Matthew Parrott

Contemporary Art

Judith Allen

February 25 2004

Genetic Art:

A Brief History and Application to Immersive Narrative

Genetic algorithms allow artists to evolve, rather than design, complex forms. Such algorithms could assist in the design of complex immersive narratives. The rise of photo-realistic computer graphics over the past decade has necessitated a dramatic increase in data management requirements. This increased complexity results in an increased cognitive load for the artist. Computer scientists have developed a number of techniques for dealing with increasingly large databases, including genetic algorithms. Closely related to the field of artificial intelligence, genetic algorithms reproduce the theoretical processes of evolution to refine a base process until that process produces acceptable results (Grau 297). For computer graphics, this involves creating a simple form, then mutating and recombining the form with others to create offspring that are similar to but more complex than their parents. At each generation, the artist or computer selects undesirable results to be removed from the list of potential parents. This process emphasizes the artist’s aesthetic choices, rather than his original inspiration and technical aptitude, hiding the vast amounts of data required for the work.

Installations have been the venue of choice for genetic artists. Forerunners in the graphic research community worked under the influence of the theoretical biologist Aristad Lindenmayer (Prusinkiewicz, Linemayer 1). Fine art innovators within the field of genetic installations include Karl Sims, Thomas Ray, and Jane Prophet. Beginning with single automata, the art form progressed through the 90s until artists synthesized entire cyber-organisms. Simple, abstract 2D images quickly gave way to fully articulated 3D environments. Natural events are now reproducible as well, allowing the artist to use environmental processes as analogies for the creation of art.

Moreover, a level of complexity is attainable which would be impossible if explicitly defined by the artist (Grau 300). Interactive installations and modeling interfaces have been the primary applications for these techniques. Obviously, if genetics is applicable to image production, scene creation, and object interaction, narrative design could also benefit from genetic techniques.

Genetic processes center around the following four concepts: grammars, replacement, mutation, and recombination (Grau 298). All living organisms are represented abstractly by DNA structures containing genotypes. The genotype is the genetic encoding while the organism created is the phenotype (Prusinkiewicz, Linemayer 28). In an abstract sense, DNA is like a language. Amino acids act as a code which follows rules, rules which may be thought of as a grammatical structure. A grammar contains a set of rules which acts upon a set of symbols (Prusinkiewicz, Linemayer 2). Each symbol points to a physical result. Symbols may be concatenated to form complex expressions. In English, a common combination is <subject><verb>. This can be considered a simple rule. Using this one rule, a speaker may construct a number of complex thoughts, such as “Steve runs,” “bazooka shoots,” “lawyers lie,” or “vacuums suck.” A writer may begin with a simple idea, a character sketch or plot line which answers the question “what if?” Then, using a language’s grammar and his own stylistic preferences, he continues to combine ideas, weaving an increasingly complex story. An author may spend his life expounding on the same central theme, all the while remaining within the bounds of the same simple rules. DNA, within its set of simple rules, is capable of creating infinite variations on the same theme.

Replacement is the action of replacing a simple symbol in a grammar with a more complex one. Returning to our earlier example, we can use replacement to make our <subject><verb> rule more complex. A subject does not necessarily have to be a single noun. It could be a noun, an adjective with a noun, or a compound subject. Using this slightly more complex rule, we can write the grammatical rule:

<subject> =

<subject><subject> |

<adjective><noun> |

<noun>

This rule implies that subject may be replaced by three possible options. Thus, the simple thought structure <subject><verb> could become <noun><verb>, <adjective><noun><verb>, or <subject><subject><verb>. The last result is interesting because it contains the <subject> symbol which can be replaced and expanded still further. So, the thought could then possibly become <subject><adjective><noun><verb>, resulting in a fragment like “Bob and Big Bubba wrestle.”

Mutation is the random adjustment of genotypes within the bounds of the grammar (Sims 320). The randomness of the mutation determines how much the new child will be like its parent. A low randomness could result in the noun “tree” becoming “bush” while a high degree of randomness would result in the noun “boat” becoming “pizza.” Mutation results in the pattern that siblings all appear to be a combination of their parents but are unique. Generally, forces constrain mutations to those that are beneficial.

Recombination is the process of taking portions of two genotypes and forming a new one from these portions (Sims 322). In this way, information is passed from one or more parents to offspring, which resemble aspects of the entire parent generation. A recombination of the two sentences “Joe eats ice cream” and “Sally ties her shoes” might result in “Joe ties ice cream.”

