Previous Top

5.4 Applications

Having shown that machine SRS is, in principle, both theoretically possible and feasible in terms of engineering systems design, their usefulness in some economic or commercial sense remains to be demonstrated. That is, what might such systems permit humankind to do that could not be done before? The main advantage in using SRS over other methods of space exploration and industrialization is that a very large capability for performing any desired task can be rapidly achieved at arbitrary remote locations, starting with a relatively small investment of time, money, energy, and mass in the original "seed" mechanism.

The team has identified four general criteria for determining the most probable and profitable application of replicating systems technology:

  1. A large number of identical or similar products is required;
  2. Excessively long production periods for alternate approaches are required;
  3. Raw materials or parts are available onsite; and
  4. Sufficient physical space is available for replication.

Each of these criteria should be applicable, or largely so, in a specific case before the use of SRS technologies is considered.

Replicating systems will find many applications on Earth, in near-Earth and lunar space, throughout the Solar System, and in the interstellar realm, for both exploration and utilization. SRS also provides a number of fascinating applications in basic and applied research in automata theory, theoretical biology, experimental evolution, and machine intelligence and robotics architecture.

5.4.1 Terrestrial Applications

The early development of replicating systems technology on Earth will be the history of modern industrial automation. The United States at one time enjoyed the highest productivity in the world, and still partakes of the prosperity that that has brought. Recently, however, competition from other nations who are more rapidly automating their industries is seriously eroding the U.S. position of leadership. The resulting economic forces are impelling domestic industry to accelerate the automation of its factories.

The space program is viewed by many as a high technology venture which predominantly makes use of computers, robot spacecraft, and other trappings of automation. In reality, NASA's activities are strongly people intensive. For example, large teams of trained technicians and scientists are required to operate a robot space probe by remote control. The same economic forces at work in the marketplace are forcing NASA to rethink its traditional way of doing business. Not only will there be more automation in the space program for this reason, but also there will be missions that are difficult or impossible to conduct without using advanced machine intelligence and robotics technologies. The harsh environment of space, the significant costs of life support systems for human beings and of "man-rating" space systems for safety, and the communications problems caused by the immense distances involved in interplanetary travel have given NASA additional incentives to develop systems of total automation beyond those commonly employed in industry. The sheer magnitude of many potentially interesting missions requires massive automation.

Accordingly, NASA should strongly participate in automation research and development in anticipation of spinoffs to industry of great potential value. The agency also should closely monitor industrial R&D efforts, remaining alert for new developments on the commercial front which might prove beneficial to the space program. The infusion of NASA funds at critical points could allow the agency to exert subtle influence on industrial development so as to provide for NASA's special needs at less cost than an independent program to achieve the same ends.

Similarly, the Department of Defense (DOD) is embarking upon an ambitious program of industrial automation. The aim is to produce war materiel in the most economical and flexible manner possible, and to shorten the time between concept and field deployment of weapons systems.

Much of the DOD effort will produce results useful for the space program. To take maximum advantage of this, NASA should maintain close liaison with DOD and should join in various cooperative efforts in areas of overlapping interests.

Computer-aided design {CAD), manufacturing (CAM), and testing (CAT), and robotics. Automation for replication will require extensive application of computer science and robotics. At the initial stage of development, and during periods when repair or reconstruction operations must be performed, computers can be used in many ways to aid the design process (CAD). They are excellent for generating and maintaining documentation. Computer-executed graphics are invaluable in assisting human operators to visualize complex objects in the absence of a real, physical construction. Simulation using computer models is used in place of, or as a cost-saving adjunct to, physical models or prototypes. Recent developments in machine intelligence research has made far easier the complete automation of the entire design process. Ultimately, the capability will exist for a human to carry on a dialog with a computer system in which the person merely defines the functional specifications of the desired product and the computer determines the remaining design details autonomously.

Computers have been used in manufacturing (CAM) for more than two decades. The most common modern application is business data processing. Computerized inventory control and scheduling are two promising uses rapidly gaining prominence today. Process control using analog computers began many years ago in chemical plants, steel mills, and paper mills. Newer facilities rely instead upon digital computing. An important subset of process control is numerical control (N/C) of machine tools, with instructions traditionally recorded on punched paper tape. Today it is feasible to connect N/C machine tools directly to a computer able to generate and store instructions in electronic memory, and increasingly this is being done, especially in the aerospace industry.

Computers can also be used to great advantage in the testing of products (CAT). (This is distinguished from measurements of process variables, which is considered a process control function.) Highly complex products such as microprocessor integrated circuits cannot realistically be tested without the aid of computer technology. A standard interface protocol (the IEEE488 bus) has been defined for the interfacing of test instrumentation to a host computer.

In the context of a factory, robotics generally is understood to refer to materials handling and assembly functions. Typical operations include loading/unloading machine tools and spot-welding automobile bodies. Hard automation (special-purpose robots of very limited versatility) commonly are used in applications requiring high volume output. But computer-controlled general-purpose robot manipulators are becoming increasingly popular, as exemplified by the rather anthropomorphic PUMA device (a robot arm system manufactured by Unimation).

Replicative automation. CAD, CAM, CAT, and robotics technologies could be combined to produce an almost totally automated factory. The Department of Defense has instituted an ongoing program designed to promote this very concept, called Integrated Computer-Aided Manufacturing or ICAM. The technology now exists to design integrated circuits in one location (CAD), then fabricate the masks for microelectronic manufacture in another (CAM) under the direction of several intercommunicating computers. Further developments and advances in ICAM techniques are imminent.

