It’s Alive!
In the earliest days of computing, Alan Turing and John von Neumann, considered the fathers of modern computer science, sketched the theoretical outlines of a scheme that remains one of the most wild-eyed and chimeric of the last half-century of science: the creation of artificial life.
At a time when the most advanced computers could barely match the power of today’s hand-held calculators, Turing and Von Neumann believed it was possible to build machines that incorporate some essential pieces of intelligent life--such as self-reproduction and thought.
The public in 1950 would have laughed out loud at Turing and Von Neumann’s idea--if it had the chance. The research was so obscure that few outside the world of computer science even knew about it.
Fast-forward to the present and a bustling factory floor cranking out heavy agricultural seeding equipment in Moline, Ill. John Deere’s factory produces more than 75 different models of machinery, each with dozens of different options. The factory must schedule worker teams, component inventories and assembly areas to turn out this extensive product line. In all, there are more than a million scheduling combinations possible--most of them not particularly efficient.
To solve the problem, John Deere four years ago turned to a type of program that uses what are known as genetic algorithms, which are modeled on the Darwinian concepts of evolution, random mutation and natural selection. The program pits mutations of itself against one another in a battle in which only the best solution survives.
“We’re using it in six factories now,” said Bill Fulkerson, the John Deere technology analyst who first sought help from artificial life researchers. “Hey, when you consider it has moved to our competitor too, it can’t be that bad an idea.”
Once on the furthest fringes of computer science, the field of artificial life has slowly begun to emerge from the laboratory and creep into the realm of commercial application.
While no one has created anything that could remotely be called alive, the decades of research have resulted in a variety of powerful tools that have helped lift mere machines from their dull, rigid existence into a realm that allows a degree of serendipity and the unexpected.
Today mutating genetic algorithms are used to predict stock market behavior and to find new drugs to combat cancer and AIDS. Animation houses such as Walt Disney and DreamWorks SKG use complex, biologically inspired software tools to create lifelike hordes of Huns charging through the snow or masses of slaves in the deserts of Egypt for the latest in animated films. The military has embraced advanced programs to evolve new strategies using virtual opponents that get more cunning with each battle they fight.
*
Even in the realm of the arts, machines have begun to make their first entrances, intruding--often hilariously--into a domain that humans have guarded as their own. At the sixth Artificial Life Conference at UCLA last month, a group of robots programmed by a USC computer science student staged a performance of a short play titled “The Self-Made Man and the Moon” before an attentive crowd of biologists, computer scientists, mathematicians and others.
The performance, in which each robot was programmed to express a few basic emotions and then turned loose to interact with other robots, had all the emotional impact of a power sander jiggling across a floor on its own.
Creator Barry Brian Werger apologized to the crowd for his robots’ off day, adding that acting was just one of many skills expected of his metallic troupe. A few days after the performance, Werger’s robots were headed to Paris to participate in an international robot soccer meet.
“They do research in the lab on a daily basis, and next year they’ll be doing weddings,” he said. “These are the hardest-working robots in show business.”
The field of artificial life, or a-life as it has come to be known, is about as broad a pursuit as any in science. The pieces of the puzzle are so numerous and complex that studying even the simplest of them can entail a lifetime of research.
*
Even in the earliest days of computer science, researchers realized that some aspects of life--evolution, adaptation and complexity emerging from simple elements--provided models for resolving some of the most difficult limitations of traditional computation.
While computers were fast at calculating figures and manipulating symbols, they were also maddeningly literal and rigid in everything they did. Artificial life researchers believed the solution lay in a “bottom-up” approach, in which machines could learn and evolve from simple instructions, as opposed to the “top-down” approach of cramming megabytes of information and instructions into a computer from the beginning.
It wasn’t until 1987 that a computer scientist working at the Santa Fe Institute in New Mexico noticed the number of people trying to model life in computers and robots. He gathered them together for a conference that year and coined the term “artificial life.”
Disney and DreamWorks were two of the earliest adopters of a-life techniques. One of the biggest problems faced by animators is the drawing of lifelike crowds. Animating each character in the crowd is too time-consuming. On the other hand, drawing one big crowd and moving it as a group looks unrealistic.
