Artificial Life: Can Computers Discern the Soul?
LEAD: NO longer content with dissecting tissues, analyzing proteins and breeding fruit flies, an increasingly diverse group of scientists has decided that the best way to study life is to make some of their own.
NO longer content with dissecting tissues, analyzing proteins and breeding fruit flies, an increasingly diverse group of scientists has decided that the best way to study life is to make some of their own.
They are creating a field called artificial life, mixing the impulses of biology with the tools of computation. By looking beyond the usual materials of life - beyond the familar biochemistry of earthly animals and plants - they hope to capture its spirit: the animated, the energetic, the replicating, the evolved.
Most of the would-be organisms of artificial life exist solely in the electronic environment of the computer, where they have little danger of being confused with the real thing. The first conference on artificial life, held last week at Los Alamos National Laboratory, offered models of processes from protein formation to plant growth to animal predation - processes meant to be, if not life, then at least lifelike.
The simulations of biology address some of the most troubling questions of the life sciences: how the primitive precursors of DNA gained the ability to store information and copy it; how the senseless force of natural selection created structures of such extraordinary complexity and beauty; how the laws of ecosystems arise from the whims of individual animals.
They also reflect an expanding sense within science of what life is. Artificial life seeks ''the ghost in the machine,'' as the conference organizer, Christopher Langton of Los Alamos, put it - an essence arising out of matter but independent of it. For the first time in generations, some researchers believe, science has a legitimate way of talking about life's soul.
''It lies in the complexity of organization,'' said Richard Dawkins, an evolutionary biologist at Oxford University. ''It's not a substance; there's no living material. It's just an incidental fact that in real living things the entities that happen to be organized happen to be made of organic, soft, squishy stuff, whereas in a computer they're made of hard, nonmoving chips.''
The creatures of artificial life already make up a strange menagerie. There are flocking birds and schooling fish just a few generations removed from the cartoons of Walt Disney. Invisible bugs breed and die out as they leave trails through a mound of electronic food. Computer flowers bud and unfold, their timing controlled by computer chemicals running up and down computer stems. Stick-figure shapes evolve in a few dozen generations into startling butterflies and shellfish.
Some simply imitate real organisms. Most, however, depart from reality to capture some abstract quality of living things, preferably a quality that arises not from the designer's intent but from unplanned processes. 'What Is the Soul?'
''What keeps me awake at night is not correspondence to reality,'' said Steen Rasmussen of the Technical University of Denmark. ''I want to know what is the soul in this that creates order - what is the engine.''
Stripped of bone and sinew, leaf and petal, ribosome and chromosome, life still has a unique logic that can be abstracted in a computer - that, at any rate, is the belief driving the new discipline. Nor is the computer essential. Some scientists are trying to create microscopic carriers of information in fragmentary protein strands or pieces of clay crystal.
''Surely there must be a more general sort of biology,'' said Graham Cairns-Smith of the University of Glasgow, author of ''Seven Clues to the Origin of Life.'' ''This is the aeroplanes-don't-have-feathers principle. Yes, birds have feathers and fly beautifully, but we have different requirements.''
Those explicitly seeking to create life, within a computer or a test-tube biochemical system, form a group that now brings together microbiologists, evolutionary theorists, physicists, chemists and computer scientists. At Los Alamos, they spawned rooms full of computer demonstrations, wandered from place to place wearing buttons asking ''What is a genetic algorithm?'' and showed videotapes of robots taking five hours to weave across a room. A Problem of Definition
They face a problem of definition. Most modern biologists think of an organism's abilities to process matter and energy, to replicate itself and to evolve as the essential, defining qualities of life. Some computer models already have those abilities, in more or less trivial ways.
So scientists debate the question of how they would recognize a genuine artificial creature if they had one. After one particularly testy exchange, a scientist proposed that a key criterion should be ''irritability.'' Others recommended purposefulness and unpredictability as qualities any good organism should have.
Gerald F. Joyce of the Salk Institute in San Diego suggested the biologist test: put the artificial organism into a room with a biologist. If the biologist comes out and says it's alive, that would be encouraging.
''And if your organism comes out and says it's alive, then you're on the right track,'' Dr. Joyce said.
Many of their colleagues will accuse those promoting artificial life of overoptimism and exaggerated claims, particularly about the capabilities of computers. They remember the overselling of artificial intelligence in the 1970's, and the Los Alamos conference, too, drew a healthy contingent from what one biochemist, Hyman Hartman of the University of California at Berkeley, called the ''computers are the next form of life, so let's get on with it'' school. Striking Patterns Emerge
Dr. Hartman warned against relying too blindly on computer models. As his own model undulated and sparkled hypnotically on the giant screen behind him, he told the audience, ''One of the great dangers of artificial life is that you can be very, very clever and invent beautiful machines that do beautiful things, but you've gotten very, very far away from what you're trying to understand.''
