Artificial Life

For millennia humans have wondered what divides the living from the non-living. Artificial Life, a field fusing computers and biology, gives startling answers to that ancient question.

by Kai Wu

In the heavens as it is on Earth?
    When NASA first organized the Viking Lander missions to Mars, one of the most vexing problems its scientists faced was devising and agreeing upon tests to detect life. Besides imaging and chemical tests, a variety of experiments were loaded onto the spacecraft in the hopes of detecting the telltale signs of microbial life: growth, food consumption, and waste production. Despite every effort at biological impartiality, the tests still reflected a strong Earth bias: assumptions about what Martian microbes would eat, "breathe", or live in rested upon our own biochemical preferences for carbon and water. Ultimately, Viking obtained no conclusive data for or against life on Mars - an important reason for sending new probes to the enigmatic red planet in the near future.

    Whether through probes or radio astronomy, our search for life on Mars and elsewhere in the Universe carries with it substantial assumptions about the nature of life. Is carbon-based life the only viable form? Is multicellular life inevitable? Does "intelligence" necessarily imply a technological civilization like ours? Is life abundant or rare in the cosmos? If we don't receive a neatly coded radio message from benevolent aliens, and instead have to do the hard work of actual exploration, just how will we recognize life if it's there? What is life, really? This ultimate question eludes an easy definition because of our own earthbound limitations: without other forms of life to compare and look at, we're unable to characterize life in general. However, one possible solution may come from the seemingly foreign world of digital computers, in the form of artificial life (AL), the study of man-made systems that demonstrate life-like behavior. The surprising results of AL suggest that life is a dynamic process, independent of the material basis, arising from the complex flow and interaction of information. AL proposes that to realize with machines the spectacular complexity and diversity of biological life, the same forces of emergence and evolution must be harnessed.

Not quite human...
    Computers have brought unprecedented changes to science and society. Their theory and operation grew largely from the work of John von Neumann, a Hungarian mathematician whose genius led many of his colleagues to wonder if he was entirely human. Besides calculating skills that seemingly exceeded his own creations, von Neumann formulated most of the theories underlying all modern computers before the middle of the 20th century. One of his most famous legacies, the von Neumann machine, held that self-replicating computers were the key to intelligence and would ultimately constitute a new form of life. He believed that life was not a special arrangement of organic molecules but a process which exhibits certain behavior. If it could be understood and formulated as a process - in particular as a Finite State Machine (FSM), then a computer (which can in principle emulate any FSM) could exhibit lifelike behavior and ultimately be deemed as alive as any other organism.

    Is life really a FSM, i.e., at any given moment, given the environment and the internal state of an organism, are there a finite number of possible behaviors that the organism will exhibit in the next moment? The richness of biological systems and their behavior, from bacteria to trees to dolphins to humans, seems to defy any easy categorization. However, the genetic code of DNA is the foundation of all biological life. Viewing DNA at the more abstract level as information given by the 4 letter alphabet of Adenine, Cytosine, Thymine and Guanine, life can be seen as the processing of that information. Since there is a finite amount of DNA per organism, it follows that there is a finite (albeit staggeringly large) number of possible combinations that the products of DNA (proteins, enzymes, cell structures) can form. Life understood in these terms is a FSM, although it functions at a level of complexity greater than any contemporary computer. Some might balk at such a statement but at the very least, life can be said to be an emergent process: from the relatively simple basis of DNA, certain chemical rules give rise to very complex results, such as ourselves.

    In the late 1940's von Neumann foreshadowed the discovery of DNA in stating that any organism had to have at minimum the instructions dictating its behavior and reproduction, along with a copy in some form of those same instructions passed on to descendants. In focusing on reproduction, von Neumann took a very different approach towards artificial intelligence (AI) than most other computer scientists, who worked towards understanding and emulating the human mind itself. Giving machines the ability to replicate and mutate allowed for further complexity and would unleash the powerful and unpredictable forces of evolution. Paralleling biological evolution on Earth, relatively simple machines would evolve, perhaps even directing their own mutations, ultimately leading to far more complex machines unrecognizable to their original human creators. Von Neumann machines have been used in many science fiction stories: for instance, in Gregory Benford's hostile universe, von Neumann probes wreak havoc in the galaxy as they replicate and destroy competing forms of life. A more optimistic use for von Neumann machines (actually studied by NASA in the early '80s) proposes a single "space-seed" self-replicating factory that could be sent to another world, where upon landing it would collect raw materials and build more copies of itself to accomplish some enormous engineering project. With exponential efficiency, a job that might take several millennia, such as terraforming or atmospheric conversion, could be finished in mere decades.

