The word ‘nanotechnology’ covers a broad range of scientific and technical disciplines. Fortunately, it will not be necessary to consider each separately in order to discuss the global catastrophic risks of nanotechnology, because most of them, which we will refer to collectively as nanoscale technologies, do not appear to pose significant global catastrophic risks. One discipline, however, which we will refer to as molecular manufacturing, may pose several risks of global scope and high probability.
The ‘nano’ in nanotechnology refers to the numeric prefix, one-billionth, as applied to length: most structures produced by nanotechnology are conveniently measured in nanometres. Because numerous research groups, corporations, and governmental initiatives have adopted the word to describe a wide range of efforts, there is no single definition; nanotechnology fits loosely between miniaturization and chemistry. In modern usage, any method of making or studying sufficiently small structures can claim, with equal justice, to be considered nanotechnology. Although nanoscale structures and nanoscale technologies have a wide variety of interesting properties, most such technologies do not pose risks of a novel class or scope.
Interest in nanotechnology comes from several sources. One is that objects smaller than a few hundred nanometres cannot be seen by conventional microscopy, because the wavelength of visible light is too large. This has made such structures difficult to study until recently. Another source of interest is that sufficiently small structures frequently exhibit different properties, such as colour or chemical reactivity, than their larger counterparts. A third source of interest, and the one that motivates molecular manufacturing, is that a nanometre is only a few atoms wide: it is conceptually (and often practically) possible to specify and build nanoscale structures at the atomic level.
Most nanoscale technologies involve the use of large machines to make tiny and relatively simple substances and components. These products are usually developed to be integral components of larger products. As a result, the damage that can be done by most nanoscale technologies is thus limited by the means of production and by the other more familiar technologies with which it will be integrated; most nanoscale technologies do not, in and of themselves, appear to pose catastrophic risks, though the new features and augmented power of nano-enabled products could exacerbate a variety of other risks.
Molecular manufacturing aims to exploit the atomic granularity and precision of nanoscale structures, not only to build new products, but also to build them by means of intricate nanoscale machines, themselves the products of molecular manufacturing techniques. In other words, precisely specified nanoscale machines would build more nanoscale machines by guiding molecular (chemical) processes. This implies that once a certain level of functionality is reached (specifically, the level at which a generalpurpose machine can build its physical duplicate and variants thereof), the nanoscale can become far more accessible, products can become far more intricate, and development of further capabilities could be rapid. The molecular manufacturing approach to nanotechnology may unleash the full power of the nanoscale on the problems of manufacturing – creating products of extreme power in unprecedented abundance. As we will see, the success of this approach could present several global catastrophic risks.
It should not be overlooked that molecular manufacturing could create many positive products as well. Stronger materials and more efficient mechanisms could reduce resource usage, and non-scarce manufacturing capacity could lead to rapid replacement of inefficient infrastructures. It is also plausible that distributed (even portable), general-purpose, lower-cost manufacturing could allow less-developed and impoverished populations to bootstrap themselves out of deprivation. Reductions in poverty and resource constraints could work to reduce some of the risks described in this chapter. Although the focus here is on the most severe dangers, we support responsible development of molecular manufacturing technology for its benefits, and we promote study and understanding of the technology as an antidote to the threats it poses.
A common characteristic of nanoscale technologies is that they are not suitable for manufacturing finished products. In general, the method of production for nanoscale components is simple and special-purpose, and can only build a material or component that subsequently must be included in a larger product through other manufacturing methods. For example, a carbon nanotube may be deposited or grown by an innovative process on a computer chip, but the chip then must be packaged and installed in the computer by traditional means. Ultraviolet-blocking nanoparticles may be manufactured by an innovative process, but then must be mixed into sunscreen with mostly ordinary ingredients. A related point is that nanoscale technologies make use of machines far larger in scale than their output.
Access to the nanoscale is typically indirect, painstaking, or both, and most manufacturing processes are unable to convey detailed information to the nanometre realm. The resulting nanoscale products, therefore, cannot be highly structured and information-rich, even by comparison with today’s manufactured products; every feature must be either miniscule, highly repetitive, or random. Large quantities of identical or nearly identical particles and molecular constructs can be built, but they will be either amorphous, highly ordered to the point of simplicity (like a crystal), organized by a prebuilt template, or have a partially random structure influenced by only a few parameters. Computer chips, which now contain millions of nanoscale features, are an exception to this, but they can only be built with the aid of ‘masks’ that are immensely expensive, slow, and difficult to produce. This is another reason why nanoscale technologies generally are not suitable for manufacturing finished products.