Finally, selection controls the processes of genetics (Sims 322). A fitness function is defined that determines which offspring will survive and which will perish. In nature, “survival of the fittest” is a classical, theoretical fitness function. A writer uses fitness functions based on style, impact, and coherence to prune and refine his text through successive drafts. Often with genetic art, the artist’s aesthetic intuition determines the fitness of each successive generation. Of course, the shortcoming of comparing writing with genetic art is that in writing the processes are carried out explicitly by the writer. In genetic art, the computer implicitly progresses through revisions; the artist is merely the referee.

A theoretical biologist, Aristid Lindenmayer developed a grammatical system for representing plant-life (Prusinkiewicz, Linemayer 1). Known as the L-System, the grammar could replicate the growth pattern of simple plants utilizing the concepts of developmental algorithms and self-similarity. Developmental algorithms attempt to replicate the growth patterns of organisms. Self-similarity implies that small parts of an organism resemble the organism as a whole. Simple parts of the grammar, such as branches, may be replaced with more complicated symbols representing forks and twisting branches (see Figure 0).

Natural growing organisms, however, rarely appear purely mathematical or systematic. Plants often appear to flow through their environment, reacting to sunlight, weather, soil type, and terrain. Adding supple, organic geometry to the scientific models creates a more appealing and realistic form (Prusinkiewicz, Linemayer 54). Jules Bloomenthal’s Acer Graphics 1984, exhibits a more advanced L-system featuring curved limbs and a uniform surface (see Figure 1). The work shows that the theoretical L-system could be used to create a photorealistic form.

While accurate branching creates general forms, sub-branch features are also a large part of the essence of a plant. P. Prusinkiewicz’s Lilac inflorescences 1990 shows the potential of L-systems to produce apex affects such as blossoms and leaves (see Figure 2). Here, Prusinkiewicz applied the same grammatical techniques used in branching to replicate the self-resembling floral patterns (Prusinkiewicz, Linemayer 63). As a bud grows into a flower, petals are added in a systematic manner that algorithms can easily recreate.

D. R. Fowler’s Water-lilies 1990, shows the L-system employed in a complete, complex environment attempting a fine-art application of grammars (see Figure 3) (Prusinkiewicz, Linemayer 108). Cluade Monet’s Water-lillies pool – Harmony in green was the major inspiration for this work. While the flora of the scene are accurate to nature, the lack of unified lighting diminishes the aesthetic appeal of the piece. Of course, Fowler is a scientist not an artist. Yet, the piece demonstrates a problem within genetic art – advanced processes do not create artistic insight. The needed insight would come as artists explored the latent content within the processes of genetics.

At the time of Fowler’s work, grammars were still static entities. Artists created environments that were snapshots of nature, lacking latent life. Something was needed to progress genetic art from explicitly dictated forms to self-determined organisms. Karl Sims pioneered the concept of genetic graphic techniques in his classic technical paper “Artificial Evolution of Computer Graphics” (Sims 319). Sims proposed that the evolutionary techniques of mutation and recombination could be used to create complex organic simulations of structures, animations, textures, and interactions.

Beginning with the concept of the L-system, Sims created the animated work, Panspermia in 1990 (Sims index.html). Based on the concept that the universe is saturated with the spore of life, the three-minute animation explores evolution on several magnitudes of scale (see Figure 4). Beginning with an expanding universe, the work progresses to the growth of star systems, planets, continents, terrain, and finally plant life. L-systems grow, mutate, and combine with other plants to create new species. Growth functions allow plants to start as seeds and rapidly grow using simulated time-lapse animation. Additionally, large patterns of growth can be seen in cultures of plants as they migrate (via seed dispersion) across landforms. The piece unifies the themes of complexity, genetics, self-replication, and chaos to explain the complex dualities of life.

In his 1993 interactive installation Genetic Images, Sims explores an interface that allows viewers to select and breed 2D images (see Figures 5 and 6) (Sims index.html). The images are generated from mathematical functions that create variations in color across the display. A large display shows the current abstract form while sixteen monitors in front of the image feature possible children of the current image. The viewer selects which child is most appealing by standing in front of a certain monitor. Selected images are allowed to remain in the environment and pass their genetic code on to the next generation, propagating their unique visual style to the next generation. Successive generations thus tend towards the aesthetic desires of the participant. Images are swapped, new children are generated, and the process is repeated until the viewer creates a piece he is pleased with. At any time, the user may request that the system start over at which time the computer creates a new, simple genotype and the process begins again. This work is important because it hides the complexity of the formula and genotype from the artist. Instead of designing a complex form, the artist is invited to participate in its evolution. Also, the distinction between viewer and artist is blurred, a concept common to many installations, happenings, and interactive artworks and which would continue through the exploration of genetic art. The piece also questions traditional concepts of creativity. Not only is personal craftsmanship removed, but the artist’s creativity is limited to their choices. A nearly infinite number of control paths allow every user to create works that are unique and often the style of the individual user is apparent in finished images.

