Get Alife! Artificial Life, That Is

Dateline: 05/25/97

CAN you be alive and not know it?

Sure you can. You might be asleep, or unconscious, or watching Married with Children (or any one of a distressingly large number of mindless TV shows worldwide). Or you might be an automobile.

Whoa! An automobile? Alive?? Well, between you and me, I didn’t think so either, until I was reminded by mathematical physicist Frank Tipler’s 1994 book The Physics of Immortality of something he and astronomer co-author John Barrow had pointed out nearly a decade earlier, in The Anthropic Cosmological Principle. Namely, that automobiles are alive. Coincidentally, Oxford biologist Richard Dawkins came (in effect, though not in so many words) to the same conclusion, in his equally famous book of the same year, The Blind Watchmaker.

Tipler explains: "They self-reproduce in automobile factories using human mechanics. Granted, their reproduction is not autonomous; they need a factory external to themselves. But so do male humans: to make a male baby, an external biochemical factory called a ‘womb’ is needed. Granted, their reproduction requires another living species. But so does the reproduction of the flowering plants: such plants use bees to pollinate and animals to disperse their seeds. Viruses require the entire machinery of a cell to reproduce. The form of automobiles in their environment is preserved by natural selection: there is a fierce struggle for existence between various ‘races’ of automobiles. Japanese and European automobiles are competing with native American automobiles for scarce resources—money for the manufacturer—that will result in either more American or more Japanese and European automobiles being built. By my definition of life, not only automobiles but all machines—in particular computers—are alive." (Emphasis added.)

Aha! Definition is key. (I should say, before proceeding, that Tipler has taken quite a bit of flak for The Physics of Immortality and for his definition of life. That he's far out, there is no question. But where else would one expect an Einstein-level mathematical physicist to be? Dawkins, too, has his detractors, and in fact I don't agree with everything he says, but if we lambast our greatest scientists for telling us frankly what's on their minds, we'll be the poorer for it.)

Definition of Life

"Life is a form of information processing," and "a ‘living being’ is any entity which codes information (in the physics sense of the word) with the information coded being preserved by natural selection," says Tipler.

Using this definition, it is easy to accept automobiles and computers—and programs that act like computers—as being alive. The definition makes no mention of carbon, on which all Earthly plants and animals and microbes are based. Any physical substrate will do, provided it can code information. Indeed, Scottish chemist Graham Cairns-Smith showed in the 1960s how mineral (non-carbon) substances such as clay could have been the first lifeforms to arise out of the primordial soup. The "pet rock" fad of yesteryear seems (marginally!) less silly.

Note that our definition also makes no mention of sentience or self-awareness.

The Study of Life

With a good definition of what to look for in life, it’s easier to study it. Definitions that insist on (or have simply assumed as self-evident) life’s being carbon-based lead naturally to a reductionist method of scientific study. That means we take things apart to see what makes them tick. With plants and animals, we cut them up and study the bits through microscopes and other instruments.

But with Tipler’s more modern definition we can study life from the opposite direction: We can literally create life, and watch it grow, reproduce, and evolve. And that’s essentially what the new scientific discipline—Artificial Life—is all about. It has its roots in biology and computer science, and represents one of those wonderful evolutionary events that occurs whenever two gene pools combine at the right time and place to produce something radical.

To expand on this theme, here’s a rather long quote (but it’s worth it) from The Santa Fe Institute (a private, independent, multidisciplinary research and education center founded in 1984 and devoted to "creating a new kind of scientific research community, pursuing emerging science"):-

"Biology is the scientific study of life—in principle anyway. In practice, biology is the scientific study of life on Earth based on carbon-chain chemistry. There is nothing in its charter that restricts biology to carbon-based life; it is simply that this is the only kind of life that has been available to study. Thus, theoretical biology has long faced the fundamental obstacle that it is impossible to derive general principles from single examples.

Without other examples, it is difficult to distinguish essential properties of life—properties that would be shared by any living system—from properties that may be incidental to life in principle, but which happen to be universal to life on Earth due solely to a combination of local historical accident and common genetic descent.

In order to derive general theories about life, we need an ensemble of instances to generalize over. Since it is quite unlikely that alien lifeforms will present themselves to us for study in the near future, our only option is to try to create alternative life-forms ourselves—Artificial Life—literally "life made by Man rather than by Nature."

Artificial Life ("AL" or "Alife") is the name given to a new discipline that studies "natural" life by attempting to recreate biological phenomena from scratch within computers and other "artificial" media. Alife complements the traditional analytic approach of traditional biology with a synthetic approach in which, rather than studying biological phenomena by taking apart living organisms to see how they work, one attempts to put together systems that behave like living organisms.