The ultimate game of checkers...
    Von Neumann invested extensive effort to make his dreams of automatons, or reproducing machines, come true. He postulated that an automaton would need at least six elements: a computer, a sensor, a manipulator, a cutter, a fuser, and "girders" from which the automaton was built and which encoded instructions to act as "memory". Although the basics of such a machine sound simple enough and are possible in principle, the technology required is still beyond us, nearly half a century later. Current robots, for example, still have a difficult time maneuvering in complicated real-world environments, and endowing machines with a sense of sight we take for granted has proved formidable. However, von Neumann again had a brilliant insight: although physically building an automaton capable of reproducing was too demanding, creating its equivalent in the virtual world of a computer was within reach. Thus was the idea of cellular automata born.

    For example, imagine the universe existed purely as a 2-dimensional grid, with an infinite number of cells, i.e. an infinite checkerboard. Each cell could have a certain state, and there exists rules for the relations between cells that dictate if the cell should remain in its state or change to another state. Von Neumann devised the rules in order that a computer could be composed of appropriately arranged cells, along with other cellular elements that would enable the whole entity to build a copy of itself in the cellular universe. Such entities and systems were dubbed cellular automata (CA). Again, the degree of complexity required was great, involving hundreds of thousands of cells, but given the rapidly growing power of computers, it was only a matter of time before a logically (as opposed to physically) replicating artificial entity appeared.

    In 1968 John Conway, an Oxford professor, decided to test if CAs really could form the basis of computers and more complex things, and in particular, to do it in a far simpler manner than von Neumann proposed. He invented the now famous game of "Life", which exhibited dazzling complexity from deceptively simple rules. Using a finite grid, each cell in the grid is said to be either alive (occupied) or dead (empty). Some starting pattern is laid down, and then the rules of Life are applied. Each cell has 8 neighboring cells. If a live cell has no living neighbors or only one, it dies from isolation. If it has 2 or 3, it continues to be alive, and if more than 3, it dies from overcrowding. If a dead cell has exactly 3 live neighbors, it becomes alive. A generation is complete once each cell has been checked according to the preceding criteria, then a new generation is begun and each cell is checked again. What Conway and other enthusiastic computer hackers found was that even such a simple universe as Life could indeed exhibit sufficiently complex behavior to form computers, and possibly more. Playing with various Life patterns became a full time obsession for Conway and others. Some patterns rapidly broke up into a chaotic mess, whereupon mass death ensued. Others oscillated over tens or hundreds or even thousands of generations between 2 or more patterns. Some, called Gliders, had the power to preserve their shape while moving slowly across the board. Larger starting patterns and larger grids or universes allowed for mysterious and astounding complexity as enormous Life patterns grew, interacted, and died with their fellows. Along with a growing fascination with Life and CAs for their own sake, the original quest of computing ability was not forgotten. For instance, there was one Life pattern that oscillated between two different states when a Glider happened to hit it; a clear form of binary memory. Further subunits made up of Life cells that could add or act as logical gates were also found, and a simple computer was deemed possible. Larger questions were raised as to the possibilities of a Life entity that could reproduce itself; indeed, how big a Life board would it take for complex, lifelike behavior and perhaps intelligence to emerge? Some Life adherents believed that the universe itself was a giant CA system, operating by some as yet unfathomable rules.