Evolved Virtual Creatures, 1994, features a system that evolves a vast array of artificial life forms to perform certain tasks (see Figure 7) (Sims 15). The creatures a bred using grammars and then compete in games to determine their fitness. Events include locomotion, hide and seek, following, and combat. Creatures grew which could hop, slither, bounce, and roll. Interesting creature designs also appeared, some having tentacles, some with legs, and others with a-symmetric blob-like structures. Unique and viable solutions are created which may have been counterintuitive and thus eliminated by a human designer. The preconceptions of the artist no longer limit his creative experiences. Instead, the machine helps the artist expand his train of thought. Sims pushes the concept of artificial creativity begun in Genetic Images to the point where the machine is apparently more creative than the artist. The fact that it is apparent creativity must be stressed, as the artist has created the process for the generation of the work. Thus, the artist takes on a god-like role in the creation of creative creatures.

Galapagos 1997, allows the user to not only select appearance, but also animation (Sims index.html). Our current biosphere exhibits the only case of theoretical evolution, limiting our ability to study the process. Galapagos allows viewers to observe other possible outcomes for life (see Figure 8). Making the ideas of Creatures more accessible, Sims allows people to select modes of animation similar to the image selection process of Genetic Images. This time, the aesthetic space which viewers navigate is motion. Appendages are added and mutated to creatures, along with varying skin textures and personalities. Movement allows for the extra dimension of time to be added to the evolutionary installation experience. The complexities and nuances of animation are now easily managed by evolving them.

One step beyond the genetics of Karl Sims is a concept known as “artificial life.” Closely related to artificial intelligence, artificial life seeks to simulate the life patterns of unconscious organisms (Grau 308). The key concept is not representing the entire system at once, but rather simply controlling individual agents. Flocking behavior results and the beings appear to take on aspects of intelligence. Artificial life’s inventor, Thomas Ray seeks to allow the evolutionary processes more control within the computer (Ray index.html). His artworks feature very large colonies of individual agents which exhibit the development of multi-species environments. His work Tierra 1996, features self-contained programs that are allowed to mutate and breed with other programs. The advance of the colony is shown visibly by a complex animation of abstract entities representing programs. Darwinian concepts were applied to allow the programs to compete among themselves (Grau 315). The programs were given basic instincts as self-preservation, consumption, and reproduction. These instincts were encoded in eighty or so simple instructions which the programs themselves were allowed to alter. Programs were actually able to evolve themselves, a concept borrowed from biologist Nils Baricelli, creating more complicated versions of themselves (Grau 316). Such freedom allowed for startling results. Species of programs formed with personalities such as Parisites, Helpers, and Combatants. Also, some species developed unique attributes such as immunity, cheating, and the ability to reprogram other units. Extremely complex behaviors resulted. While not strictly a visual work, Tierra is important because it validates many evolutionary theories. Ray claims that the substrates of silicon and electricity are in fact capable of producing hormones and nervous systems. The programs exhibited a primitive intelligence in their ability to adapt and improvise, implying that creativity could be within reach of machines.

Thomas Ray has recently moved to testing Tierra over the internet. Netlife is intended to reproduce the explosion of variations and complexities we see in the natural world (Grau 318). By releasing artificial life to be distributed among thousands of network nodes, Ray is seeking that “Big Bang” of computer-aided evolution through the nearly unlimited bandwidth of a global network. The envisioned software capable of creating images of inconceivable complexity, capable of surpassing the creative input of its creator and exhausting available hardware capacities. Like the neo-realists, who sought to create images indistinguishable from photographs, artificial life attempts to create organisms which are indistinguishable from life.

The great problem that arises out of Netlife is whether it is life or not. Artificial life supporters claim that the work does not simulate life, but rather that Netlife is alive. Indeed, given an imaging system sophisticated enough, artificial life could create scenes that viewers accept for life, similar to how audiences fled from the train in the Lumiere brothers’ film. Also, biologists have accepted the idea that life consists of systems – that life is alive because it is organized matter (Grau 318). If machines are capable of organizing themselves then they may be defined as alive. Taking the idea further, the question becomes whether a work of art could in fact create itself. The problem here is that, despite artificial life’s success, an initial starter culture or set of parameters is still required. Thus, the artist is never completely removed from the loop. The artist maintains ultimate control (and aesthetic choice) because of their ultimate necessity.