The process of synthesis has been an extremely important tool in many disciplines. Synthetic chemistry—the ability to put together new chemical compounds not found in nature—has not only contributed enormously to our theoretical understanding of chemical phenomena, but has also allowed us to fabricate new materials and chemicals that are of great practical use for industry and technology.

Artificial life amounts to the practice of "synthetic biology" and, by analogy with synthetic chemistry, the attempt to recreate biological phenomena in alternative media will result in not only better theoretical understanding of the phenomena under study, but also in practical applications of biological principles in the technology of computer hardware and software, mobile robots, spacecraft, medicine, nanotechnology, industrial fabrication and assembly, and other vital engineering projects.

By extending the horizons of empirical research in biology beyond the territory currently circumscribed by life-as-we-know-it, the study of Artificial Life gives us access to the domain of life-as-it-could-be, and it is within this vastly larger domain that we must ground general theories of biology and in which we will discover practical and useful applications of biology in our engineering endeavors."

Let’s be quite clear about this. Artificial Life is not "just" about biology as we have come to think of that science. It is about a new kind of biology, one that accepts the definition of life—or one like it—we discussed above, and which therefore embraces machines as appropriate objects of study.

Alife Techniques

In keeping with its joint biological and computing underpinnings, it is fitting that Alife should employ a combination of biological and computing terminology. The two primary and related techniques used in the study of Alife reflect this duality: Cellular Automata (CA) and Genetic Algorithms (GA).

CAs

"Cellular automata (CA) were originally conceived by Ulam and von Neumann in the 1940s to provide a formal framework for investigating the behavior of complex, extended systems," writes Dr. Moshe Sipper of the Swiss Federal Institute of Technology. "CAs are dynamical systems in which space and time are discrete." A cellular automaton consists of a grid of cells, just like a Go (Japanese checkers) board. A set of simple Go-like rules determine whether a cell "lights up" or not. Typically, the rules will say: If two squares adjacent to you are lit up, then don’t light up. If three adjacent cells are lit up, then you light up too. What happens then is almost magical: strange shapes start to evolve and move around the grid. Sometimes they slide off the edge, only to reappear from another edge. Sometimes the shapes break up into several new or identical shapes, and sometimes they collide and form a new entity.

This would be interesting enough, but there’s more. Computer scientists are now learning how to get CAs to compute—to act as "universal computers" able to solve problems in mathematics and physics. They are being used to study communication, computation, construction, growth, reproduction, competition, and evolution. One of the most well-known CA programs, the "Game of Life'', was conceived by John Conway in the late 1960s. (The Santa Fe Institute lists a number of more recent Alife programs you can download and run on your Unix, Mac, or DOS/Windows platform.)

GAs

"The GA was developed by John H. Holland in the 1960s to allow computers to evolve solutions to difficult problems, such as function optimization and artificial intelligence. The basic operation of a GA is conceptually simple: (1) maintain a population of solutions to a problem, (2) select the better solutions for recombination with each other, and (3) use their offspring to replace poorer solutions. The combination of selection pressure and innovation (through crossover and mutation) generally leads to improved solutions, often the best found to date by any method." (Professor David E. Goldberg, Director of the Illinois Genetic Algorithms Laboratory)

Switzerland’s Dr. Sipper provides a more detailed and more technical explanation:

"The idea of applying the biological principle of natural evolution to artificial systems, introduced more than three decades ago, has seen impressive growth in the past few years. Usually grouped under the term evolutionary algorithms or evolutionary computation, we find the domains of genetic algorithms, evolution strategies, evolutionary programming, and genetic programming. Evolutionary algorithms are ubiquitous nowadays, having been successfully applied to numerous problems from different domains, including optimization, automatic programming, machine learning, economics, operations research, ecology, population genetics, studies of evolution and learning, and social systems. . . . As a generic example of artificial evolution, we consider genetic algorithms.

A genetic algorithm is an iterative procedure that consists of a constant-size population of individuals, each one represented by a finite string of symbols, known as the genome, encoding a possible solution in a given problem space. This space, referred to as the search space, comprises all possible solutions to the problem at hand. Generally speaking, the genetic algorithm is applied to spaces which are too large to be exhaustively searched. The symbol alphabet used is often binary due to certain computational advantages . . . ; this has been extended in recent years to include character-based encodings, real-valued encodings, and tree representations.

The standard genetic algorithm proceeds as follows: an initial population of individuals is generated at random or heuristically. Every evolutionary step, known as a generation, the individuals in the current population are decoded and evaluated according to some predefined quality criterion, referred to as the fitness, or fitness function. To form a new population (the next generation), individuals are selected according to their fitness. Many selection procedures are currently in use, one of the simplest being Holland's original fitness-proportionate selection, where individuals are selected with a probability proportional to their relative fitness. This ensures that the expected number of times an individual is chosen is approximately proportional to its relative performance in the population. Thus, high-fitness ("good'') individuals stand a better chance of "reproducing'', while low-fitness ones are more likely to disappear.

