A robot that duplicates itself? It's entirely possible, says this expert.

(c) 1983 Robert Freitas, Jr.

Omni 5(July, 1983):44-47,117-118.

The next 20 years may witness the birth of a man-made life form that could lead us into space – much of the preparatory work toward this dream has already been done.

Picture one possible result of these efforts:

From a rocket that left Earth several years before, an enormous egg drops to Saturn's ice moon Enceladus and cracks open, releasing the robot inside. Stilting spiderlike on the uneven surface, the automaton immediately sets about reproducing itself, using only the materials at hand and feeble light energy from the distant sun.

Soon the robot and its descendents begin their real task: mining Enceladus's ice and building small light-sail tugs to carry the chunks toward the inner solar system. For a time, earthly astronomers see nothing unusual, but eventually a new ring begins to appear around Saturn, surrounding the old ones at about twice the distance. A cloud of replicating robot vessels spirals outward from Enceladus and then spills in a long stream toward Mars. Their shipments of ice fall like sparkling meteors onto the Martian surface, melting on impact and thawing the frozen ground. First rivers, then whole seas form. The air grows thick and warm, and soon it rains on Mars for the first time in perhaps a billion years. Within a decade of this transformation human colonists arrive on their new world.

According to Robert A. Frosch, former NASA administrator, such missions are not only possible but necessary. In a talk before the Commonwealth Club of San Francisco, he told his startled audience that to support ourselves in space, we will need self -reproducing robots. They would. he declared, "provide easy access to the resources of the solar system for a relatively manageable investment."

The key to the scheme is a machine that can use solar energy and local materials to build a replica of itself with little or no human guidance. Since generation after generation of offspring would be built. the total number of machines would grow exponentially, the way biological populations expand. So would the machines' output of manufactured products.

NASA is taking this concept very seriously. In 1980 it held a ten-week summer study session at the University of Santa Clara, in California. My group, called the Replicating Systems Concepts Team, studied the idea of setting up a self-reproducing factory on the moon that would eat raw lunar soil and manufacture anything we need from what it ingested.

Our basic plan would put a 100-ton seed full of machinery on the moon. The first .robots would emerge to fuse the lunar topsoil into a circular factory site of cast basalt 100 yards across. Then they would install the factory itself and erect a canopy of solar cells to power the system.

The factory has three major sections: One extracts purified elements from the soil: another forms them into machine parts, tools, and electronic components; and the third assembles the parts into useful products. In a year a 100-ton seed could extract enough materials to duplicate itself. If allowed to grow undisturbed for 18 years, the factory output would total more than 4 billion tons per year-roughly the current industrial output of the entire world.

A growing, self-replicating factory could be programmed to mass-produce robot miners and spacecraft-almost anything we need. "It could build a few thousand meter-long robot rovers equipped with cameras, core samplers, and other survey instruments," suggests Georg von Tiesenhausen, assistant director of the Advanced Systems Office at the Marshall Space Flight Center, in Huntsville, Alabama, and a member of the replicating systems team. "They could cover the moon like ants, mapping it in just a few years, By conventional methods, it might take a century or more to do the same thing."

How soon could such a system be in operation? Von Tiesenhausen says that within 20 years after the project is begun, the United States could produce the first robot able to duplicate itself from raw materials. Former NASA administrator Frosch is even more optimistic. "We are very close to understanding how to build such machines," he says. "I believe that the technology is already available and that the necessary development could be accomplished in a decade or so."

Long before NASA became interested in self-replicating machine systems, the basic theory had already been worked out in some detail; much of it has been around for more than 30 years.

It began in 1948 with the late John von Neumann, a brilliant Hungarian mathematician famed for his early work on electronic computing. In a series of lectures delivered at Princeton, he discussed how automata might reproduce themselves.

According to Von Neumann a self-replicating machine must have at least four distinct components: the builder. the copier. the controller. and the blueprints. Reproduction starts when the controller commands the builder to construct exact replicas of all mechanical systems according to instructions in the blueprints, a sophisticated computer code. The robot would pick the right machine parts from its stockroom and assemble them in order. Then the controller would command the copier to duplicate the blueprints, insert the copy into the replica. and turn the new robot on. Voila-two machines!

