Readings Index
 

 

After Humans: Speculations on Artificial Life

A Review of Steven Levy's Artificial Life: A Report from the Frontier Where Computers Meet Biology

 

Mary Shelley’s Monster

In the early years of the last century, scientists working with the newly discovered phenomenon of galvanism (i.e., electricity) found that electrical impulses could cause the limbs of dead animals to twitch. This discovery led the nineteen-year-old Mary Shelley to write what has become the archetypal tale about the creation of artificial life, Frankenstein, or the Modern Prometheus. Though strikingly original in its use of contemporary scientific ideas, Shelley’s novel nonetheless drew on a theme at least as old as Genesis—the danger of seeking forbidden knowledge. Adam and Eve, who ate of the forbidden Tree of Knowledge; Prometheus, who stole fire from the gods and gave it to humans; the sorcerer’s apprentice, who brought disaster on himself by trying out his master’s wizardry; Faust, who sold his soul to the devil to become the greatest of scholars—these are but a few of the precursors of Shelley’s protagonist, Victor Frankenstein, whose cautionary tale warns us against the dangers of taking the creation of life into our own hands. 

Today, Shelley’s story seems remarkably prescient, for the fabled “secret of life” that Victor Frankenstein sought is now clearly understood. The key was the discovery by James Watson and Francis Crick of the DNA molecule, which carries the genetic code for producing an organism and is capable of self-replication with variations (due to mutation and exchange of information between the DNA of sex cells). These variations provide the raw material for natural selection, making it possible for organisms, over time, to change, to speciate, to develop to the point where they can discover the secret to their own development. 

Knowledge of the workings of DNA has led to amazing discoveries. Journalists around the world recalled the tale of Victor Frankenstein when they wrote their stories about the successful cloning of a sheep named Dolly by a researcher in Scotland. If a large, complex mammal like a sheep could be cloned, would humans be far behind? And didn’t cloning mean that people, in the mode of Victor Frankenstein, were taking the creation of life into their own hands? Well, not really. 

Cloning isn’t a technology to get very excited about. Plant breeders have been doing it for many years, and cloning simply results in the birth of an identical twin, something that occurs naturally anyway. It’s ironic that the cloning story should have received such attention from the media when other technologies for the creation of really new forms of life exist. Two that deserve the kind of media attention that cloning received are the possibility of recombination of parts of the Human Genome and the new computer science of artificial life, or AL. Both have dramatic implications for the future of life on earth. 

The Genetic Engineering of People

The largest scientific undertaking in the history of our species, now underway, is the Human Genome Project, funded by the United States government and being carried out by dozens of labs around the country. The object of this project is to sequence the entire human genome, the genetic blueprint carried by the DNA of human beings. Soon, this complete blueprint will reside in the memory banks of a computer at the National Institutes of Health (NIH) in Bethesda, Maryland. Recombinant DNA techniques already perfected allow scientists to snip out parts of the DNA of a cell and replace them with other parts. The rationale for the research now underway is that by mapping the Human Genome and then studying its parts, we shall be able to identify the codes implicated in the creation of genetically determined diseases, such as cystic fibrosis and breast cancer, and do gene therapy to replace these strings of code in sex cells so that children will not be born with such genetic defects. The potential exists, however, for current technologies to be used for more ambitious undertakings, such as creating designer people. The real scoop in the arena of genetics is that human beings are close to being able to take evolution of the species into their own hands. While nearly everyone can agree that eliminating genetic diseases is a laudable goal, what about such other goals as improving skin tone or visual acuity or memory? The potential of recombinant DNA research is infinitely more profound than that of cloning, for soon our species will have to confront the question of whether we wish to genetically engineer people to our specifications. Unfortunately, whatever people can do, they have a tendency to try out. It is doubtful that any United Nations resolution against the genetic engineering of people would be effective in keeping such a thing from occurring. 

The Triumph of the Mechanical View

There was a time, before the discovery of natural selection and of its mechanism, DNA, when one could easily hold to a view of life known as vitalism, which posited some extramaterial force, some elan vital, that set life apart from nonlife, that made possible such characteristics of living things as movement, metabolism, growth, and reproduction. Theologies around the world posited such a life force, or spirit, often identified with the breath (an identification that survives in our word inspiration). In the story of Genesis, Jehovah shaped a man of clay and breathed into this figure the breath of life. To the Hindus, the life force was likewise prana, or breath. Aristotle and Descartes both posited an immaterial soul that animated the body. To the galvanist, this life force was electricity. To New Age faith healers and teachers it is some mysterious “energy” that can be manipulated by meditation or the laying on of hands. Nowadays, it is difficult to find scientists who hold to such a view. The standard scientific view today is that living organisms are complex machines, operating according to clearly understood physical principles. This view is clearly expressed by philosopher Daniel Dennett in his recent book Darwin’s Dangerous Idea: 
    Darwin was offering a skeptical world . . . a scheme for creating Design out of Chaos without the aid of mind. . . . Darwin had discovered the power of an algorithm. An algorithm is a certain sort of formal process that can be counted on—logically—to yield a certain sort of result whenever it is “run” or instantiated. . . . 