The data must flow...
    With the explosion of interest in CAs generated by the popularity of Life, CAs became seen not merely as interesting systems in themselves, but as directly connected to the physical world. However, outside the enthusiastic inner core of CA pioneers, few computer scientists, let alone biologists, took CAs seriously. One of the people willing to rigorously study CA phenomena was Stephen Wolfram, creator of the now famous Mathematica program. Wolfram was interested in complex systems in general, and saw CAs in particular as a key to understanding them. The Life game was an example; Wolfram also devised his own 1-dimensional CA which, despite using even simpler rules than Life, also developed into intricate and diverse patterns. With the simpler 1-dimensional case, Wolfram was better able to explore all possible combination of rules for his CA. He devised a classification system for the types of CA universes he observed: Class 1 CAs quickly degenerated into stagnant patterns of all alive or all dead states; Class 2 CAs were rigidly stable and also uninteresting; Class 3 CAs showed no patterns and were random messes. The final Class 4 CAs were the complex yet organized patterns that seemed closest to being alive. Later, Christopher Langton, another pioneer in the fledgling field of AL, showed that the class of universe observed depended highly on the ease or difficulty with which information flowed and was retained. Life or lifelike behavior seemed to manifest itself most strongly in the virtual universes were information neither flowed too freely (random Class 3) nor too rigidly (dull Class 1 and 2). Its abode lay in the dynamic balance of retention and transfer of Class 4 universes. It was an insight that further reinforced the idea of understanding life as a processing of information, independent of the particular material basis.

All for one, and one for all...
    Cellular automata had demonstrated a key feature of biological life, complexity. The strange, intricate patterns arising on computer screens and on playing boards all arose from applying a set of simple rules to simple elements, but the collective result that emerged far surpassed any single element in complexity. Thus CA, and many other systems in nature, posses emergent properties: from starting simplicity one can have an incredibly complex, collective result. The behavior of the system as a whole is unpredictable from examination of its constituent parts alone, an idea that may be everywhere in the universe.

    Seeing any large group of social animals in action is a marvel; the collective behavior of flocks of birds, of colonies of ants, of herds of zebras, or of schools of fish, defy simple rules. But what of the individuals within the group? In the case of flocking birds, ornithologists had long thought that the acrobatic, cohesive and complex movement of a flock was the result of each bird following some equally complex set of rules. Computer animator Craig Reynolds, however, suspected that the collectively complex movement of a flock was an emergent property of each bird following a simple rule-set. His computer simulation had birdlike entities, named Boids, follow 3 simple rules: maintain a minimum distance from fellow Boids and other objects, match velocity with neighboring Boids, and move toward the perceived center of mass of nearby Boids. When applied to a group of Boids, such simple rules gave rise to startlingly familiar behavior: Boid flocks moved about a computer screen with the familiar grace of their organic cousins. Success was achieved in similar attempts to mimic the collective behavior of ants and fish; by having each individual follow a set of simple rules, the emergent collective behavior was remarkably similar to what was seen in the natural world. Besides the convincing demonstration of emergent properties, Reynold's Boids and other simulations reinforced the idea that the forces of biology could indeed be duplicated on computers. Although a Boid was far from a bird, the behavior of a flock of Boids was comparable to a natural flock; different elements of a different basis, once given appropriate rules, had duplicated a natural phenomena.

    Nature is rife with other examples of emergent properties, of mind-boggling complexity from starting simplicity. For instance, human intelligence seems a likely candidate as an emergent process. Individual neurons are relatively simple and behave much like discrete switches, firing chemical signals in an "all or nothing" fashion depending upon the particular conditions. However, with 10 trillion neurons connected in some fashion that we've yet to understand, something quite unexpected of mere switches appears: consciousness, along with language, abstract thought, culture, technology, and the national deficit. The organization of neurons in brains has inspired the study of neural networks and parallel processing in computer science, which thus far have had mixed success, but teach an important lesson. Merely having a sufficient number of switching, interconnected elements is as unlikely to give rise to intelligence as a few million beer cans strung together will produce passable music. For highly ordered complexity such as life, the emergent properties we observe are due to the forces of self-organization, and evolution.