In a very real sense, an industrialized nation is a symbiotic self-replicating, growing "organism" consisting of humans and machines working together. At the beginning of the industrial revolution the "organism" consisted chiefly of human beings, who, aided by a few machines, performed logical and physical functions. In later years more and more of the heavy and most dangerous work was delegated to machines. As ICAM increasingly enters the mainstream of industrial automation, the logical processes of man-machine manufacturing "organisms" will begin to be taken over by sophisticated computer systems and the physical functions will be dominated by commercial robot devices.

When ICAM techniques are directed toward the production of components of their own systems (CAD, CAM, CAT, and robot machines), a regenerative effect occurs in which each generation of automated factories is cheaper to construct than the preceding one. By the time this regeneration, which has been termed "superautomation" (Albus, 1976), is achieved on Earth, there may be very little human intervention in the replication process except for supervisory and top-level guidance functions. The final step in achieving totally autonomous machine replication requires the replacement of the human top-level managers with computers and turning over any remaining physical tasks to robot devices.

The near-term removal of all human intervention from the industrial "organisms" on Earth is highly unlikely. Certainly people may want to continue to perform various logical and physical functions for social or psychological reasons, and man may always remain the decision maker in control of which products are produced. Certain tasks are likely to prove more difficult to automate than expected, and human beings will continue to perform these jobs for economic reasons for a long time to come. Superautomation on Earth will proceed only as far and as fast as is economically advantageous.

The long-term future almost certainly will see the development of full replicative automation capability on Earth. Whether it is economical remains an open question at present. The main advantage of pure machine replicating systems over man-machine symbiotic systems is that autonomous factories can be sent to locations where there is not, or cannot be without great expense, a population of human workers adequate to operate and maintain the factory complex.

Prime candidates for terrestrial replicating systems applications will most likely be mass-produced products for use in inaccessible or hostile places requiring large spaces to perform the specified tasks. Possibilities include large photovoltaic arrays for centralized power plants in the southwestern regions of the United States (Leonard, in-house document, Bechtel Natl. Inc., San Francisco, Calif., 1980), desert irrigation and soil conditioning equipment covering vast areas, agricultural or military robots, ocean-bottom roving mineral retrievers and seawater extractors patrolling the vast continental shelves, or solar power satellite ground receiver (rectennae) devices. Each of these machine systems could probably be made to selfreplicate from a basic feedstock substrate, possibly even from a raw material substrate ultimately.

A few somewhat more speculative terrestrial applications have been proposed by imaginative writers. For instance, Moore (1956) suggested the idea of an artificial living plant able to extract its own nutrients from the sea. These machines could obtain energy from sunlight to refine and purify materials, manufacture them into parts, and then assemble the parts to make duplicates of themselves. Such plants could be harvested for a material they extracted or synthesized. Thus, an artificial plant which used magnesium as its chief structural material could be cannibalized for its metal content. Like lemmings, schools of artificial living machines could be programmed to swim to a harvesting factory when they reached adulthood.

Clearly there would be need for international controls and allocation of areas for production and harvesting. This would involve not only the political rights of nations but also questions of natural conservation. Social problems could arise in connection with the selection of products to be manufactured. An artificial plant might be designed to make a product useless to the plant itself. It might extract gold from seawater, refine it, and cast it into an ingot, which would be harvested as the crop from the plant. But this would be a shortsighted choice. Multiplying at an exponential rate, the gold-making plant would soon produce so much that gold would lose its scarcity value and probably end up being worth very little. An excellent candidate for production by an artificial plant is fresh water, which is needed in great quantities in various parts of the world.

Dyson (1979) suggests a small self-reproducing automaton well adapted to function in terrestrial deserts. It builds itself mainly out of silicon and aluminum which it extracts from ordinary rocks wherever it happens to be. Its source of energy is sunlight, its output electricity and high-tension transmission lines. There is bitter debate in Congress over licensing this machine to proliferate over our Western states. The progeny of one robot can easily produce ten times the present total power output of the United States. Legislation is finally passed authorizing the automaton to multiply, with the proviso that each machine shall retain a memory of the original landscape at its site, and if for any reason the site is abandoned the device is programmed to restore it to its original appearance.

After its success with the rock-eating automaton in the United States, the company places on the market an industrial development kit, designed for the needs of developing countries. For a small down payment, a country can buy an egg machine which will mature within a few years into a complete system of basic industries together with the associated transportation and communication networks, custom made to suit the specifications of the purchaser. The vendor's guarantee is conditional only on the purchaser's excluding human population from the construction area during the period of growth. After the system is complete, the purchaser is free to interfere with its operation or to modify it as he sees fit. (A technological spinoff is the Urban Renewal Kit - a city's architects and planners work out a design for urban rebuilding, then the kit is programmed to do the job for a fixed fee.)

Theodore Taylor calls all such devices "Santa Claus Machines" because of their almost "magical" behavior (Calder, 1978). In his version of SRS, a fully automatic mining, refining, and manufacturing facility gathers scoopfuls of raw lunar materials and then processes them by means of a giant mass spectrograph with huge superconducting magnets. This device converts mined material into an ionized atomic beam which is deflected by the magnetic field. Lighter elements curve more than heavier atomic species, so the material is sorted into stockpiles of constituent elements atom by atom. To manufacture any item, the Santa Claus Machine selects the necessary metals and plastics, then vaporizes and sprays them onto a mold. Instructions for manufacturing, including directions for adapting to new processes and replication, are stored on magnetic tapes in the machine, perhaps activated by radio command from Earth. Conceivably, costs eventually could fall to zero; and if the workload grows too large, the machine simply reproduces itself.

5.4.2 Near-Earth and Lunar Space Applications

While terrestrial self-replicating systems may be limited for some time to coevolution with Earth-based industry constrained by normal economic factors, the prospect for extraterrestrial applications is quite different. The difficulty of surmounting the Earth's gravitational potential makes it more efficient to consider sending information in preference to matter into space whenever possible. Once a small number of self-replicating facilities has been established in space, each able to feed upon nonterrestrial materials, further exports of mass from Earth will dwindle and eventually cease. The replicative feature is unique in its ability to grow, in situ, a vastly larger production facility than could reasonably be transported from Earth. Thus, the time required to organize extraordinarily large amounts of mass in space and to set up and perform various ambitious future missions can be greatly shortened by using a self-replicating factory that expands to the desired manufacturing capacity.