    What Darwin discovered was not really one algorithm but, rather, a large class of related algorithms that he had no clear way to distinguish. We can now reformulate his fundamental idea as follows: 

    Life on Earth has been generated over billions of years in a single branching tree—the Tree of Life—by one algorithmic process or another. (50-51)

Most scientists today would agree that life is completely explainable as a set of emergent properties of complex physical systems that are themselves ultimately explainable in terms of the laws of physics and chemistry. (An emergent property is one that appears as the unpredictable result of the complex interactions of parts that themselves obey simple rules or laws. More about this later.) This view, known as mechanism or reductionism, was given a boost by the emergence of modern genetics, which explains the mechanical basis of the previous mysterious process of reproduction. The mechanical view is the underpinning of the new science of artificial life, described in Steven Levy’s Artificial Life: A Report from the Frontier Where Computers Meet Biology. 

What Is Artificial Life?

Practitioners of the science of artificial intelligence distinguish between two views of what they do: so-called “weak AI,” in which the scientists’ work is held to be about simulating intelligence in computers, and “strong AI,” in which the work is held to be about actually creating intelligent computers. Similarly, Levy draws a distinction, early in his book, between weak and strong artificial life: 
    Artificial life, or a-life, is devoted to the creation and study of lifelike organisms and systems built by humans. The stuff of this life is nonorganic matter, and its essence is information: computers are the kilns from which these new organisms emerge. Just as medical scientists have managed to tinker with life’s mechanisms in vitro, the biologists and computer scientists of a-life hope to create life in silico. 

    The degree to which this resembles real, “wet” life varies; many experimenters admit freely that their laboratory creations are simply simulations of aspects of life. The goal of these practitioners of “weak” a-life is to illuminate and understand more clearly the life that exists on earth and possibly elsewhere. . . . 

    The boldest practitioners of this science engage in “strong” a-life. They look toward the long-term development of actual living organisms whose essence is information. These creatures may be embodied in corporeal form—a-life robots—or they may live within a computer. Whichever, these creations, as [physicist James Doyne] Farmer insisted, are intended to be “live under every reasonable definition of the word”—as much as bacteria, plants, animals, and human beings. (5-6)

Levy traces the origins of the science of artificial life to John Von Neumann, the brilliant mathematician and physicist who, among his other accomplishments, invented game theory and the stored program computer, made seminal contributions to quantum theory, and helped to develop the atomic bomb. Shortly before his death in 1955, Von Neumann conceived of an abstract mathematical machine, a cellular automaton, or collection of checkerboard squares, each of which could switch between various states.  Following a simple set of rules, this automaton was capable of generating a copy of itself that in turn contained the blueprint for generating another copy of itself, ad infinitum. The Von Neumann cellular automaton could not only reproduce itself but also was a version of the “universal computer” first envisioned by Alan Turing. In other words, it was capable of mimicking the operations of any other conceivable computing machine. Then, in the late 1960s, a British mathematician named John Conway created a vastly simplified version of the Von Neumann cellular automaton, now known as the Game of Life. 

Conway’s Game of Life

Von Neumann’s cellular automaton was very complex, containing cells that could be in any of twenty-nine separate states. Conway envisioned a much simpler version. Conway’s rules for the Game of Life were quite simple: 
    1. Any given cell on the checkerboard is either alive (on) or dead (off). 
    2. If a cell is alive, it will continue to be alive in the next iteration, or generation, if and only if it has either two or three neighbors that are also alive. 
    3. If a cell is dead, it will continue to be dead in the next iteration, or generation, unless exactly three of its neighbors are alive, in which case it will be born.
Delightfully, these simple rules gave rise, on the life game board (which was envisioned as being infinite in all directions), to an amazing variety of emergent forms, including blinkers (objects that oscillated between two states), gliders (objects that reproduced themselves after a series of transitional states and appeared to move across the board), and guns that produced gliders in a periodic stream. Examples of Conway’s Game of Life can be found at  (Before reading the rest of this article, it would be a good idea to visit one or two of these sites to find out more about how Conway's system works.) “By using glider streams to represent bits,” Levy explains, “Conway was able to produce the equivalent of and-gates, or-gates, and not-gates, as well as an analogue of a computer’s internal storage (Levy, 57).” The Game of Life, as simple as its rules were, could be made to emulate a universal computer, or Turing machine. 

The Flocking of Boids and Other Emergent Behaviors

One of the most interesting aspects of the Game of Life and other cellular automata is that from simple rules, very complex phenomena emerge, phenomena that could not otherwise be predicted. An example of such emergence, discussed by Levy, is Craig Reynolds’s Boids (a variant of “Birdoid”). Reynolds created onscreen, paper-airplane-shaped objects and a few simple rules governing their behavior—moving away if they came within a certain distance, moving closer and aligning if they got too far apart, and so on. Based on these rules, Reynolds’s onscreen Boids produce remarkably lifelike flocking behavior. Although each individual Boid is obeying simple rules that apply only to itself, these rules give rise to what looks like concerted group action. For example, just like real birds, a flock of Reynolds’ Boids, when encountering an obstacle, splits up, moves around it, and congregates again. Reynolds’ Boids paradigm has been used to model, on computers, the flocking of birds, the schooling of fish, and the movements of insects such as virtual ants, or vants. For examples, see  Emergence is one explanation that artificial life researchers give for how living creatures can do the astonishing things that they do while being based upon simple underlying laws. The most dramatic of all examples of emergence is, of course, evolution. From simple beginnings, manifold forms develop by the operation of the simple law of survival of the fittest. 