His simulation, a checkerboard of 65,000 cells that changed color according to simple rules, was meant to show how simple processes on the surface of a clay crystal might generate complexity. Indeed, as in several other demonstrations, strikingly rich and irregular patterns arose - large-scale structure emerging from the interplay of small-scale rules.
The recognition in recent years that complexity can arise spontaneously from simple systems gives the field of artificial life its strongest motivation. The scientists agreed that the most promising demonstrations were those whose lifelike qualities emerge unbidden, surprising even their programmers. 'Miracles Aren't Allowed'
They want to play god, but not a god who directs the motion of every sparrow, as Dr. Langton said. ''We don't want to do that - no global controller,'' he said. ''And no miracles -miracles aren't allowed except at the very beginning.''
A computer graphics expert trying to create a flock of birds that will fly convincingly around obstacles, for example, must create a free-flowing yet tightly coordinated pattern of motion. Instead of programming a flock from the top down, Craig W. Reynolds of Symbolics Inc. let each of hundreds of imaginary birds follow a set of rules for avoiding their neighbors.
A natural-looking flock took shape, sweeping gracefully but not rigidly around blocks and cylinders. And unexpected behavior emerged as well - one bird crashed into an obstacle, fluttered in a momentary daze, and then staggered onward. Organizing Principle
The spontaneous emergence of organization is a central problem of life at all scales. Those studying the origin of life are acutely aware that, without some self-organizing principle, it would take many times the age of the universe before chance would bring amino acids together in just the right combinations necessary to form the elaborate machinery of DNA.
Self-organization must also guide the combination of embryology - the unfolding of individual creatures according to the rules of development built into their genes - and evolution. These remain deep mysteries, and computer models are intended to show not how they do occur, but how they might plausibly occur.
Scientists have discovered in recent years that some seemingly complicated patterns, like the branching, jagged structures of plants, have simple descriptions in the language of fractal geometry, in which patterns are built up from rules repeated on different scales. No one knows just how such rules are encoded in the genes of real plants; nevertheless, several demonstrations at Los Alamos created lifelike ferns, trees and even flowers from relatively modest fractal instructions.
One program, by Przemyslaw Prusinkiewicz of the University of Regina in Canada - the winner of an ''Artificial 4-H Contest'' for most lifelike organism at Los Alamos - mimicked the growth of a variety of flower species. It combined geometric instructions with a set of timing signals, like the chemical signals that real plants use to control branching and budding. The results were vivid images of plant growth. A Surprising Look at Evolution
Such models illustrated rich development with no possibility for evolution. By contrast, Dr. Dawkins, the Oxford zoologist and author of ''The Blind Watchmaker,'' offered a stick-figure version of embryology with surprising evolutionary power.
Through random mutation and a somewhat arbitrary version of natural selection, the program manages to evolve into shapes with surprising complexity and often a surprising resemblance to earthly creatures. Each experience with the model brings new evolutionary paths, none of which could be predicted.
The results are just drawings on a computer screen, with neither the attributes nor the potential of real life, as Dr. Dawkins himself noted. In the long term, he said, electronic versions of evolution could produce something more. Ways of Thinking About Life
In general, by creating a variety of computer environments, universes with their own sets of rules, scientists intend to provide ways of thinking about universal principles of life -principles more general, perhaps, than those observed in nature. Computer scientists since John von Neumann, one of the fathers of computer in the 1940's and 50's have known that such artificial environments can create ''self-replicating automata,'' organized structures that reproduce themselves.
''If they don't have the whole enchilada, at least they have a few pieces of lettuce,'' said A.K. Dewdney of the University of Western Ontario, Scientific American's computer columnist. ''If we have a system that can organize itself or can evolve in some sense, then the hope is that down the road the system, without any thought or care on our part, will become intelligent.''
For artificial life to become a successful approach, Dr. Dewdney and other scientists said, models will have to become much richer than the first efforts. They will have to combine processes of growth, competition and evolution, only pieces of which have been seen so far.
Still, many of them are optimistic -willing descendants of Dr. Frankenstein, who remains ''the bugaboo metaphor for artificial life,'' in the words of J. Doyne Farmer of Los Alamos, an expert on chaotic dynamics who is modeling the body's immune system. He echoed some other scientists in calling the prospects frightening, perhaps not so much because of what might be created as because of what it might tell us about people.
No comments:
Post a Comment