By limiting the overpowering complexity of artificial life through careful pruning of base cultures, an artist may create aesthetic arrangements of synthetic organisms. Jane Prophet uses artificial life techniques to synthesize entire ecological systems down to the smallest detail. However, she progresses past her predecessors by using Darwinian processes to create content, not just form. Her installations use process to make a statement; it is not just the complex organisms which are important, but the social implications of those organisms (Prophet biog.html).

SWARM 1997, synthesizes the interactions that take place between 20,000 individual agents, exploring the concepts of communal society and the “hive mind” (Prophet art.html). Audio of buzzing bees drones over countless, faceless voices of a large crowed. Three different stations resembling large hives display different visualizations of the agents within the hive. One displays images of society on sliding LCD screens, similar to the grates used by bee farmers to collect honeycomb. The second station displays trackers of the participants within the room along with the virtual agents. Here, the viewers are integrated directly into the hive. The final image displays MRI brain images with agents moving through the mind’s pathways like ants in an ant farm. Taken in totality, Prophet uses Hive to illustrate that while a hive is made up of many individuals, the entire group may also be seen as a single organism. The hive is neither an individual, nor a crowd, but rather a continual flux of action and reaction; individuals are both self-determined and subservient. The technical interpretation points to the capability of artist to create entire organisms made up of cellular units. However, the deeper conclusion of the work recalls the anxiety within humanity that we are merely mindless products of mass culture. Contradicting this desire for identity is her use of the medium of genetic art, which would appear to some to hijack the sacred act of creativity from humans.

Prophet elaborates on the theme of the complete organism in her work The Internal Organs of a Cyborg Sarcophagus, 1998 (Prophet cyborg.html). Viewers are able to interact with a figurative sculpture upon which is projected the complex processes of human anatomy (see Figure 10). The projections also contain the futuristic narrative of two characters, one which becomes a host to the other’s personality after a botched cybernetic surgery. While not a truly genetic work, the piece does exhibited the potential to combine art patterned after life with narrative and the uses of narrative within interactive installation.

Technosphere 1999, features creatures that viewers evolve (see Figure 11). Technosphere is actually a rework of a previous piece of the same name, redone to be fast, interactive, and fun (Prophet technosphere.html). Installed National Museum of Photography, Film & Television at Bradford, UK, the installation features touch screens which allow users to select different growth options. Organism types include plants, herbivores, carnivores, and parasites. Users select aspects of organisms from an extended list of possible attributes. The relationship between environment and creature design is clearly visible. An ill-paired form of locomotion with the environment will result in a species dieing out after a few generations. The importance of this work though lies in the fact that the creatures live from day to day in a vast environment and are able to interact and compete with each other. Organisms exist within their natural, real-time context, rather than in an isolated creation environment. Thus, the entire biological system is replicated along with the genetic process.

Decoy, the Landscape Room 2001, features images of modern British landscape overlaid with evolved images of landscape (see Figure 12). In this way, Prophet allows viewers to see the countryside before humankind’s intervention (Prophet landscape.html). Many people adore the countryside of Britain but fail to realize that the vistas are far removed from nature. Centuries of careful manicuring have created a sterile façade devoid of natural trends. Decoy pictures scenes surrounding Holkham Hall in North Norfolk. Currently on display at Norwhich Castle, the work exists alongside idealized landscape works from the 17 th century. Five large screens display scenes of pristine looking-pools, arched bridges, and graceful running fences. As the installation progresses, flora is introduced the scene which grows and reproduces in natural ways. In a few minutes, land forms return to their possible natural appearance. More than an artist’s interpretation, the piece shows how natural biological processes would actually affect the environment. Humankind’s ideal landscapes are replaced with the freedom of natural forms. The artificial facades of modern society often hide the natural patterns of life. Prophet asks people to consider their current constrained existence in comparison with the reality in which they might live.

Genetic algorithms could be use to evolve narrative events, rather than design them, simplifying the process of creating realistically complex, immersive plots. The sense of “presence” an immersive viewer experiences is directly linked to the quality of the narrative surrounding the immersive environment. A fully articulated environment contains far too many variables to be explicitly designed. Natural life is filled with a rich amount of life that occurs at various levels of magnitude. Recreating these details within a real-time environment is near impossible. Using genetic techniques could add complexity to basic environments. Background activity and minor characters could be evolved rather than designed. An empty street scene could be filled with people. These people could then have their interactions with each other evolved, a process much simpler than scripting all of those extras.