Growing Programs by Means of Genetic Algorithms

One of the most fascinating subjects treated in Levy’s book is that of genetic algorithms. This approach to programming, pioneered by John Henry Holland of the University of Michigan, uses evolutionary methods to develop useful code. Randomly generated strings of ones and zeros are subjected to fitness tests. The fittest of these strings (the ones closest to a useful organization for accomplishing a given purpose, such as sorting a group of numbers from highest to lowest) are allowed to reproduce, swapping bits of their code with those of other fit strings to produce a second generation. At times, mutations are introduced, random swapping of ones and zeros in the code. By these means, over thousands of generations, usual programs can be developed. As Levy explains, 
    It seemed an almost absurdly simple recipe for optimization: take a string of random numbers and treat them as computer programs. Grade them according to how well they do at executing the work of a custom-designed computer program, and then reward them to the extent of their excellence by allowing them to reproduce to that degree. Then take the revised population, pair the strings, and have each marriage partner swap a part of itself with a mate. Change a few bits for mutation, and do it again. One would intuitively expect this process to take a very long time to match the results of a computer program specially written for a task—in fact, it might be difficult to envision something that good ever resulting from this elementary process. . . .
    But computer muscle telescoped millions of generations worth of evolution into a lunch hour, and the GA turned out to be a stunningly powerful tool. Indeed, it seemed to deliver on Holland’s original perceptions of the benefits of evolution: “perpetual novelty” and “something out of nothing.” (164-65)
Subsequent developments of the concept of genetic algorithms treated not only ones and zeros but parts of programs in higher-level languages. Others, such as the Tierra program developed by Thomas Ray (see http://www.talkorigins.org/faqs/tierra.html) go beyond the imposition of artificial fitness tests to natural selection based on environmental conditions; in Tierra’s case, competition among bits of code for processor time and memory allocation. One striking possibility raised by the use of genetic algorithms is the existence of self-organized programs whose workings are completely mysterious to the programmers who first set them into motion. 

The Future of Artificial Life

The science of artificial life is in its infancy, but already it has produced astonishing results. When researching his book, Levy interviewed many of the most famous figures in computer science—Rodney Brooks, Fred Cohen, Doyne Farmer, Ed Fredkin, Danny Hillis, John Holland, Chris Langton, Pattie Maes, Marvin Minsky, Thomas Ray, Mitchel Resnick, Craig Reynolds, Alvy Ray Smith, and Stephen Wolfram, among others. Many of these people look to the day when creatures created in silico have evolved to such a point that human beings have become, in the words of one NASA report, “just an example of the generic class ‘intelligent creatures’” (quoted in Levy, 346). Some, like Marvin Minsky and Norman Packard, imagine that we are, in fact, creating our silicon successors. Victor Frankenstein created his creature out of body parts stolen from graves. Some people, at least, believe that creatures can be created out of ones and zeros. One thing is certain: we reached a turning point when we first developed programs that could evolve on their own. We already have at least one existence proof that evolutionary processes can create intelligent creatures: we have ourselves. 
 

References

Books

Dennett, Daniel. Darwin’s Dangerous Idea: Evolution and the Meanings of Life. New York: Simon and Schuster, 1995. 

Levy, Steven. Artificial Life: A Report from the Frontier Where Computers Meet Biology. New York: Random House, 1992. 

Artificial Life Sites of Interest Not Listed in the Article Above
 
 
Questions for Discussion and Review 

The following questions are based on the preceding text. Clicking on a question will take you to the place in the text where the question is discussed. To return to these questions, simply click the "Back" button in your browser. 

1. What makes Mary Shelley's novel seem, today, "remarkably prescient"? 

2. What two present-day technologies receive far less press than cloning did and yet are far more dramatic in their consequences? 

3. What is the difference between weak and strong AI? 

4. What is vitalism? How does it differ from reductionism? 

5. What is the science of artificial life? 

6. What is the difference between weak and strong Artificial Life? 

7. What is a cellular automaton? Who first conceived of the idea of a self-reproducting cellular automaton?  

8. Who is John Conway, and for what is he known? 

9. What is "emergent behavior," and in what way are Craig Reynolds's Boids a perfect example of such behavior? 

10. What is a "genetic algorithm," and how does such an algorithm lead to the development of programs not explicitly described by a programmer? 

11. What reason does this essay give for thinking that genetic algorithms might provide a means for the development of "intelligent" artificial life? 
 
 
 EMCParadigm Publishing