In the not-too-distant future such facilities could be sited either in Earth or lunar orbit, or on the surface of the Moon. The chief advantages of orbital factories are near-zero gravity, absence of lunar dust or atmosphere, convenience in choice of orbit, proximity to Earth (relative ease of transport of finished products), and unobstructed view of virtually the entire celestial sphere. For some applications, however, the lunar surface may be the preferred location. Many manufacturing processes require at least small amounts of gravity, and the availability of solid ground for physical support may be important too. The main advantage to factories on the lunar surface is that the raw materials to be processed into finished products are right at hand - only relatively low-mass final products need be lifted from the lunar surface, rather than bulky raw materials as in the case of an orbital factory. The Moon can also be used as a shield to block sunlight or electromagnetic interference from Earth during highly sensitive observations.

The useful applications of replicating factories with facilities for manufacturing products other than their own components are virtually limitless.

Manufacturing. Huge solar power satellites with dimensions 1-10 km on a side could be constructed in Earth orbit by a fleet of free-flying assembly robots or teleoperators manufactured by a replicating factory complex using material from the Moon. Components for very large structures, including communications, storage, recreational, penal, or even military platforms could be fabricated, and later assembled, by an SRS. Another exciting mass-production possibility is the notion of orbital habitats, or "space colonies" (O'Neill, 1974, 1976), by which increasingly large populations of human beings could be safely and comfortably maintained in a support capacity for the space program. Additionally, a replicating factory could build more copies of itself, or new variants of itself capable of manifesting different behaviors and producing different outputs, in almost any desired location. Possible useful output of such facilities already has been summarized in section 5.3.4.

Observation. Exceedingly large sensor arrays for Earth or astronomical observations could be rapidly constructed from nonterrestrial materials by a self-replicating manufacturing facility. This technology could be used to make feasible such advanced missions as optical extrasolar planet imaging (using millions of stationkeeping mirror assemblies arranged in an array with an aperture diameter on the order of kilometers); complex multisensor arrays; very large, high-resolution x-ray telescopy; and other self-organizing optical or radio telescopic arrays of grand proportions to permit such ambitious undertakings as galactic core mapping, continuous observation of large numbers of passive fiducial markers for Earth crustal plate motion monitoring, and various SETI (Search for Extraterrestrial Intelligence) observations including beacon acquisition, radio "eavesdropping," or, ultimately, active communication. Automated mass production will make possible arrays with heretofore unattainable sensitivity and spatial resolution.

Experimentation. Replicative automation technology will permit a tremendous expansion of the concept of a "laboratory" to include the Earth-Moon system and ultimately all of the bodies and fields in the Solar System. A number of grand experiments could be undertaken which would prove too costly if attempted by any other means. For example, an Earth orbital cyclotron could be constructed as a series of thousands of robot-controlled focusing coils and stationkeeping target assemblies within the terrestrial magnetosphere, with operating energies possibly as high as TeV for electrons and GeV for protons. Additional experiments on magnetospheric propulsion and energy generation could be conducted by free-flying robot drones manufactured on and launched en masse from the lunar surface. Gravity field probes, including mascon mappers and drag-free satellites, could be coordinated to perform complex experiments in kinematics, special and general relativity, and celestial mechanics. Investigations of artificial in situ lunar crater formation dynamics, solar wind composition and utilization, unmanned ecological simulation modules, and isolation or "hot lab" module manufacturing for conducting dangerous experiments with explosive, radioactive, or biologically engineered materials are still further possibilities.

Exploration. The Moon is largely unexplored. A growing, self-replicating factory could be reprogrammed to massproduce modified mining or other mobile robots, including orbiters and rovers, for detailed investigation of the lunar surface. This would augment orbital sensing and intelligent image processing systems (see chap. 2) around the Moon, and could be linked to lunar subsurface explorers and other automated surface prospecting equipment to assist in new resource location, colony siting, and the further acquisition of scientific knowledge. Subselene or subterrene (see discussion of the "Coal Mole" in Heer, unpublished draft notes, Pajaro Dunes Workshop, 1980) mining robots could burrow deep into the lunar or terrestrial crust in search of pockets or veins of useful substances, and then dig them out. A self-replicating manufacturing facility could produce thousands of meter-long robot rovers equipped with cameras, core samplers, and other instrumentation which could survey the entire Moon - or any other planet, for that matter - in just a few years. Such exploration would take a century by more conventional methods. Similarly, due to the low gravity, lack of atmosphere, and relative abundance of energy and raw materials, the Moon is an excellent location for the construction and launching of future generations of interplanetary exploratory spacecraft.

Human resources. The augmentation of human services and the extension and safety of the human habitat is yet another near-term application of self-replicating systems. In principle, it is possible to construct a completely autonomous lunar-based facility, but it may turn out to be inefficient or uneconomical in the future unless a few human beings are present onsite to handle unforeseen problems with the machinery. (Humans are the most compact and efficient general-purpose self-replicating systems of which we have certain knowledge.) Initial crew quarters and supplies can be transported from Earth, but much larger and more pleasant living accommodations could be manufactured in situ by lunar or orbital replicating systems. The inexpensive mass-production of habitation and agricultural modules (or their components) could help open the door to more extensive lunar and space colonization by people, including recreational, industrial, medical, and educational uses, especially because of the abundant solar energy and the expected ability of replicating factories to manufacture and implement a low-cost lunar-surface-to-orbit launch capability. A comprehensive, highly sophisticated automated astronaut search and rescue system may also become necessary as the human population in space begins to grow, with system components mass-produced by SRS.