The Liquid Narrative Group has explored methods for recombination within virtual narratives (Young, Riedl, Branly, Martin, Saretto 24). Plot lines may now be expanded based on existing patterns (Young 2). Organizing a grammar for plot will allow artists to generate new and creative narratives. For example, archplot narratives generally consist of an inciting incident, complications, crisis, climax, and resolution. Each of these sections can then contain plot twists, value changes, and subplots. Symbols could be assigned to each plot device and replacement could be used to increase the complexity of plot lines. An artist could begin with a simply starting culture, possibly character sketches or climactic scenes. The story would then be evolved through successive generations. Artist could move beyond explicitly creating stories to creating methods of storytelling.

Genetic art has itself evolved. Grammars provided the basic symbols and rules to model organic life forms. Evolutionary processes allowed works to be mutated and recombined into future generations. Artificial life modeled the interaction between organisms and their ability to adapt to situations. Finally, agents allow for entire systems and organisms to be replicated.

Genetic art proposes the idea that the artist is not necessarily someone who creates art, but rather defines a method for its creation. Like conceptual art, it is the idea itself, expressed in genotypes, which creates the work. Artists no longer need concern themselves with minute details, but can leave them to organic processes. Evolving a design is far more intuitive than designing it. Yet, control is often lost and specific results are difficult to achieve. Processes for tailoring evolutions are required to create more predictable results. Questions of the nature of creativity also point to the need to better define the relationship between the artist and the work that his creations create.

Genetic artists face the challenges of overcoming the direct analogies between genetic algorithms and biological processes. Organisms are not the only things that could be represented. Sims broke briefly from the trend with his Genetic Images which featured abstract images. However, further growth of the medium could be accomplished by applying the techniques to narrative design. While genetics has been fully applied to applications of form, Prophet is one of the few artists who has explored its importance to content. Building upon her work and the work of the Liquid Narrative Group, genetic art will become a medium for the creation of methods of narration.

Figures

Figure 0. Branching patterns of herbaceous plants.

Figure 1. Acer Graphics Jules Bloomenthal (1984).

Figure 2. Lilac Inflorescences P. Prusinkiewicz (1990).

Figure 3. Water-lilies D. R. Fowler (1990).

Figure 4. Panspermia Karl Sims (1990).

Figure 5. Genetic Images Karl Sims (1993).

Figure 6. Selections from Genetic Images Karl Sims (1993).

Figure 7. Evolved Virtual Creatures Karl Sims (1994).

Figure 8. Galapagos Karl Sims (1997).

Figure 9. Swarm Jane Prophet (1997).

Figure 10. The Internal Organs of a Cyborg Sarcophagus Jane Prophet (1998).

Figure 11. Technosphere Jane Prophet (1999).

Figure 12. Decoy the Landscape Room Jane Prophet (2001).

Bibliography

Grau, Oliver. Virtual Art: from Illusion to Immersion. Cambridge: The MIT Press, 2003.

Prophet , Jane. Jane Prophet: Biography. 2002. CARTE. Jan 28 2004. < http://www.cairn.demon.co.uk/biog.html>.

Prusinkiewicz, Przemyslaw, Aristid Lindenmayer. The Algorithmic Beauty of Plants. New York: Springer-Verlag, 1990.

Prusinkiewicz, Przemyslaw. Visual Models of Morphogenesis: A Guided Tour. May 1997. Department of Computer Science, University of Calgary. Feb 23 2004. <http://www.cpsc.ucalgary.ca/Research/bmv/vmm-deluxe/Section-09.html>.

Ray, Thomas S. Artificial Life. Jul 15 1996. ATR Human Information Processing Research Laboratories. Feb 2 2004. < http://www.isd.atr.co.jp/~ray/pubs/fatm/>.

Sims, Karl. “Artificial Evolution for Computer Graphics.” Computer Graphics 25.4 (1991): 319-328.

Sims, Karl. “Evolving Virtual Creatures.” Proceedings of the 21st annual conference on Computer graphics and interactive techniques (1994): 15-22.

Sims, Karl. Karl Sims Homepage. 1987 – 2001. GenArts, Inc. Feb 4 2004. < http://www.genarts.com/karl/ >.

Young, Michael, Mark O. Riedl, Mark Branly, R.J. Martin, C.J. Saretto. “An architecture for integrating plan-based behavior generation with interactive game environments”. Journal of Game Development 12.1 (2003): 24-32.

Young, Michael. “Steps Towards a Computational Theory of Interactive Narrative in Virtual Worlds.” Unpublished manuscript of North Carolina University (2003).