Other scientists have also given serious thought to self -replicating automata. Physicist Freeman Dyson, of Princeton's Institute for Advanced Studies, suggests that a small robot adapted to earthly deserts might duplicate itself from the silicon and aluminum in the rocks around it. Powered by sunlight, it would manufacture electricity and high-tension lines. Its progeny could eventually generate ten times the present electrical output of the United States.

Though the robot's potential for destroying the natural environment would be enormous. Dyson believes it would eventually be licensed for use in the deserts of the western United States. probably after bitter debate in Congress. The robot would probably have to carry within itself a memory of the original landscape so that it could restore the site's appearance whenever a location was abandoned.

"After its success here," he speculates. "the company that built it might market an industrial-development kit for the Third World. For a small down payment, a nation could buy an egg machine that would mature within a few years into a complete system of basic industries. along with the associated transportation and communications networks." A spinoff, he suggests, might be the urban renewal kit, equipped with self replicating robots programmed to build brand-new neighborhoods from the debris of the old ones.

Computer scientists have received such schemes enthusiastically. Ewald Heer, a robotics specialist at NASA's Jet Propulsion Laboratory, in Pasadena. California, calls self-replicating robots one of the most fascinating ideas for the future of space. "This offers a way to create a self-supporting economy by robot labor," he observes. "Immigrants from Earth could set out, knowing that the means of their survival had already been provided."

Michael Arbib of the department of computer and information sciences at the University of Massachusetts at Amherst, suggests that they might also be used for interstellar communication. "A self-reproducing machine might carry out its own synthesis from the interstellar gas," he offers.  These machines, Arbib says, could reproduce in space creating an expanding sphere of explorers moving outward toward the far reaches of the universe.

With, this scheme in mind, Frank J. Tipler of Tulane University in New Orleans, has even argued that intelligent aliens cannot exist: If they did. they would have had to build such machines to explore and use the galaxy, and we would see glaring evidence of these machines all around us.

Some machines have already managed to reproduce in primitive ways. Self-replicating computer programs have been written in nearly a dozen different languages. and small machines that can copy themselves from simpler parts have proved remarkably easy to build.

One basic model was developed years ago by British geneticist L. S. Penrose at University College, London. It is an ingenious set of interlocking blocks with clever arrangements of springs, levers, hooks, and ratchets. A two-. three-, or even higher-block assembly can replicate when placed in a box with other loose blocks and shaken gently. One end of the completed assembly hooks on to the loose blocks in the right sequence. building up a duplicate chain and then releasing it when the final block has been connected.

Homer Jacobson, a physicist at Brooklyn College. in New York, built another such device, using an HO train set. In his invention, there are two kinds of programmed, self-propelled boxcars, called heads and tails, that circulate randomly around a loop of track with several sidings. If a pair of boxcars, a head and tail, is assembled on a siding, it can reproduce itself.

Here is how it is done: The head car in the pair waits for a loose head car to come by and shunts it onto the next open siding. Then the next loose tail car to come by is shunted onto that same siding to make a new head-tail pair. Once this happens the first boxcar couple turns itself off and the second pair becomes the active, reproductive one. It. can reproduce using the next open siding. and this chain of pair reproduction continues until all sidings are filled or all components are used.

Such experiments sound much too simple to justify calling them reproduction, nothing like the mysterious processes that form a new human life. You might even object that Von Neumann's whole concept is just a general-purpose assembly robot whose output happens to be copies of itself. But after all, observes W. Ross Ashby, a biophysicist at the Burden Neurological Institute. "living things that reproduce do not start out as a gaseous mixture of raw elements." Even human beings require a specialized environment supplied with air, water, and nutrients in order to procreate. Von Neumann's robots happen to be just a little less independent.

In fact some scientists already feel that computers are more than mere machines. John G. Kemeny, president of Dartmouth College and one of the inventors of the computer language BASIC, believes that computers should be considered a new species of life. "Once there are robots that reproduce," he declares, "it would be easy to program them so that each offspring differs slightly from its parent. It would be a good idea to let each robot figure out some improvement in its offspring so that an evolutionary process can take place."