Presently, there are about 6000 known and tracked pieces of debris orbiting the Earth at various altitudes and inclinations, and countless additional shards which lie below observational thresholds in near-Earth space. These represent an ever-increasing danger of collision with spacecraft. Debris-catchers or "scavengers" mass-produced by SRS technology could be automatically launched into various Earth orbits, seek out and recognize space debris, report ephemerides in the case of satellite-like objects to avoid destruction of operational equipment and, upon go-ahead, collect the debris. Scavengers would be programmed either to enter the Earth's atmosphere after a specified time in orbit and self-destruct, or to return their collections to orbital manufacturing facilities for recycling of high-level components and materials to help build new robots. A more advanced network could offer protection from possible ecological disasters caused by terrestrial meteorite impacts (Alvarez et al., 1980).

Another possibility, however controversial, is meteorological and climatological intervention on both a local and global scale. A number of interesting alternatives were discussed by the participants of the recent Pajaro Dunes Workshop (Heer, unpublished draft notes, 1980), including:

The replicative manufacturing facility needed to economically produce such large numbers of similar system elements would make possible at least a rudimentary global homeostatic environmental control by humanity.

Given the exotic conditions prevailing on the lunar sur face and in space, and the novel materials and processes that may become available, it is highly probable that a self replicating growing lunar facility will be able to economically produce many goods directly for use in space and for export to Earth. What these goods might be is not now certain. However, the economic importance of the telephone, steamboat, airplane, television, office copying machine, etc., during their early stages of development like wise were not at all obvious to most people.

5.4.3 Solar System Applications

The technology of replicating systems will become increasingly important as humanity expands its theater of operations from near-Earth space out to encompass the entire Solar System. Mankind has fallen heir to an incredible treasure trove of nonterrestrial energy and material resources (see sec. 4.2.1). It is likely that replicating machines will provide the only "lever" large enough to explore, and ultimately manipulate and utilize in a responsible fashion, such tremendous quantities of organizable matter. Lacking this advanced automation capability, most of the more ambitious Solar System applications appear uneconomical at best, fanciful at worst.

Observation. Exceedingly far-reaching observational possibilities may become feasible with the advent of SRS technology. Very large baseline interferometry (VLBI) may be attempted with components distributed across the entire Solar System, perhaps located at the stable Trojan points of the Jovian planets or their moons, providing multiplanar baselines of from 1 to 100 AU and complete spherical coverage with the use of out-of-ecliptic robot sensor devices that are mass-manufactured by replicating factories. The solar wind could also be mapped in three dimensions, and by using the entire Sun as a gravitational lens focal lengths on the order of the size of the Solar System can in theory be obtained (Ingel, 1974). This may permit simultaneous observation of the entire celestial sphere across the full spectrum of gravitational radiation using fleets of gravity-wave detectors manufactured by SRS and stationed along the focal plane. A Solar System surveillance network could be constructed to track and warn of objects approaching human habitats, facilities, or the Earth on collision courses, allowing mankind to avoid potentially severe catastrophes.

Exploration. The technologies developed for a generalized lunar autonomous replicative manufacturing facility should be directly applicable in the exploration of all planetary and satellite surfaces. One early possibility is a mission to land a single replicative "seed" on Mars which would then use local materials to produce large numbers of rovers (including, perhaps, fliers, crawlers, walkers, or rollers) and orbiters. A population of 1000 to 10,000 surface rovers each perhaps 100 kg in mass, coupled with a chain of orbital monitors, might continuously monitor and explore the planetary surface and leave stationary probes (active or passive) behind in permanent emplacements. The probes need only have lifetimes on the order of a year or so, since they could constantly be repaired and replenished by the rovers (each of which could last 10 years or more). This system would provide complete surface exploration and continuous status monitoring of all areas on the planet, including temperatures, pressures, wind velocities, seismic events and crustal creeps, meteorite impacts, surface and subsurface compositions, illumination, precipitation, and numerous other phenomena of interest. Automated balloon explorers could be mass-produced and released in Jovian atmospheres, and "trains" of deep solar probes (Heer, unpublished draft notes, 1980) could be hurled into the Sun to obtain direct information on internal conditions there.

Materials retrieval. Replicating systems would make possible very large-scale interplanetary mining and resource retrieval ventures. Nonterrestrial materials could be discovered, mapped, and mined using teams of surface and subsurface prospector robots manufactured en masse in an SRS factory complex. Raw materials could be dug up and sent back to wherever they were needed in the Solar System, or could be refined along the way and the waste slag used as reaction mass, or could be utilized in situ for manufacturing useful products which would then be exported. Atmospheric mining stations could be established on many different planets - Jupiter and Saturn for hydrogen, helium (and rare isotopes potentially useful for fusion power generation, Martin, 1978), and hydrocarbons, using "aerostats" (Parkinson, 1978); Venus for carbon extraction; Europa for water; Titan for hydrocarbons; etc. Comets could be intercepted to obtain large quantities of useful volatiles, and Saturn's rings could be mined for water-ice by large fleets of mass-produced robot craft. Heavy metals may be retrieved in great quantities from asteroids. Replicating systems might manufacture huge mining, processing, even ground-to-orbit and interplanetary transportation capabilities using local materials in surprisingly short periods of time.

The general product factory. The team has proposed the design and construction of an automatic multiproduct replicating lunar factory. The reason for the factory having multiproduct capability is to permit it to be able to respond to any changing requirements in kind or amount of product output. This leads to a still broader concept - the notion of a general product factory.