Because the outputs of nanoscale technologies can typically be used only as inputs to traditional manufacturing or material-use steps, their scope and impact necessarily will be limited by the non-nanotech components of the products and the non-nanotech manufacturing steps. For the most part, nanoscale technologies can be viewed as an addition to the existing array of industrial technologies. In some cases, nanoscale structures including nanoparticles may be released into the environment. This could happen deliberately as part of their use, due to accidents including unintended product destruction, or as part of the product lifecycle (including manufacturing processes and product disposal).
The risks of nanoscale technologies appear in two broad classes, which are analogous to existing industrial risks. The first class is risks resulting from the availability and use of the eventual products. In this way, the nanoscale technology contributes only indirectly, by making products more powerful, more dangerous, more widely used, and so on. The second class is risks resulting from new materials that may cause inadvertent harm.
Depending on the application, risks may be increased by making improved products. Improved medical technology frequently must confront ethical issues. Some applications of nanoscale technology (such as using nanoparticles to treat cancer) will have strong demand and little downside beyond the standard medical issues of creating good treatment protocols. The use of nanoscale technologies for military medicine (ISN, 2005) or advanced goals such as anti-aging may be more controversial. But in any case, these risks and issues are only incremental over existing risks; they are not new classes of risk, and there do not appear to be any global catastrophic risks in nanoscale health care technology.
Nanoscale technologies may contribute indirectly to the development of weaponized pathogens, which arguably could lead to an existential risk. Smaller and more powerful research tools could be used to fine-tune or accelerate the development process. However, the same tools also would be useful to counter a biological attack. The SARS virus was sequenced in only six days in 2003 (Bailey, 2003). The development of sufficiently fast, sensitive, and inexpensive tests could greatly reduce the threat of almost any infectious disease. Thus, it is not yet clear whether nanoscale technologies will increase or reduce the risk of globally catastrophic pandemics.
Improved computers may produce new classes of risk, some of which arguably might be existential. Surveillance technology is improving rapidly, and will continue to improve as data mining, data processing, and networking become cheaper. A state that knows every movement and action of its populace would have power unequalled by history’s most repressive regimes. Misuse of that power on a global scale could pose a catastrophic risk. However, this risk cannot be blamed entirely on nanoscale technologies, since it seems likely that sufficiently powerful computers could be developed anyway. Another possible risk stemming from improved computers is artificial intelligence (see Chapter 15, this volume).
Environmental or health risks from inadvertent releases of nanoscale materials will not be existential. That is not to say that there is no risk. Just like any industrial material, new nanoscale materials should be evaluated for toxicology and environmental impacts. Some nanoparticles may have high stability and environmental persistence, may migrate through soils or membranes, and may be chemically active; these are all reasons to study them closely. Information about the health and environmental impacts of nanoparticles is still far from complete, but at this writing their risks seem to fall into the familiar range of industrial chemicals. Some nanoparticles will be mostly harmless, while others may be quite toxic. Although it is possible to imagine scenarios in which the manufacture of sufficient quantities of sufficiently toxic particles would threaten the world in the event of accidental release, such scenarios seem unlikely in practice.
As nanotechnology develops, it is becoming possible to plot a continuum between near-term nanoscale technologies and molecular manufacturing. It begins with a convergence of increasingly complicated molecules, improved facility in mechanical (e.g., scanning probe) chemistry, and more powerful and reliable simulations of chemistry. It is now possible for mainstream nanotechnology researchers to imagine building machine-like constructions with atomic precision. Once this is achieved, the next step would be to harness these machines to carry out an increasing fraction of manufacturing operations. Further advances in machine design and construction could lead to integrated large-scale manufacturing systems building integrated largescale products, with both the manufacturing systems and the products taking advantage of the (expected) high performance of atomically precise nanoscale machinery and materials.