Nature's free lunch...
    In nature very few systems lend themselves to the elegant but simplified equations and models of the introductory science classroom. Almost any phenomena of interest is going to be dauntingly complex due to the enormous number of atoms involved, feedback between the elements of the system, the complicated interactions of many forces, and so on. Yet despite the bewildering amount of variables, the results are often not chaotic, random jumbles but systems of marvelous order, from the whirling of the planets and galaxies, to the humblest bacteria, and thence to humans and their societies. How does such order arise? For life, besides evolution, the answer may lie in a mysterious force of self-organization that drives systems towards greater order, just as the 2nd Law of Thermodynamics drives systems towards entropy and disorder. Life's existence is a dramatic instance of this force at work; any living system maintains or increases its internal order through drawing energy (sunlight, food) from the environment, and is thus a local violation of the law of entropy. Life's origin may be an example of this force as well.

    Despite great strides in molecular biology and our knowledge of early life, the actual origin of life remains a scientific mystery. The Miller-Urey experiment, which simulated the early Earth's environment, showed the ease with which organic molecules could form and encouraged biologists who worried that life seemed an extremely unlikely event at best. The difficulties that arose later were far more daunting than the generation of amino acids, however. Analogous with the classic dilemma of the chicken or the egg, it was not known how DNA or RNA could have arose without the right catalyzing enzymes, and vice versa. The chance of "naked" RNA forming by itself in the environment seemed too remote and the chain of chemical events needed to create a self-replicating molecule seemed too improbable. But Stuart Kauffman, an origin of life theorist and experimenter, came to see life's origin as not only likely but inevitable, given the power of self-organization. His computer network experiments had demonstrated this force at work: 100 elements, each connected to 2 others, and each subject to random logical rules, allowed for a huge number (2 to the 100th power) of possible states for the system as a whole to be in. Surprisingly, instead of progressing from one random state to another, Kauffman found that the system went into a 4 state loop after only 10 cycles. Changing the rules had little effect, as the system again quickly settled into a cycle. In chaos theory such a phenomenon is known as a periodic attractor, a preferred state that a system will "settle down in" despite an enormous number of alternative states. To Kauffman it indicated that in systems of interconnected elements, networks, a powerful force of self-organization tended to drive them towards order and stability.

    Kauffman's theory for the origin of life was inspired largely by the surprising results of his network experiments, with the unexplored power of self-organization playing a pivotal role. Like physicist Freeman Dyson's, his theory emphasized the coupled processes of replication and metabolism; feedback reactions between the two gave rise to life. In the prebiotic soup of early Earth, single amino acids and other organic molecules, monomers, linked up to form simple polymers. Some of these polymers would have a catalyzing effect to form further varieties of polymers, perhaps longer and more complex ones, which in turn could catalyze further reactions and polymers, and so on. The number of possible combinations would grow at an incredible rate, but Kauffman was confident that the same force of self-organization witnessed in his computer networks and elsewhere would have taken hold. At some critical point of complexity, everything would catalyze something else in the system, with the collective result of metabolism for the whole; an autocatalytic, chemical network performing a crucial function of life. Combine it with an equally rich network of reactions responsible for replication, add the same self-organization force, and suddenly the possibilities for life's origin seem far better. Unnatural selection...
    Like Newton's laws of motion and Maxwell's equations of electromagnetism in physics, Darwin's theory of evolution through natural selection served as the great unifying principle of biology. With evolution, the immense variety, patterns, and relationships of the living world could at last begun to be understood. Evolution as natural selection through descent with modification served as the most basic definition from which AL workers began to emulate biology's central maxim on computers.

    Christopher Langton took a decidedly different approach with his CAs than Conway or Wolfram. Langton skipped the idea of making CAs into universal computers and instead chose to more closely parallel biology by creating a computer entity able to reproduce itself. His CA, the Langton Loop, contained the minimum requirements postulated by von Neumann: information to be copied from generation to generation (the genotype, found as DNA in biological life) and the instructions, giving rise to characteristics and behavior, derived from that information (the phenotype). A series of core cells served as the genotype, and through the particular rules of Langton's CA universe, they dictated the "growth" of sheath cells surrounding the core as well as the reproduction of further Loops. In time, Langtonšs CA universe rapidly filled up with a central "dead" core of Loops no longer reproducing for lack of cell-space, and an active, growing boundary layer, a situation reminiscent of the growth of coral reefs. Langton Loops would come to be recognized as the simplest self-replicating systems known. The next step would be to impose a fitness criteria. Along with random mutations thrown into the genotypes, the power of evolution would be unleashed.