A general product factory is one which can be instructed to manufacture anything which is physically possible to make. Such a system is the physical realization of von Neumann's "universal constructor" automaton, which can construct anything constructable, given an adequate substrate and the rules of operation of his artificial cell-space universe. In the context of drawing upon planetary resources, we should think of each celestial body in terms of its menu of possible materials and the repertoire of processes theoretically available there (see sec. 4.5.4). The following questions should then be considered:

The notion of a general product factory using asteroidal material was briefly considered at the Pajaro Dunes Workshop. The "Hive," as it was called, would consist of "an autonomous space island 'beehive' of independently intelligent machines . . . specialized in mining and production, experts in planning, navigation and repair." The product of the Hive would be solar power satellites, "asteroids turned into space colonies, vacuum-filled balloons of nickel floated down to a resource-hungry Earth, spaceships, telescopes, or even another Hive." The Hive was envisioned as an independent economy, using raw materials gathered from the Asteroid Belt, refined and processed with solar or fusion energy, then fashioned into useful output by robot hands. Workshop participants suggested a timetable in which the first fully autonomous replicating system could be in operation in the Asteroid Belt by 2040, commencing exponential growth with a replication time of 5 years, resulting in a total of 1000 new Hives available for production by the year 2080.

Table 5.4. - Developmental Milestones For A General Product Factory
  1. Design and construct a system which, when supplied only with parts and subassemblies, can duplicate itself.
  2. Design and construct a system which can duplicate itself, and in addition produce some useful product.
  3. Design and construct a system which, when supplied only with feedstock, can duplicate itself.
  4. Design and construct a system which, when supplied with raw materials only, can duplicate itself.
  5. Design and construct an automated, reprogrammable, multiproduct system which can, from raw materials, duplicate itself.
  6. Design and construct an automated, reprogrammable, multiproduct system which, using only lunar materials and employing only those processes possible in the lunar environment, can duplicate itself.
  7. Design and construct an initial automatic "seed" system which, if placed on the lunar surface, could unpack itself and develop into an automated, reprogrammable, multiproduct replicating system, using lunar resources and lunar processing modes only.
  8. Design and construct an initial seed which can, in the lunar environment, develop and augment itself so as to become a general-product factory, relative to the lunar environment.
  9. Design and construct a seed which can, in an arbitrary planetary environment, develop into a general product factory.

Human resources. From the human standpoint, perhaps the most exciting consequence of self-replicating systems is that they provide a means for organizing potentially infinite quantities of matter. This mass could be so organized as to produce an ever-widening habitat for man throughout the Solar System. Self-replicating homes, O'Neill-style space colonies, or great domed cities on the surfaces of other worlds would allow a niche diversification of such grand proportions as never before experienced by the human species.

SRS provides such a large amplification of matter-manipulating capability that it is possible even to consider the "terraforming" of the Moon, Mars, Venus, and other worlds. Terraforming is a theoretical concept in which a planetary environment with otherwise inhospitable conditions for life is purposefully and artificially altered so that humans may live there with little or no life support equipment. The "traditional" approach is to suggest biological means, such as the proposal to seed the atmosphere of Venus with genetically tailored algae to convert excess carbon dioxide into combined carbon and free oxygen. This would have the incidental salutary effect of lowering the planetary surface temperature so that people could live unaided on the surface. However, it is not known whether biological organisms can be found or developed which are able to withstand present conditions in the Venusian atmosphere.

An alternative approach is to use nonbiological replicating systems which may be far more durable under extreme conditions. A few simple calculations reveal the approximate magnitude and duration of such an enterprise. Consider the terraforming of Mars. For simplicity it is assumed that the planetary crust is largely silicon dioxide and that a general-purpose 100-ton SRS factory "seed" which lands there can replicate itself in 1 year. In just 36 years such a system could theoretically manufacture an SiO2 reduction capability able to release 220,000 tons/sec of pure oxygen into the Martian atmosphere, which in only 60 years is sufficient to produce 4X1017 kg O2. Assuming negligible leakage through the Martian exosphere, this is enough oxygen to establish a 0.1 bar breathable oxygen atmosphere planet-wide - approximately equivalent to normal air on Earth at an altitude of 3000 m (16,000 ft). This plan requires a solar power satellite system in near-Mars orbit with a total generating capacity of about 1017 W, a network which would take less than a year for the finished replicating factory system to produce. The total material thus excavated to terraform Mars is of the order of 1018 kg SiO2, enough to fill a surface depression 1 km deep and 600 km diameter. This is roughly the size of the crater Edom near the Martian equator, or Mare Crisium on the Moon.

Of course, far more efficient methods for terraforming planets may eventually be found, such as Dyson's proposal to mine the Saturnian moon Enceladus for its water-ice and return the material to Mars (Dyson, 1979). But the utility of self-replicating systems is clear, and it appears that terraforming times on the order of one century are conceivable using the SRS approach.

Technology requirements. Additional technology over and above "superautomation" (sec. 5.4.1) will be required for the highly ambitious ventures described in this section using advanced space-based self-replicating systems. The most important new technology in this regard is "closure engineering," discussed in section 5.3.6. Some of the enterprises proposed above are of such large scale that it is difficult to envision a feasible mode of operation with anything less than 100% materials and energy closure and virtually 100% information closure as well. No doubt there exist manufacturing operations which are not economically

viable candidates for total automation in terrestrial industry - in these instances the functions either must be redesigned for full automation or else people must be permanently incorporated as symbionts of a locally teleoperated or remotely human-supervised system. Manufacturing processes developed for terrestrial environments must be re-engineered to accommodate the input and production environments found in space or on the surfaces of other planets, and output streams must be sufficiently flexible to make feasible the notion of a general products factory.

5.4.4 Interstellar and Galactic Applications

Replicating systems technology is the key to exploration and human habitat expansion beyond the confines of the Solar System. Although these kinds of missions necessarily are highly speculative, and admittedly exceed the limits of current or projected technology in many areas, a consideration of possible interstellar and galactic applications is nonetheless a useful exercise because it serves to illustrate the fantastic power and virtually limitless potential of the SRS concept.