The molecular manufacturing approach is expected by several researchers, including Robert Freitas, Ralph Merkle, and Eric Drexler, to lead to ‘nanofactories’ weighing a kilogram or more, capable of producing their own weight in product from simple molecules in a matter of hours. Because of the immense number of operations required, the factory is expected to be entirely computer-controlled; this is thought to be possible without advanced error-correcting software because of the high precision inherent in molecular construction (Phoenix, 2006). And because the manufacturing process would be general-purpose and programmable, it is expected to be able to produce a range of products including nanofactories and their support structure.
Any step along the pathway to molecular manufacturing would represent an advance in nanotechnology. But the point at which nanofactories become able to build more nanofactories seems particularly noteworthy, because it is at this point that high-tech manufacturing systems could become, for the first time in history, non-scarce. Rather than requiring cutting-edge laboratory or industrial equipment to produce small amounts of nanoscale products, a nanofactory would be able to build another nanofactory as easily as any other product, requiring only blueprints, energy, and feedstock. Thus, the cost of nanofactorybuilt products would owe very little to either labour or (physical) capital.
The supply of nanofactory-built products, as well as their cost, would depend on which resource – information, feedstock, or energy – was most difficult to obtain at point of use (the nanofactory itself).
• Information costs very little to store and copy; although it might be limited by regulations such as intellectual property laws, this would not represent a natural limit to the use of nanofactory-based manufacturing.
• Feedstock would be a small molecule, used in bulk. The molecule has not been specified yet, so its cost and availability cannot be determined, but the relatively low cost and high production of manufactured gases such as acetylene and ammonia indicates that nanofactory feedstock may contribute only a few dollars per kilogram to the price of the product.
• The largest cost of product fabrication may be energy. A preliminary analysis of a primitive nanofactory architecture was carried out recently (Phoenix, 2005), which calculated that building a product might require approximately 200 kWh/kg. Although this is significant, it is comparable with the energy cost of refining aluminum, and the material fabricated by the nanofactory would probably have significantly greater strength than aluminum (comparable to carbon nanotubes), resulting in lower weight products.
As general-purpose manufacturing systems, it is quite possible that nanofactories would be able to build feedstock processing plants and solar collectors. In that case, the quantity of products built would seem to be potentially unlimited by any aspect of the present-day industrial infrastructure, but rather by the resources available from the environment: sunlight and light elements such as carbon, none of which is in short supply. The time required for a nanofactory to build another nanofactory might be measured in days or perhaps even hours (Phoenix, 2005). With strong, lightweight materials, the time required to build another nanofactory and all its supporting infrastructure, thus exponentially doubling the available manufacturing capacity, might be as little as a few days (though without detailed designs this can be only speculation).
Nanofactory-built products (including nanofactories) would potentially enjoy a number of advantages over today’s products. Certain atomically precise surfaces have been observed to have extremely low friction and wear (Dienwiebel et al., 2004), and it is hoped that these attributes could be constructed into nanomachines. Smaller machines work better in several respects, including greater power density, greater operating frequency, far greater functional density, and greater strength with respect to gravity (Drexler, 1992). Being built at the scale of atoms and molecules, the machines would be able to perform a number of medically significant procedures (Freitas, 1999). Atomically precise materials in appropriate structures could be far stronger than today’s building materials (Drexler, 1992).
Building on the basic set of capabilities analysed by Drexler (1992), a number of researchers have worked out designs for products in varying degrees of detail. Freitas has analysed a number of basic capabilities relevant to medical devices, and has described several medical nanorobots in detail. Freitas has proposed devices to collect and neutralize harmful microbes (Freitas, 2005), to supplement the gas-transport function of blood (Freitas, 1998), and to replace damaged chromosomes (Freitas, 2007), among others. For everyday use, masses of relatively simple robots a few microns wide could cooperate to reconfigure themselves in order to simulate a wide range of shapes and conditions (Hall, 1996). Hall has also undertaken preliminary investigation of small aircraft with several innovative features.
