[CHRISTOPH]
And indeed, if one glances at the young but growing discipline of Artificial Life (AL) it seems as if this positive notion of chaos is put to work. Chaos gets incorporated into the very ways computers and machines perform, it advances to a fundamental element of how computers are programmed and set to fulfill their tasks.
Artificial Life has changed the strategy for programming computers: instead of trying to exert complete control over the working process of machines, principles of biological life and its organization as we know it - for example as in neural networks or evolutionary mechanisms - are introduced as metaphors for putting computers to work. Normally computers are programmed in a "top-to-bottom" manner, that is to say for every possible instance or usage of an application a specific execution order has been preestablished. If there are cases which haven't been thought of by the engineers and haven't been incorporated into the software exceptions occur - in the best case one gets an error message, in the worst one has to unplug the machine.
With artificial life this "top-to-bottom" strategy where the performance is determined and controlled form "above" gets turned around. Computers can work now in a "bottom-to-top" manner. On the basis of chaos, out of random initial conditions, computers evolve on their own solutions for a desired task. The crux is that one does not specify anymore how a solution for a specific problem has to be achieved; there is not a predetermined, finite structure which "foresees" all possible events. Biological "functions" allow machines an autonomy to generate - independent of the programmers ability to map all possible events - their own solutions; solutions which are often even out of the scope of the programmer: We don't understand anymore how it functions.
The change from a "top-bottom" to a "bottom-top" strategy in programming has developed "lifelike" behaving machines, where "lifelike" is that which is nonmechanistic, nonpredictalbe and spontaneous[1]. And this behavior encounters us first of all in a visual way. Besides the actual construction of little hardware robots, it is on the screen in form of simulations that we perceive this artificial life. In more appropriate terms one has to say that we begin to explore a new way of seeing with the generation of visual worlds based on AL.
Currently there are different areas in which Artificial Life influences the ways digital images are produced or manipulated. "Neural networks" serve as a first model for programming in a "bottom-top" manner and for producing autonomous, non-deterministic images [2]. The problem solved by software modeled after neural networks is basically to simulate the behavior of several interdependent agents. A classic example in this field is the animation of schools of fish or flocks of birds. The main idea is that there is no central agency which controls the behavior of each individual. Rather each agent communicates with its environment, with other agents and with itself, i.e. its internal needs. The result is that the animation of several interdependent agents doesn't have to follow a preestablished script determining the position of each single individual in advance. The movements and the behavior of the agents are based on a feedback loop, so that the animation clip can be left on its own and becomes autonomous. This way of animation is already well accepted since the bats in "Batman II" or the wild bees in "The Lion King" are crossing the screen based on these just mentioned principles.
As a second example for the interweaving of Artificial Life and digitally produced pictures let's imagine the typical situation of a person engaged in computer animation. After the desired object is created in the simulation, one sets out to create a series of slightly modified single pictures which produce in their final assembly the animated movement effect: someone walks form A to B. The point of genetic algorithms [3] is, that one doesn't have to engage anymore into the production of every single object-position which makes up the movement sequence. One rather specifies the conditions - the so called "fitness criteria" - of the movement - for example distance to move, the preferred style, the freedom of joints etc. - and the computer will generate out of random initial conditions something that is quite close to the desired result [4]. A computer with genetic algorithms is therefore highly flexible regarding the problems it can solve, the movements it can generate since it finds on evolutionary mechanisms a solution on its own: through a kind of software Darwinism.
It is especially in the context of interactive media, that some implications of these autonomous animation's regarding fundamental irritations of our "used" ways of seeing appear more clearly. Artificial Life is not only used to produce some sort of animation sequence but to simulate "lifelike" behavior and interaction in general. Suddenly we can fish for the fish or feed the flock of birds. Interactivity in the field of AL has broken the form of a preestablished structure which only gives the user more or less complex but always rigid, fixed options to react: once you know them there is nothing surprising anymore in it. It's only a matter of time and skill until you master the Video Game. Now user and simulation interact dynamically and influence each others actions respectively; you have to read and anticipate the action of the other, never quite sure what's really going to happen. The meaning of a picture flickering on the monitor only evolves in time, since there is no way to infer a further state from the current one. Not even the author or designer of the work knows what's exactly going to happen next - the only way to find out is to interact, to experience in time, since the possibilities of the image only evolve in time.
From that it follows that the representation itself plays a subordinate role in interpreting the image. One could say that interactivity which is still a rather additional feature of some images in our present days achieves a more and more dominant role in the process of giving meaning to images. Now images rather become an appendix to interactivity. Realistic, fancy graphics loose their importance -even if they gradually improve. The action speed of a character, for example, influences how it is perceived by the user: fast moving agents are perceived as young, upheat and intelligent. Compared to the dominance of the visual in terms of a good, flawless design in our present time, it is remarkable that this might actually be only of equal or minor importance in respect to the possibilities of a simple emotional interaction: evoking a smile on the face of your virtual encounter, knowing about it's inner state instead of realistic graphics lies at the heart of an intriguing interaction.