But compact, self -reproducing robots still lie beyond the technological horizon. According to Marvin Minsky, head of artificial-intelligence research at MIT. an automaton today would have to be the size of a factory to reproduce itself from raw materials rather than from prefabricated parts.

Fujitsu Fanuc, Ltd., a manufacturer of numerically controlled machine tools. took a giant step toward that goal with a $40 million robot factory. Robots there are built by other robots, with only 100 humans to supervise and help. The plant produced 100 robots in its first year. Once such a plant can make its own components. it can be programmed to make more of itself to reproduce. (See "Robots of Japan," January 1982.)

Since we cannot foresee all the problems these robots will have to face on a distant planet. we must supply them with goals and with the problem-solving ability to carry out their assignments in our absence. It seems at least possible that machines this complex will begin to evolve some of the social behavior common in animal populations. This brings us very close to sharing our planet with a form of near-life whose evolution we cannot predict.

At the simplest level. what would happen if one machine began to neglect its production chores in order to reproduce? Its offspring. possessing the Same trait, might soon dominate the machine population. Would some form of "kin-preferring" behavior arise'? Might the robots even develop a form of -reciprocal altruism ' in which the machines behave in seemingly unselfish fashion toward others that are not "kin to create a more stable "society"?

"If our machines attain this behavioral sophistication," notes Richard Laing, of the department of computer and communication sciences at the University of Michigan, in Ann Arbor, "it may be time to ask whether they have become so like us that we have no further right to command them for our own purposes and so should quietly emancipate them.

And one wonders: Could such self-reproducing robots someday become our enemies? The usual answer is that we can just pull their plugs to regain control over them. But is that really so? We are already so dependent on computers that to shut them down would cause general economic chaos. Of the Santa Claus machines, theologist Ralph Wendell Burhoe, of the Meadville/Lombard Theology School, in Chicago, asks, "Will we become the contented cows or household pets of the new computer kingdom of life?"

And what if the machines learned to defend themselves? The Replicating Systems Concepts Team at the Santa Clara study session concluded that to escape human control, any machine must have at least four basic abilities: It must create new ideas to explain conflicting data, inspect itself completely, write its own programs, and change its own structure at will. A machine that lacked even one of these abilities would almost surely be unable to anticipate or prevent its own disconnection. It seems unlikely that machineswill soon acquire these powers, and even if they do, it will be only because human beings make a conscious decision to supply them.

Nonetheless, a few people view the future gravely. James Paul Wesley, associate professor of physics at the University of Missouri, points out that the advent of machines has been amazingly abrupt compared to the billions of years it took carbon-based life to evolve on Earth. Yet the same laws of reproduction apply to both biological evolution and machines.

"Machines," Wesley cautions, " have also evolved toward an increased biomass [quantity], increased ecological efficiency, maximal reproduction rate, proliferation of species, motility, and a longer life span. Machines, being a form of life, are in competition with carbon-based life." The result, he fears, is that silicon life "will make carbon-based life extinct."

Yet there is another possibility. What we are approaching, says NASA computer scientist Rodger A. Cliff, is "cybersymbiosis"; eventually humans could come to live inside a larger cybernetic organism. As man and machine evolve, our interactions will cease to be voluntary and become necessary. Cliff views this development with eagerness. Our descendants could live inside large, self-replicating, mobile space habitats. which would act as extraterrestrial refuges and guarantee that humanity is never completely wiped out by some sort of earthly catastrophe.

"Flesh and blood are ill adapted to space," Cliff comments, "but silicon and metal are ideal. Just as our own DNA resides within a protective membrane, and mitochondria are locked within cells, so might humanity live as cybersymbiotic organelles of the space-colony organism. I see these traveling throughout the cosmos, searching for nutrients-asteroids, gas clouds, and so on-growing, evolving, and reproducing. And we will be inside their offspring."

Editors' note: Robert Freitas. Jr., is a physicist/lawyer and expert on extraterrestrial life, This article grew out of his participation in a NASA conference, 'Advanced Automation for Space Missions," at which he discussed the idea of using self-replicating robots for space exploration.

References To This Page

Creation date: August 28, 1998
Last Modified: August 28, 1998
HTML Editor: Robert J. Bradbury