    In nature the fitness of an organism depends on how well adapted it is to its environment: how well it can find food, escape from predators, nurture its young, etc. In emulating life on computers, the criteria for fitness of artificial lifeforms was dictated by the programmers, leading to a process of unnatural selection. An example of the power of evolutionary techniques applied to computer programs is the genetic algorithm (GA), a bottom-up approach to programming developed by John Holland, the nation's first Ph.D. in computer science, during the dawn of computing in the '50s. Analogous with biology, binary information and instructions replace DNA and chromosomes, and the fitness criteria is arbitrary. A starting collection of completely random strings of code, or later, random combinations of actual computer subroutines, were evaluated on their ability to perform some task. Those who got the best results were most likely to be included in the next generation, but before another fitness test was performed, the most successful of these digital entities traded some of their code or genes with one another, just like crossover or sex in biology. Point mutations were also emulated by having a few random flippings from 1 to 0, or vice versa, per entity. After all this swapping of digital information, the new population was again evaluated according to the programmer's fitness criteria. The departure from nature lay in speed: each generation on a computer lasted only fractions of a second, thus a progression that might take millions of years in nature would require but a few hours on a computer. Also, the final desired behavior or product was far simpler than the incredible complexity one finds in even the humblest bacteria. Even so, genetic algorithms were surprisingly and enormously successful at their tasks. For example, from starting random code, virtual "ants" in a UCLA project, each a mere 450 bits worth of computer code, evolved in less than a hundred generations to well-adapted, flexible digital entities. They were able to navigate within their artificial computer environment, containing simulated obstacles, traps and sought-after food with clever tricks and routines that human programmers would have found difficult to predict or match. Equally intriguing was the discovery that crossover played a much more significant role than mutations: turning off mutations in a system had little effect on the rate at which the GAs adapted themselves, but the crossover function was critical to successfully evolving code.

    The analogous conclusion for the biological world ran contrary to what most biologists believed, that mutation was the driving force of evolution. They were loath to cede that any computer mirroring of nature could provide anything more than a curious simulation, let alone a fundamental insight into their field. Computer scientists, for the most part, also reacted negatively at first to the success of GAs, which with their aura of randomness and "something from nothing" quality threatened their traditional method of deliberate, systematic programming. The establishments were slow to accept that the same process that gave rise to their own intricate complexity and adaptation could do the same for programs.

Infinite diversity in infinite combinations...
    Artificial evolution applied to computer graphics is another dramatic example of the power of using the evolutionary process. Richard Dawkins, a leading figure in the field of evolutionary biology, was an early experimenter in this field with his "biomorphs". From a starting simple stick figure with nine parameters or genes controlling such factors as symmetry and branching, each biomorph would be cycled through "generations", its growth dictated by its genotype, which was also mutating. The human operator chose whichever biomorph caught her or his fancy and guided the unnatural selection process by allowing favorites to continue growth in the next generation. The results surprised Dawkins most, who had merely expected increasingly complicated treelike stick figures to emerge. Instead, an incredible variety of shapes resembling insects, reptiles, plants, shells, snowflakes, letters, faces, and just about anything one could imagine appeared on the screen after a few dozen generations, from the humble beginning of a single pixel. Karl Sims, a computer artist, would later build upon Dawkins' idea to create his spectacular and haunting evolutionary images: artificial landscapes, plants, textures, and other strange unnameables were generated from starting randomness and rapidly evolved to a desired end product. With the programmer's eye guiding the process of unnatural selection, one could literally search through the entire space of all possible images, with lightning speed.

