Artificial Life: A Programmer’s Perspective
By ai-depot | August 23, 2002
Artificial Life
Artificial Life
What is life? That question may sound too difficult to be answered, but from a scientific standpoint, life is nothing more than a pre-defined course of functional activity carried out by organic entities (cells) and a series of changes determined by some bio-chemical reactions. All other complex phenomena that occur originate from these two basic attributes only. Emergent properties are inherent to life; complex behavior emerges from simple rules. For example, let the organic cell in a human body be the functionary unit and the DNA code be the set of rules. The actions of each cell are determined by the code in the DNA. Each cell is composed of various organic compounds varying in their chemical composition. An interaction between these varyingly composed cells brings about different bio-chemical reactions; the reactions in turn change the state of the cells or form new cells, the new ones being able to respond differently to the system. The process continues and at some point during the course, simple cell structures (cell barriers) are formed. These structures or barriers then work collectively to form the different organs of the human body, hence finally giving it full form and shape. It has to be noted that with growing complexity of the body, the cellular structures also get hierarchically specialized in their action. All this might give the impression that human life is highly deterministic. However, any introduction of a foreign agent, say a virus, will contribute to the indeterministic behavior of the system. In fact we will be dealing with two different behavioral systems then, and predicting the interaction in such a case is not always possible.
Artificial life is implicitly based on the theory of evolution. Chris Langton, one of the founders of artificial life, defines it as the study of the dynamics underlying biological phenomena by abstracting and then implementing them on other physical media, like the computer. The computational power of a computer enables us to manipulate it too, thus helping us find variations for it.
“Artificial life is about studying emergent objects and the evolutionary complexities that arise out of such objects.”
More specifically, artificial life is about emergent properties. It attempts to see how elaborate properties can be explained with simple rules of evolution. The beauty lies in the fact that none of the properties were built into the system. They just evolved out of a simple set of rules. Craig Reynolds’ boids are a good example in this context. Reynolds created a flock of virtual birds, flying such that each bird in the flock avoids collision with nearby mates, tries to move in the average direction of the flock, and attempts to stay near its mates. The boids were then observed to fly in a single group, dissevering into subgroups whenever any obstacle appeared in front of them, and then regrouping again after clearing it. This flocking behavior is a higher level phenomena emerging (meaning not controlled) out of vicinal interactions between relatively lower levels of fundamentally simple elements. Another example is Tom Ray’s Tierra – a virtual world created on a computer system, comprising of one single self-replicating entity (program) called the “ancestor” or the “80”. The Tierran system then started off. The results were astonishing. The Tierran system had evolved into an entire ecosystem where multiple variations of the ancestor were fighting for virtual world resources like memory and CPU time.
The question that automatically follows concerns the need for carrying out such experiments. Let’s try to answer this question. We have seen that in all kinds of artificial life experiments, the environment plays a very significant role. In Conway’s Game of Life, if the quadratic cells are replaced by hexagonal ones then the behavior changes. Reynolds’ boids won’t be of much interest if they are not allowed to move in groups. The Tierran system won’t evolve if there are abundant resources available to its creatures. The configuration and confinement of the environment is what gives them a dynamic nature. Moreover, a proper balancing of the rules is also required. Or otherwise, a population - group of cells - will either die out soon or go into chaos. Equilibrium is desirous, so that the fittest ones survives and breeds, while the unfit ones die out. But we cannot manipulate Nature (our environment), thereby restricting the domain under which we carry our real-life experiments. However, although it is not possible to completely duplicate a real-life environment, still we can manage it within the purview of our problem. The problem, for example, may be a simple function to optimize. The bounds of the variables in it define the environment. And the equilibrium is the minimum or maximum value of the function. On a broader scale, artificial life experiments help us understand complex systems from the simpler perspective of evolution. They will let us see what happens if the rules are slightly changed (mutation), thereby assisting us in improving the overall performance of the system. The mystery behind the origin of life lies in a simple self-replicating system. So, it’s only a question of finding how such a system evolved in the first place.
In the next section we will see how Life can be used as a test bed for artificial life experiments.
Tags: none
Category: essay |