Selection alone cannot introduce any new individuals into the population, i.e., it cannot find new points in the search space. These are generated by genetically-inspired operators, of which the most well known are crossover and mutation. Crossover is performed with probability pcross (the "crossover probability'' or "crossover rate'') between two selected individuals, called parents, by exchanging parts of their genomes (i.e., encodings) to form two new individuals, called offspring; in its simplest form, substrings are exchanged after a randomly selected crossover point. This operator tends to enable the evolutionary process to move toward "promising'' regions of the search space. The mutation operator is introduced to prevent premature convergence to local optima by randomly sampling new points in the search space. It is carried out by flipping bits at random, with some (small) probability pmut. Genetic algorithms are stochastic iterative processes that are not guaranteed to converge; the termination condition may be specified as some fixed, maximal number of generations or as the attainment of an acceptable fitness level. . . ."

An early and very simple example of GA programming is supplied by Richard Dawkins. In the 1960s, he developed a program he called Evolution containing two sub-programs—Reproduction and Development. Starting with some pre-programmed tree-like shapes on the screen, his Reproduction program would produce offspring from the parents, but designed according to the offsprings’ own genetic makeup inherited from the parents. The genetic makeup result is passed to the Development program, which determines the growth of the "trees"—their appearance on screen as they grow. Built-in random gene mutations ensured that, as in nature, the offspring would never look exactly like their parents.

The "environment" for the trees was provided by the human operator (Dawkins), selecting them for survival or elimination for no particular reason. This is a bit of a stretch from the natural selection that occurs in real evolution, but it’s not completely invalid.

After developing the program and running it for the first time, "Nothing in my biologist’s intuition," (wrote Dawkins in The Blind Watchmaker) "nothing in my 20 years’ experience of programming computers, and nothing in my wildest dreams prepared me for what actually emerged on the screen. I can’t remember exactly when in the sequence it first began to dawn on me that an evolved resemblance to something like an insect was possible. With a wild surmise, I began to breed, generation after generation, from whichever child looked most like an insect. My incredulity grew in parallel with the evolving resemblance."

(Aside for kids: If you think zapping aliens in Doom is cool, try making the sort of discovery Dawkins made with his program. Science is close to zero degrees Kelvin, and that’s as cool as it gets.)

The Philosophy of (Artificial) Life

In 1971, at a time when Alife was still closer to intellectual speculation than actual creation, the prescient Polish philosopher (and wonderful writer) Stanislaw Lem wrote a brilliant short story entitled Non Serviam. In it, he not only described the physics of an Alife universe (in some ways reminiscent of Frank Tippler’s physics of the real universe, nearly three decades later) but he also considered Alife from the perspectives of the Alife beings themselves and of the human beings who created them and their entire universe.

From the point of view of Lem’s Alife beings, even though they inhabited a totally different universe (a mathematical construct) from our own, they faced the same sort of philosophical and religious problems. They wonder who or what, if anything, created them? They argue among themselves about the existence of God. Is their universe open or closed, eternal or timebound, predetermined or a lucky accident, unique or just one universe in an infinite multiverse?

We know, of course, that they had a proximate creator—the scientists ("personeticists," in Lem’s fictional terminology, who created them—but even after making that stab in the dark, the Alifers are left with the same question we face: Who created the creator?

From the scientists’—from humanity’s—perspective, they face awful moral and ethical dilemmas: should they have created Alife in the first place? Having created it, are they bound to maintain it at all costs, even after their research grant funds end? Observing pain and suffering among the Alife, should they intervene? Should they so arrange things that there will be no pain and suffering? Should they give the Alifers (apparent) immortality? Above all, should they give the Alifers free will and if they do, is it really free will? (More on this next week.)

Our present implementation of Alife is a long way (a decade or three, perhaps) before it reaches the sophistication of Lem’s Alife, populated with beings that don’t just have life as we defined it at the start of this article but also have personhood—inner feelings, a soul, a sense of self, and at least the appearance of free will. It is to Lem’s credit and, insofar as Lem is a representative of our species, to our own credit that these deeply important philosophical issues are being considered before, rather than after the fact.

Meanwhile, before we reach that point, Alife is proving to be a useful computational and problem-solving tool.

Until next week,

NEXT WEEK: Free Will. Is it real, and if so can Machina sapiens have it?

P.S. Please convey my respects to your car.

Previous Features