Detailed designs for larger products have been scarce to date. An exception is the architectural and scaling study of nanofactory design mentioned earlier (Phoenix, 2005); this 85-page paper considered numerous factors including energy use and cooling, physical layout, construction methods, and reliability, and concluded that a kilogram-scale, desktop-size, monolithic nanofactory could be built and could build duplicates in a few hours. Larger nanofactories building larger products, up to ton-scale and beyond, appear feasible. But in the absence of detailed plans for building and assembling nanomachines, it has not generally been feasible to design other products at high levels of detail. Nevertheless, analyses of extreme high performance at the nanoscale, as well as expectations of being able to integrate nearly unlimited numbers of functional components, have led to a variety of proposed applications and products, including the rapid construction of light weight aerospace hardware (Drexler, 1986).
General-purpose manufacturing would allow the rapid development of new designs. A manufacturing system that was fully automated and capable of making complete products from simple feedstock could begin making a new product as soon as the design was ready, with no need for retraining, retooling, or obtaining components. A product designer developing a new design could see it produced directly from blueprints in a matter of hours. The cost of manufacture would be no higher for a prototype than for full-scale manufacturing. This would benefit designers in several ways. They would be able to see and evaluate their products at several stages in the development process. They would not have to spend as much effort on getting the design right before each prototype was produced. They would not have to make an additional design for full-scale manufacture. With these constraints removed, they could be considerably more aggressive in the technologies and techniques they chose to include in a product; if an attempted design failed, they would lose relatively little time and money.
Once a design was developed, tested, and approved, its blueprint could be distributed at minimal cost (as computer files on the Internet), ready for construction as and where needed. This means that the cost to deliver a new product also would be minimal. Rather than having to manufacture, ship, and warehouse many copies of a design whose success is not assured, it could be built only when purchased. This would further reduce the risk associated with developing innovative products. If a design was successful, it could be manufactured immediately in as many copies as desired.
Because weapons figure in several global catastrophic risks, it is necessary to discuss briefly the kinds of weapons that might be built with molecular manufacturing. Increased material strength could increase the performance of almost all types of weapons. More compact computers and actuators could make weapons increasingly autonomous and add new capabilities. Weapons could be built on a variety of scales and in large quantities. It is possible, indeed easy, to imagine combining such capabilities: for example, one could imagine an uncrewed airplane in which 95% of the dry weight is cargo, the said cargo consisting of thousands of sub-kilogram or even sub-gram airplanes that could, upon release, disperse and cooperatively seek targets via optical identification, and then deploy additional weapons capabilities likewise limited mainly by imagination.
The size of the gap between such speculation and actual development is open to debate. Smart weapons presumably would be more effective in general than uncontrolled weapons. However, it will be a lot easier to cut-and-paste a motor in a computer-aided design programme than to control that motor as part of a real-world robot. It seems likely, in fact, that software will require the lion’s share of the development effort for ‘smart’ weapons that respond to their environment. Thus, the development of novel weapon functionality may be limited by the speed of software development.
To date, there does not appear to have been a detailed study of molecular manufacturing-built weapons published, but it seems plausible that a single briefcase full of weaponry could kill a large percentage of a stadium full of unprotected people (to take one scenario among many that could be proposed). Small robots could implement some of the worst properties of land mines (delayed autonomous action), cluster bombs (dispersal into small lethal units), and poison gas (mobile and requiring inconvenient degrees of personal protection). A wide variety of other weapons may also be possible, but this will suffice to put a lower bound on the apparent potential destructive power of molecular manufacturing-built products.
An idea that has caused significant concern (Joy, 2000) since it was introduced two decades ago (Drexler, 1986) is the possibility that small, self-contained, mobile, self-copying manufacturing systems might be able to gain sufficient resources from the ecosphere to replicate beyond human control. Drexler’s original concern of accidental release was based on a now-obsolete model of manufacturing systems (Phoenix and Drexler, 2004). However, there is at least the theoretical possibility that someone will design and release such a thing deliberately, as a weapon (though for most purposes it would be more cumbersome and less effective than non-replicating weapons) or simply as a hobby. Depending on how small such a device could be made, it might be quite difficult to clean up completely; furthermore, if made of substances not susceptible to biological digestion, it might not have to be very efficient in order to perpetuate itself successfully.