Extrasolar exploration. Before humankind can move out into interstellar space, automated probes will scout the way ahead. The distances are so large and the volumes so vast that self-replicating probes are highly desirable, even essential, to adequately and efficiently perform a reconnaissance of extrasolar star systems in a search for human habitable worlds and extraterrestrial life. A preliminary design for a self-reproducing interstellar probe has been presented in the scientific literature (Freitas, 1980a), and another study of the comparative benefits of reproducing and nonreproducing galactic exploration strategies by unmanned probes suggests that search patterns using semi-intelligent automata involving more than about the nearest 100 stars would probably be optimized (in terms of economy and productivity) if self-replicating systems are employed (Valdes and Freitas, 1980). Reproductive probes could permit the direct investigation of the nearest million stars in about 10,000 years and the entire Milky Way galaxy in less than 106 years, starting with a total investment by humanity of a single self-replicating exploratory spacecraft.

The problems in keeping track of, controlling, and assimilating data returned by an exponentially growing number of self-reproducing space probes are staggering. Part of the solution may lie in the use of an extremely high level of autonomy in operations management and reasoning such as discussed in chapter 3 of this report; part may lie in the utilization of high levels of abstraction in the information returned to Earth after the fashion of the World Model sensing and data-processing philosophy articulated in chapter 2. Another major piece of the solution is the development of a hierarchical command, control, and information-gathering architecture in which any given probe communicates directly only with its own parent and offspring. Control messages and exploration reports would pass up and down the chain of ancestral repeater stations erected by earlier generations (Valdes and Freitas, 1980). Certain highly critical but low probability- signals might perhaps be broadcast in an omnidirectional alarm mode to all members of the expanding network (and to Earth) by individual probes which encountered specific phenomena or events - such as the discovery of an extrasolar planet suitable for human habitation or a confrontation with intelligent alien lifeforms or their artifacts.

Extrasolar utilization. Before mankind can venture out among the stars, his artifacts and replicating machines must blaze the trail. Ultimately, however, one can envision freeflying space colonies journeying through interstellar space (Matloff, 1976). Upon reaching some new solar system or other convenient source of raw materials, these mobile habitats would reproduce themselves with the human passengers redistributed among the offspring colonies. The original space habitats would serve as extraterrestrial refuges for humanity and for other terrestrial lifeforms that man might choose to bring along. This dispersal of humankind to many spatially separated ecosystems would ensure that no planetary-scale disaster, and, as people travel to other stars, no stellar-scale or (ultimately) galactic-scale event, could threaten the destruction of all mankind and his accomplishments. Replicating systems may be the only rational means to attempt large-scale astroengineering projects usually relegated to the domain of science fiction, such as the construction of "Dyson Spheres" which enclose and utilize the energy output of entire suns (Dyson, 1959).

The limits of expansion. The expansion of a population of replicating systems in any environment is restricted largely by two factors: (1) replication time, and (2) maximum velocity of the outer "envelope" which defines the physical extent or dispersion of the population. No population can accrue at a faster rate than its components can reproduce themselves. Similarly, no population can disperse faster than its medium will permit, no matter how fast components are manufactured - assuming number density remains essentially constant, corresponding to continuous maximum utilization of the environment. Neither factor may be ignored during any phase of population growth.

If envelope expansion velocity does not constrain a population because components are produced only relatively very slowly, then that population will experience exponential multiplication according to:

N(T) = exp(T/t) (1)

where N(T) is the number of replicating units comprising the population at time T (replication starts at T = 0) and t is the replication time per unit, assumed constant. On the other hand, if unit reproduction is so swift that multiplication is not constrained by replication time, then the population can grow only as fast as it can physically disperse that is, as fast as the expansion velocity of the surface of its spherical outer envelope - according to:

N(T) = 4/3 π d(VT)3 (2)

where V is peak dispersion velocity for individual replicating units at the periphery and d is the number density of useful sites for reproduction. Expansion cannot exceed the values for N(T) given either by equations (1) or (2) at any time T, provided all replication sites receive maximum utilization as stipulated (e.g., constant number density of units).

Populations of machines expanding across the surfaces of worlds with replication times on the order of 1 year will not achieve mean envelope growth speeds in excess of a few meters per hour, even in later phases of extreme enlargement when the population of SRS covers a large fraction of the available planetary surface. This figure is well within anticipated nominal ground transport capabilities, so exponential extension should remain largely velocity-unconstrained on such bodies if replication time remains constant at greater population sizes.

Similarly, three-dimensional populations of replicating systems in interplanetary space using Solar System materials and solar energy ultimately are restricted to spherical circumstellar shells where SRS units can collect virtually all energy radiated by the Sun. If a "Dyson Sphere" of 100-ton replicating "seed" units is assembled near the orbit of Earth, approximately one terrestrial mass is required to manufacture the more than 1019 individual units needed to completely enclose the star. But maximum expansion velocity even in this case never exceeds about 100 m/sec, hence interplanetary replicating systems as well in theory may spread at purely exponential rates.

In the interstellar realm, however, the situation is far more complex. Depending on the maximum dispersal velocity and interstellar probe replication time, either equation (1) or (2) may control. Figure 5.24 compares pure exponentiation and dispersal speed effects for t = 1 year (see sec. 5.3.4) and t = 500 years (Freitas, 1980a), and for V = c (since the theoretical maximum envelope expansion rate is the speed of light) and V= 10%c (Martin, 1978) for an assumed homogeneous stellar distribution of "habitable" star systems (taken as 10% of the total) in the galactic disk. In most cases, exponential multiplication soon is halted by the speed-of-light barrier to dispersion, after which the SRS population expansion proceeds only polynomially.

