm, export the images to common graphics file formats, and most importantly, explore 2D- Graph space. A critical problem that had to be overcome was the speed. To remain an interactive and captivating program, images had to present themselves as quickly as possible. Karl Sims had a CM2 to render his images with. He could dedicate a processor to every pixel in an image. To minimize rendering time, we included the ability to use other computers on a network to donate their idle time towards rendering the images. With this system, one could combine the power of several connected desktop PC’s to render all of the images in parallel. The system worked great. Users were enchanted by the program, and would spend hours “playing god” and designing swirling psychedelic patterns. At times, it can be quite addicting.

      To date, however, it has not achieved as remarkable results as Sims’ software. Sims used even higher level structures for both the image rendering and the genetic algorithm than those used in BioGraphy. Rather than the simple assembly language of arithmetic and memory operations used in BioGraphy, Sims used convoluted tree structures which were designed to be exact functions.3 As primitive operations in his language, he had both simple arithmetic and complex high-level image processing functions. Having these functions built into the system from the start makes it possible to evolve incredibly complex organic looking images from nearly random genotypes. His mutation and mating algorithms could also take advantage of the high level tree structure of a genotype. These advantages make comparing the two programs like comparing a bacterium to a mammal. My next goal, of course, is to attempt a system more similar to Sims’ software.

      What unique properties do programs using aesthetic selection hold in respects to art? In systems like BioGraphy, there is little technical knowledge one needs to know. Nearly anyone can evolve images without any difficulty. The challenge in making it art is in making good aesthetic selections. Someone with bad taste will evolve ugly images.

      The hundreds of choices a person makes at each step leads to a very expressive image of that persons’ cumulative aesthetic choices. Give two people the same seed images and their end results will be completely different. Images evolved through aesthetic selection are direct representations of an artist’s aesthetic desires. It also takes some skill to evolve really good images. Some people just can’t seem to get the hang of it. Since these images can be generated by intentional acts of skillfully chosen aesthetic choices, the results fit most people’s criteria for art.

      This is a case where the technology handles every task of creating the art, but leaves the pure domain of aesthetics to the user. As Karl Sims describes,

...the humans supply decisions of visual aesthetics, and the computer supplies the mathematical ability for generating, mating, and mutating complex textures and patterns. The viewers are not required to understand the technical equations involved. The computer can only experiment at random with no sense of aesthetics -- but the combination of human and machine abilities permits the creation of results that neither of the two could produce alone.

      A final issue to ponder is the role of the programmer of programs which enable aesthetic selection to occur. As creators of the entire system which an artist may use to evolve art, they are a sort of meta-artist. Programming can, in many cases, be viewed as an art. Any programmer can testify that aesthetics play a large role in crafting software. Programmers seek elegant solutions to their problems and strive for genius in their code. The difference between machine generated code and code written by a human is clearly visible. Machine generated code may work and do the same task in the final result as a human made program, but the human will typically write a much smaller, elegant piece of code. The machine will write a bulky, boring, and downright ugly-looking program.

Part III. Transhumanist & Extropian Art

      There exists a few art movements in which technology is invited to play an important role. Transhumanist and Extropian art movements are two related genres which have been born out of their respective general Transhumanist philosophies. Transhumanism is the view that as mankind evolves into the future, augmenting himself with his technology he will transform into something so entirely different than today’s humans that he is a new species. Extropianism is a philosophical subset of Transhumanism, which is introduced in the “Extropian Principles” as

....challenging human limits by means of science and technology combined with critical and creative thinking. We challenge the inevitability of aging and death, and we seek continuing enhancements to our intellectual abilities, our physical capacities, and our emotional development. We see humanity as a transitory stage in the evolutionary development of intelligence. We advocate using science to accelerate our move from human to a transhuman or posthuman condition. As physicist Freeman Dyson has said: "Humanity looks to me like a magnificent beginning but not the final word."

      Extropian and Transhumanist art welcome new technologies with open arms and optimism. They challenge the inertia of tradition, and move to explore, change, and grow through their art in Transhumanist style. Art has often been used as a way of exploring and understanding the many different facets of being human. As humanity transforms into post humanity, art must be adapted to the same forms in order for us to make sense of it all.


Dawkins, Richard. The Blind Watchmaker. 1987. W.W. Norton.

Goldblatt, David & Lee. B. Brown. Aesthetics: A Reader in Philosophy of the Arts. 1997. Prentice Hall, New Jersey.

Kelly, Kevin. Out of Control. 1994. Perseus Books, New York.

Sims, K. Artificial Evolution for Computer Graphics. Computer Graphics (Siggraph '91 proceedings), July 1991, pp.319-328.