Digital Yin and Yang
    Combining artificial evolution with massively parallel computers not only magnified the power of the technique but provided further insights into the workings of evolution in nature. Danny Hillis' Connection Machine, with its 65,536 connected processors (typical computers merely have 1 processor), could simulate the growth and development of a very large number of artificial organisms and their interactions with each other at fantastically high speeds. In an early experiment, Hillis started with 65,536 randomized digital "organisms" dubbed "Ramps", one per processor, and threw them at the task of sorting a set of sixteen integers in descending order in the fewest number of exchanges possible. Each Ramp was a small program that was to compete with the ingenuity of the best human programmers; Milton Green held the record with sixty exchanges in his program. Like Holland's GAs, the Ramps were evaluated at every generation on their ability to sort the set of integers; the most successful could breed with one another (Hillis gave them an ability to choose) and survived to the next generation, where the fitness test was repeated. Hillis later tossed in artificial "parasites" who also evolved and continually challenged the Ramps to greater efficiency, quickening the pace of their evolution considerably. Eventually, after hundreds of thousands of generations, the Ramps managed to sort the integer-set in sixty-one exchanges. Their highly evolved, optimized code contained many subtle tricks that would have impressed a human programmer, in addition to being more robust and flexible than the highly specific and relatively fragile program of a human. Many AL researchers feel that the future of programming lay with the evolutionary approach of GAs: if the problem could be defined well enough, one could ignore the tedious details in programming a solution and instead evolve flexible, efficient and robust code superior to perhaps any human product.

    The coevolution of Ramps and parasites provided many insights into the evolutionary process in general. The advantage of simulating a process of the natural world on a computer lay in the operator being able to examine every detail of the system, surpassing even the evolutionary biologist's dream of having the sequenced genome of every extinct animal along with the complete fossil record. Not only did Hillis find the coevolution-evolution cycle of host and parasite to be more important than tracking the evolution of a single species, but his system also put a new twist on the old evolutionary debate of gradualism versus punctuated equilibrium. Gradualism maintained that evolutionary changes were small but constant between generations of organisms; punctuated equilibrium proposed that evolutionary changes came suddenly, spurred on by large environmental disturbances. Hillis found that, despite the appearances of punctuated equilibrium in that the Ramps typically stayed at evolutionary plateaus for some time before being forced to further adapt, underneath the calm veneer of stability the steady exchange and interactions of their digital genes prepared them for the next surge of adaptation. Although most biologists were far from convinced that a computer model could tell them anything about the real world, a growing number of them appreciated the new insights given. Until much more sophisticated tools become available to study large communities of organisms in the field, increasingly sophisticated computer models may be among the best means to understand evolution's workings. But what would be needed to move beyond models, and begin to cross the boundary between living and nonliving on a computer? The answer lay in open-ended evolution.

Other realities, other worlds, other beings...
    As the sphere of AL expanded, some began to wonder at the further tantalizing possibilities of evolution on computers. In particular, how could complexity rivaling that seen in biological life be realized? Biologist Thomas Ray saw that the unnatural selection process of previous work was far too limiting. He developed an artificial environment, named Tierra, where his digital organisms were to vie with one another for computer resources (memory, processing cycles), so that their fitness criteria resembled that of nature's more than a programmer's fancy. Each artificial creature was again a collection of computer instructions, initially with the simple behavior of replicating itself; each had a digital genotype and phenotype. The world of Tierra welcomed the possibility of mutation and was more forgiving of it as well; mutations were less likely to be fatal than in the real world. On January 3 1990, Tierra received its first visitor, a creature eighty instructions long that Ray named the "Ancestor". The Ancestor's descendants rapidly propagated throughout Tierra, and much to Ray's shock and delight, Tierra soon bloomed with new "life".

    Like Hillis' Ramps and other GAs, the organisms of Tierra had many subtle, counter-intuitive coding methods that nonetheless were very efficient and complex. Mutants appeared that could replicate with seventy-nine instructions, then seventy-eight and even fewer. Parasites with only forty-five instructions soon followed, and like Hillis' system, an evolutionary tug-of-war broke out: as hosts developed defenses, parasites found new means of attack, and the war raged on. Later, "hyperparasites" arose that could attack the parasites through inspecting themselves for the marauders, and thence steal the parasite's replication process for their own growth. With the parasites driven to extinction, a cooperative cycle arose between groups of hyperparasites who relied on their neighbors for more efficient growth. Soon, a new breed of parasite appeared which took advantage of this cooperative cycle for its own ends, and so on.