Figure 5.24 -
Limits to exponential and polynomial expansion of self-replicating interstellar probe populations dispersing throughout the galactic disk.

Technology requirements. In order to sustain the expansion of a potentially infinite replicating system, new dispersal mechanisms must be developed. Initially, self-replicating machines or their "seeds" must be capable of motion across a planetary surface or through its atmosphere or seas. Later, interplanetary, interstellar, and, ultimately, intergalactic dispersal mechanisms must be devised. Supplies of energy, stored and generated, must be established if extrasolar spacecraft are to survive in the depths of interstellar space far from convenient sources of power (such as stars) for a major portion of their lives. The technologies of command, control, and communication over stellar and galactic distances ultimately also must be developed.

5.4.5 Applications to Basic Research

In addition to specific applications of replicating systems technology to future missions in space, a number of applications to basic research in biology, computer science, and automata theory have been identified by the team. These are discussed below.

Automaton theory. Automaton theory is the abstract and precise study of all mechanistic devices and processes. At times this has been restricted to the theory of discrete and deterministic machines with a fixed finite number of states. In this narrow sense it is the abstract mathematical counterpart of physical devices such as existing digital computers. In the broadest sense, though, automaton theory can include the study of all mechanisms, discrete or continuous, deterministic or probabilistic or even indeterministic, with a fixed, variable, or indefinitely large number of possible states. Included in this wider definition is the notion of devices which can alter the number of their states by growth or by contraction in respect to certain of their organs, much like the way a Turing machine or a pushdown automaton (or a linear-bounded automaton) can increase or decrease the number of its states by increasing or decreasing the length of its memory tape - but also can grow by increasing or decreasing the numbers of its more active computing components. This is representative of machines which can construct or dismantle other machines.

These machines can not only increase their memory capacity but can augment their computing power by the construction of additional active computing organs (registers, control units, etc.) and by constructing machines separate from themselves, including duplicates of themselves. Von Neumann had begun to develop a general and logical theory of automata which would have embraced all these machine types. Automaton theory has, however, never achieved the generality he sought, at least not in the sense he seems to have intended.

The very general theory of automata has become increasingly abstract, moving from describing mechanistic processes in terms of algebraic concepts such as groups and semigroups to employing category theory, the most abstract and general of algebraic theories. Although a certain level of understanding of what mechanisms might exist has thereby been developed, the applicability of such approaches to the design of complex systems of automata is very slight or nonexistent. In this regard, von Neumann once lamented that "... at a great distance from its empirical source, or after much abstract inbreeding, a mathematical subject is in danger of degeneration.... Whenever this stage is reached, the only remedy seems to me to be the rejuvenating return to the source - the reinjection of more or less directly empirical ideas." (von Neumann, 1966).

It may be that an effort to actually design and implement a system of machines which can construct more machines like themselves would encourage theorists again to attempt to develop a very general automaton theory including as a part of its subject matter the spatial and communicatory interactions of vast and increasing numbers of submachines. (Perhaps the automatic telephone system provides us with the closest physical analogy to such systems, aside from the analogy of human societies themselves.)

Such a theory would enable one to ask what is the best organization of a system of (potentially) arbitrarily increasing numbers of active components, arranged in various spatial geometries. How might the interacting activities of vast numbers of submachines be optimized? What rules of interaction and of interconnection can be imposed on such a system in order to attain efficient and stable behaviors? What are the safest physical and behavioral interactions, and which lead to instabilities and pathologies?

A general theory would also take as part of its subject matter the flow of parts and materials. It might, like the von Neumann cellular system, treat the creation and flow of materials and the movement of machinery as a form of information flow. It might distinguish information, materials (raw materials, feedstock, and parts) and the movement and siting of machines, but treat them in an identical format so that tradeoffs and exchanges in these categories could be computed (while retaining the essential differences among these types of flow important to the working of the system). The theory would answer such questions as: When will more information be the best substitute for more parts or more feedstock? Under what conditions in the vast assemblage of machines should parts be made anew, from raw materials and feedstock, and when should information or already finished parts be employed to the same purpose? When should machines which are likely to fail be abandoned? When should machines in the assemblage which are still in good condition nevertheless be shut down, moved, sacrificed for parts or dismantled, or sealed off? Under what local and global conditions should submachines be retired, repaired, or replaced?

Theoretical biology. Machines which can construct machines, and machines which can construct replicas of themselves, display behavior which in many ways is analogous to that of natural organisms. Furthermore, as machines are designed to examine their own structure and the structures of other machines, to repair themselves and other machines, and generally to become more autonomous and more reliable, the analogies become even more apparent.

The ways in which machines carry out these processes of growing, repairing, regenerating, and reproducing may or may not be similar to those carried out by natural organisms - which, in many cases, are not yet even known.

One goal of theoretical biology is to develop an under standing of the mechanisms of living systems, to the point where these systems can be characterized in a precise mathematical fashion (Miller, 1978). To attain such a characterization one needs a good intuitive feeling for the full possible range of lifelike forms. For example, a theory of biology that takes as its subject matter only Earth-evolved forms would be as unlikely to be capable of providing adequate explanation for non-Earth forms as were attempts to account for the forms of extant organisms quite apart from their extinct progenitors.

It seems, therefore, likely that an adequate explanatory theory of biology of any elegance and simplicity must embrace not only all biological forms which presently exist, but all those which have ever existed, or will exist, or could exist. Indeed, the proper subject matter for a true theoretical biology in its broadest sense would be the study of life like behavior wherever it occurs whether now, or in the past, or the future; whether on Earth or elsewhere in the universe and whether it is exemplified in artificial or natural forms (Freitas, 1980b), a field of study termed "xenobiology" by one author (Freitas, 1981). This suggests that research on complex automata able to reconstruct, reproduce, and repair themselves might serve as a fertile source of hypotheses as to the logical control and organizational aspects of how living organisms in fact carry out these processes. Such explanatory hypotheses can apply to life like systems generally and have the advantage that they are likely to be simpler and more elegant than the necessarily ad hoc explanations of behavior for the particular organ isms of particular worlds, at particular times.