Craig Reynolds, now a software developer for DreamWorks, was one of the first animators to crack the problem. His technique, demonstrated in 1987 in a two-minute video titled “Breaking the Ice,” involved programming each element in groups of virtual fish and birds with a few simple rules, ordering characters to move away when they got too close to one another, to move closer when they got too far apart and to follow the general direction of the main body.
Once all the elements were programmed, the system was set in motion. The behavior of the crowd emerged on its own. Fish and birds glided between pillars with a chaotic, yet orderly movement that seemed lifelike in its strange complexity.
The technique has become a mainstay of animation, giving modern animated films a sense of grandeur and dynamism unknown in the past.
“I don’t expect this to replace traditional animation, but it makes it practical to do things with lots of characters,” Reynolds said. “There’s a richness and complexity that would be hard to do with traditional techniques.”
One of the most fruitful avenues of research has been the development of genetic algorithms and evolutionary programs, whose modern development has been credited to John H. Holland, a computer science professor at the University of Michigan, and Lawrence Fogel, a former General Dynamics researcher who now heads an evolutionary programming company in San Diego aptly named Natural Selection.
Evolutionary programs and genetic algorithms both rely on the concept of random variation and survival of the fittest. The programs have largely been used to find solutions to complex problems with many variables, such as the scheduling problem at John Deere’s manufacturing plant.
“Selection focuses you, and random variation is your exploration,” said David Fogel, chief scientist at Natural Selection and the son of Lawrence Fogel. “Exploitation and exploration are the keys.”
One of the company’s programs is being used to search for drugs to treat AIDS. The program matches the infinitely variable shape of certain chemicals to odd-shaped slots on the body of the AIDS virus. The theory is that plugging the slots will hamper the activities of the virus.
*
Finding a molecule that fits is like working on a three-dimensional jigsaw puzzle. The program begins by trying a random selection of shapes, which are modeled in a computer and fitted according to rules governing the bonding of different molecules. The shapes that come closest to fitting are then slightly mutated and set in competition with each other to create another generation of shapes that fit a little better. The process is repeated over hundreds of generations--a process that takes about two minutes on a high-powered computer.
Agouron Pharmaceuticals of San Diego has been using the tool for the last year.
In the case of John Deere’s factory problem, the genetic program follows a similar course, except that instead of varying the shape of a molecule, it juggles the dozens of variables in farm equipment assembly. The program begins by creating 20 random schedules to assemble several hundred seeding machines over a one-month period.
Out of those schedules, the best ones are crossbred with other good candidates by randomly mixing parts of their daily schedules. The crossbreeding continues for about five hours, during which time about 600,000 schedules are evaluated. Although the program constructs a one-month schedule, the company only uses the first day. For the next day’s schedule, the whole process is run again from scratch.
A Santa Fe, N.M.-based firm called Prediction Co. is now using a-life programming techniques to predict the movement of stock markets. Computer games, such as “Creatures” from CyberLife Technology, that use some form of a-life have already begun to appear on the market. And in the world of art, a group of artists has organized an informal school it calls the Algorists.
In many ways, the strange pieces of computer art provide the clearest illustration of the unpredictable union of machines and life. From intricately patterned prints to autonomous computer animations with imaginary creatures flitting about the screen, there is an engaging sense of mystery over how such complexity can spring from a mere machine.
*
At the UCLA a-life conference last month, a steady stream of viewers stopped to play a sort of computer game titled “The Bush Soul,” presented by Rebecca Allen, chairman of UCLA’s design department. Viewers controlled one character in the program with a joystick, prompting a shift in the behavior of the other characters gliding across the screen. Each character was programmed with a basic personality, and the ballet of movements on the screen was the result of different personalities coming in contact with one another. Even though most modern computer games have wilder graphics, “The Bush Soul” drew a continual flow of onlookers intrigued at the subtle interplay.
“It is the movement that is the art to me,” Allen said as she watched groups of people take their turns with the joystick. “They really seem to be alive, and that is what is freaky.”
*
Times staff writer Ashley Dunn can be reached via e-mail at [email protected].