    Such remarkably rich interactions were typical in Tierra, which generated fame not only for its creator but for AL as well. Even diehard biologists took note of it and similar systems, which so clearly illustrated the cycles and interactions of evolution they observed in nature, and also provided a way to test out some theories impractical to verify in the field. There was little question that the creatures of Tierra were lifelike; there was emergent and collective complexity, self-organization, and evolution. The question haunting every AL researcher was when the boundary would be crossed (some think it already has been), and life, by any reasonable definition, would appear on the computer screen. In the mean time, the field of open-ended evolutionary computer systems continues; as computer power grows exponentially, increasingly sophisticated organisms and environments will appear. At some point they may well surpass the complexity of a bacteria, then an insect, perhaps reaching the level of a human, and thence beyond...

The will to live...
    Some AL researchers argue that computer viruses already stand teetering on the boundary between the living and nonliving, like their organic cousins. These often malevolent and notorious computer programs satisfy many of the criteria for a living system: they reproduce and propagate themselves across computers, they divert their host computer's "metabolism" for themselves, and they could react to their environment to avoid detection and reproduce at opportune times. Some even rival real viruses in complexity of code (taking DNA as a 2-bit code giving A,C,T,G). The crucial missing quality is the ability to evolve in an open-ended fashion towards greater complexity, but many see that as a likely addition in the near future. It's a possibility dreaded by those in computer security; soon, any networked computer may become a battleground between evolving and competing computer viruses and anti-viruses.

    What of returning to von Neumann's prophecy of creating physically reproducing and evolving machines? Although robotics has begun to adopt increasingly adaptive and flexible "learning" systems more akin to living systems, the physical and engineering difficulties of having one large macroscopic machine build another remain formidable. One possibility for achieving this variety of "real" artificial life lies in nanotechnology. Like the scenario played out in an episode of Star Trek where "nanites", molecular scale self-replicating machines, are accidentally released aboard the Enterprise, the expected ability of nanotechnology (or molecular manufacturing) to build up systems from the "bottom-up", atom by atom, promise to make von Neumann machines a reality. Giving such machines the ability to mutate, exchange "genes", and evolve would then be trivial, with results just as unexpected and unpredictable as what befell the Enterprise, where the nanites grew in number to become a collective superintelligence.

    Many would like to believe that we can postpone worrying about AL indefinitely in that it will never be realized, but time and time again the unpredictable power of evolution, even within the confines of a desktop computer, have weakened the arguments of naysayers. Whether artificial life manifests itself on the digital plane of a computer or as a von Neumann machine in our world, accepting the possibility of their eventual creation brings about serious questions. What are they, and what will they become, working with evolution at a rate millions of times faster than ourselves, and with the ability to direct their evolution far more effectively than we ever could with biotechnology? Will they recognize us as creators, equals, pests, or as undesirable "biological filth"? Could they become everything we are, with emotions, creativity and culture, and go beyond? Could we impart to them our best qualities of compassion and wisdom? Would they then have our blessings as our successors? Once they are created, there will likely be little we can do to guide them; the same forces of evolution unleashing their diversity will almost certainly place them beyond our control as they seek their own genetic fitness. These and other possibilities have been raised by those in AL, who have begun to advocate open discussion on the subject. As Farmer and Berlin, two prominent AL researchers, have said, "With the advent of artificial life, we may be the first species to create its own successors." No matter what the eventual outcome, their emergence will be one of the most profound and far-reaching events in our planet's history.

Further information:
    There's a rapidly growing amount of literature on AL; the best introduction to the field is Artificial Life by Steven Levy, which served as the primary source of information for this article. It's very accessible, quick-paced, well-written and asks the deep questions. Also of note is Artificial Life I, a collection of papers submitted at the first AL conference in '87- more technical, but chock-full of great ideas. If you've got Internet access, check out the newsgroup comp.ai.alife. You can also subscribe to the AL mailing list: alife-request@cognet.ucla.edu

Author's Note:
    A junior studying physics in the College of Sciences & Arts, Kai Wu firmly agrees with Calvin and Hobbes that the surest sign for the existence of intelligent life in the universe is that we have not been contacted.