Thus, research in self-growing and self-replicating machine systems can be viewed as a contribution to, even as F a central part of, a true theoretical biology which takes as its subject matter not merely the evolved, naturally occur ring living organisms of Earth, but lifelike mechanisms, natural or artificial, having existed or possible, wherever in L the universe they might arise.

Design of biological and hybrid organisms. The forms and processes of artificial organism-like systems are not bound to follow the particular structure and logical organizations of known naturally evolved organisms. As the design of increasingly complex artificial systems capable of drawing materials and energy from natural surroundings and possessing more and more organism-like properties proceeds, it may become apparent that there are artificial organism functions which, if embodied in biological organisms, would be of value. With advances in "genetic engineering" it may become possible to create new biological forms, possessing the desired features.

Just as the design of artificial mechanisms can be inspired by contemplation of evolution's apparent solutions to various design problems, so might new biological systems also be created, drawing upon designs originally conceived for artificial systems - a kind of inverse bionics. Taking this a step further, one can envisage as a research goal the gradual elimination of the perhaps arbitrary line now drawn between artificial and natural organisms, and the consideration of a more deliberate systematic investigation of the creation of hybrid biological-mechanical systems.

Experimental evolution. Studies of form and function in biological and artificial systems may contribute to an understanding of the design and construction of both biological and mechanical organisms. This interdisciplinary exchange should not be limited to studies of the relationship between individual classes of lifelike entities, but should also extend to studies of the consequences of large numbers of such entities interacting and competing for resources. Replications of programs and creation of new machines (including replicas), and compounds and combinations of initially existing machines, can be a feature of the proposed machine replicating systems. It seems clear that development of a science of evolving systems is needed (Miller, 1978). (This would again be a part of a very general "true" theoretical biology, which takes all possible lifelike systems as its subject matter.)

For example, one putative value of sexual over asexual reproduction is the enormously increased mobility of genetic variation in the species population. This widely available variation tends to ensure that environmental changes can be accommodated or exploited with great swiftness by at least some members of the population (Smith, 1978). In a "designed" universe, one is free to consider the advantages (if any) of three or more sexes (Freitas, 1980c; Smith, 1978) or of the consequences of other, even more radical redesignings of existing natural systems. In particular, the actual behavior of largely autonomous growing replicating machine "species" with differing capabilities and reproducing strategies certainly should be an object of study by evolutionary biologists who might be able to predict the forms which would persist and come to dominate in systems left unperturbed by external pressures and commands.

The existence of large interacting populations of entities whose "genetics" are precisely known, but whose global behavior over time cannot readily be predicted, may be of great experimental value to evolutionary biologists. At present, computer simulation is the usual tool of choice for such problems. However, if the physical creation of machine populations becomes sufficiently inexpensive, experimental situations might be created in remote nonterrestrial regions. Machine growth and population changes could be monitored over time for their adherence (or not) to hypothesized consequences. The advantage of this approach over the computer simulation would be in the much greater detail and fidelity to real situations, and the consequent likelihood of serendipitous useful observation.

Machine intelligence architecture. Very general symbol manipulating devices (such as stored program computers) are at the heart of efforts to demonstrate that machines can exhibit behavior which in human or other animals would be considered intelligent. In one sense, such devices are computationally universal. That is, certain mathematical technicalities aside, they can carry out any arbitrary Turing machine computation and, accepting the Church-Turing Thesis, can also carry out any algorithmic process. Thus, if any machine can be intelligent one need look no further than to a general-purpose computer, for there is some program which will cause the machine to display the desired intelligent behavior. This is so even if one insists that brains, for example, are machines, but are not at all like digital computers. This is because digital computers, again accepting the Church-Turing Thesis, can be programmed to simulate any known mechanistic process to any fineness of detail, whether the process of interest be analog, frequency coded, probabilistic or other.

Even though ordinary computing machines do not, for example, reproduce themselves, they can be programmed to simulate the behavior of machines that do in fact reproduce. From this point of view, the concept of machines which posses the power to construct other machines and to replicate themselves can be represented to any degree of detail in the computation of an ordinary general-purpose computing machine which cannot itself reproduce. Even though existing general-purpose cannot generally inspect themselves and draw conclusions therefrom, computers can be programmed to simulate such unlikely machine actions if such a simulation is thought useful or interesting. Hence, the construction of the kinds of machines considered here - machines that can compute, construct, reproduce, and inspect, repair, simulate, and observe both themselves and other machines - would not enlarge what a general-purpose device can in theory already do but rather our perception of their capability to exhibit more sophisticated mindlike behavior.

It should also be noted that machines can be designed and constructed so as to do things beyond what any known evolved organism (including man) can do. We are already aware of this superiority of machines in regard to strength, speed, accuracy, flight, and the like. There are already many ways in which machines can be designed and constructed so as to exceed human mental capabilities for specific tasks.

For example, though we are constantly reminded of the social value of being able "to see ourselves as others see us," our evolutionary history has left us with only a very limited capacity for accurate introspection and self-examination - though in this respect we admittedly exceed all other known evolved creatures. Machines, however, can be designed to secure far greater access to their internal structure and states than we are ever likely to posses as individuals, and this capacity might mean that machines can be programmed to achieve mindlike powers far beyond ours. A trivial case of this "introspective" superiority of machines is seen in their ability to "remember." Computers can be programmed to methodically search all of their memory with a thoroughness that can evoke human envy.