Pattie Maes writes that it is still a long way down the track until one understands computer generated behavior and it's illustration and until one can refrain from giving the user verbal guidance regarding the ways in which to interact with the simulation. But maybe these difficulties which still accompany our understanding of autonomous pictures respectively our interaction with them don't just call for a more perfect, realistic and detailed simulation of the reality we already know. Maybe the just mentioned difficulties are an expression of our awareness that simulating behavior in forms we already know misses the point. We know that the dog [5] we see on the screen is a virtual dog and not identical with its real, alive counterpart and therefore it is only "natural" to expect different interaction patterns. The challenge consequently is to experiment with new combinations of behavior and visuals which do not have to live up to our expectations, real world experiences and their "real-world" models. One should start form scratch and allow a development of images that are not aligned at real world models but evolve their own specific computer laws of usage, understanding and seeing. Simulated dogs or the simulation of a face are nothing else than the first means to appropriate the new terrain of simulated behavior. But what do we "see" if we start to really explore this new terrain and dare to animate pure forms, space or ...?
Marcos Novak visualizes liquid Cyberspace - a space which does not obey anymore a static, earthbound notion of space but rather follows fluid and continuos transformations. Based on genetic Algorithms space ceases to be something inert and fixed and is more likely perceived in terms of behavior instead of scale and form. Architecture as the art of (Cyber-) Space is "no longer a single edifice, but a continuum of edifices, smoothly or rhythmically evolving in both space and time. Judgments of a building's performance become akin to the evaluation of dance and theater" [6]. In Cyberspace the contour and forms of space, respectively of edifices are constantly changing, developing new shapes, edges, configurations of blocks and objects. What we see in Cyberspace and what we normally would expect to be lasting for centuries suddenly exposes signs of permanent flux and continuous transformation.
But how are we to read what we see, how do we have to understand space with behavior? Do we have to read the visuals in terms of a narration where previous states explain future ones or is it a structure which is displayed: the permanent transformations mirror different states of one and the same object. Or do we rather have to concentrate on the rhythm displayed - comparable to music - in which the changes in space take place and which suggest a judgment in terms of temporal composition. There are different criteria after which we can judge behavioral space and it is far from clear which ones to choose, that is to say that there is a variety of effects still to explore how this transformations in the image, these morphings are perceived. As Gene Youngblood states: "One can begin to imagine a movie composed of thousands of scenes with no cuts, wipes or dissolves, each image metamorphosing into the next, [with] infinite possibilities, each with unlimited emotional and psychological consequences" [7].
Be it the transformations in space or the composing of a movie with morphing techniques instead of cuts, in both cases our used and everyday practice of seeing gets interrupted. There is no sedimented way of reading autonomous images, no established form in which we can access the meaning of liquid, morphing space and forms of behavior. It is rather an open field for experimentation, where we have to slowly develop an understanding for the possibilities as well as for the specific ways one can use these images.
Viewed from a slightly different perspective it is suggested that through the transfer of behavior into the computer a new aesthetics of (visual) behaviors [8] will develop. Previous transfers of one medium into the other have show, that this simple transfer itself affects the characteristics of that which gets placed into a new medium. The simulation of the typewriter with the computer gave birth to a much more scattered and fragmented writing experience; through the simulation of the mailing system by e-mail a style of communication with its own laws and symbols developed. If one conceives the simulation of autonomous action in this avenues it is not too much of a speculation that there will evolve a new aesthetic of forms of behavior with its own laws and rules which are specific for the simulation of behavior. This aesthetic will follow its own premises instead of being oriented at or concerned with a perfected mimesis of real world living creatures. Time will show what happens to the behavior (of a dog as well as to the dog itself) when it gets simulated. In this sense Roy Ascott writes: "Classical Aesthetics dealt with the behavior of forms, Technoetic Aesthetics deals with forms of behavior" [9].
The freedom of visuality consists in acknowledging this fundamental difference
between a real world phenomena and its simulation and allows for experimentation
with ways of "seeing" (in the sense of understanding) behavior
without confining it to the already well known.
[1] See: Pattie Maes (1996): Artificial Life Meets Entertainment. Lifelike Autonomous Agents. In: Lynn Hershman Leeson: Clilcking In. Hot Links to a Digital Culture. Seattle: Bay Press. p. 213
[2] See: Craig Reynolds BOIDS. http://hmt.com/cwr/boids.html
[3] Genetic Algorithms are one instance of the broader field of Artificial Life. Algorithms are tools for solving problems. If they have the prefix genetic, algorithms are capable to modify themselves where each new variation of "problem solver" is called a generation. Genetic Algorithms are capable of passing their achieved capability to solve a specific problem to the next generation, which means that they sloley "breed" better and better solutions. It is the survival of the best problem solver.
[4] Gritz, L. J. K. Hahn (1995): Genetic Prgramming for Articulated Figure Motion. In: Journal of Visualization and Computer Animation.
[5] The ALIVE project at MIT tries to simulate a virtual dog with whom users can dynamically interact. See: Maes, 1996, p.213 ff.
[6] Marcos Novak (1994): Liquid Architectures in Cyberspace. In Micheal Benedikt (ed.): Cyberspace. First Steps. Cambridge, Massachusetts: MIT Press. p. 225 ff.
[7] Gene Youngblood (1989): Cinema and the Code. In: LEONARDO Computer Art in Context Supplement Issue p.28
[8] The success of tamagotchi in Japan indicates the growing populariyation and the overall fascination to create artificial life: Orignially conceptualized as a kids toy it soon found its way into every age group. Basially it is a plastic egg which shows likfelike behavior, you have to feed it, clean it, care for it and if you don't it ultimately dies. You have to hatch it like a real animal without a choice to turn it off... that's all!
[9] A Glossary by Roy Ascott
Go back to Table of Contents