Molecular manufacturing, if it reaches its expected potential, may produce three kinds of risk: (1) as with nanoscale technologies, molecular manufacturing may augment other technologies and thus contribute to the risks they present; (2) molecular manufacturing may be used to build new products that may introduce new risk scenarios depending on how they are used by people; (3) molecular manufacturing may lead to self-perpetuating processes with destructive side effects. At the same time, however, molecular manufacturing may help to alleviate several other catastrophic risks.
Rapid prototyping and rapid production of high-performance nanoscale and larger products could speed up innovation and R&D in a wide variety of technologies. Medicine is an obvious candidate. Aerospace is another; constructing new airframes and spacecraft tends to be extremely labourintensive and expensive, and the ability to rapid-prototype finished test hardware at relatively low cost may allow significantly more aggressive experimentation. If molecular manufacturing is developed within the next 20 years, then it would be capable of building computers far more advanced than Moore’s Law would predict (Drexler, 1992). Each of these technologies, along with several others, is associated with catastrophic risks; conversely, medicine and aerospace may help to avert risks of plague, asteroid impact, and perhaps climate change.
If molecular manufacturing fulfils its promise, the products of molecular manufacturing will be inexpensive and plentiful, as well as unprecedentedly powerful. New applications could be created, such as fleets of high-altitude uncrewed aircraft acting as solar collectors and sunshades, or dense sensor networks on a planetary scale. General-purpose manufacturing capacity could be stockpiled and then used to build large volumes of products quite rapidly. Robotics could be advanced by the greatly increased functional density and decreased cost per feature associated with computer-controlled nanoscale manufacturing. In the extreme case, it may even make sense to speak of planet-scale engineering, and of modest resources sufficing to build weapons of globally catastrophic power.
21.2.3.1 Global war
If molecular manufacturing works at all, it surely will be used to build weapons. A single manufacturing system that combines rapid prototyping, mass manufacturing, and powerful products could provide a major advantage to any side that possessed it. If more than one side had access to the technology, a fast-moving arms race could ensue. Unfortunately, such a situation is likely to be unstable at several different points (Altmann, 2006). A number of players would want to enter the race. Uncertainty over the future, combined with a temporary perceived advantage, could lead to preemptive strikes. And even if no one deliberately launched a strike, interpenetrating forces with the necessary autonomy and fast reaction times could produce accidental escalation.
During the Cold War period, the world’s military power was largely concentrated in two camps, one dominated by the United States and the other by the Soviet Union. As both sides continued to develop and stockpile massive amounts of nuclear weapons, the doctrine of Mutually Assured Destruction (MAD) emerged. Full-scale war could have resulted in the annihilation of both powers, and so neither one made the first move.
Unfortunately, many of the factors that allowed MAD to work in deterring World War III may not be present in an arms race involving nano-built weapons:
1. The Cold War involved only two primary players; once a rough parity was achieved (or perceived), the resulting standoff was comparatively stable. Unless the ability to make nano-weapons is somehow rigorously restricted, a large number of nations can be expected to join the nanotech arms race, and this could create a situation of extreme instability.
2. Acquiring the capability to use nuclear weapons is an expensive, slow, and difficult process. It is therefore relatively easy to track nations that are seeking to gain or expand a nuclear fighting potential. By contrast, the capability to make weapons with molecular manufacturing will be very inexpensive, easy to hide (in the absence of near-total surveillance), and can be expanded rapidly. A ‘starter’ nanofactory could be smuggled from place to place more easily than a stick of gum, then used to build more and larger nanofactories.
3. Rapidly shifting balances of military power may create an atmosphere of distrust. Greater uncertainty of the capabilities of adversaries could foster caution – but it also could increase the temptation for preemptive strikes to prevent proliferation.
Following the collapse of the Soviet Union, another factor emerged to keep the likelihood of full-scale war at a low level. This was the growing economic interdependence of nations: the global economy. It is said that democracies rarely attack one another, and that’s also true for trading partners (Lake, 1996; Orr, 2003).
But the proliferation of civilian molecular manufacturing capability could reduce global trade (McCarthy, 2005), at least in physical goods. Any nation with nanofactories would be able to provide virtually all their own material needs, using inexpensive, readily available raw materials. As economic interdependence disappears, a major motivation for partnership and trust also may be substantially reduced (Treder, 2005). Trade in information may represent an important economic exchange between countries that have compatible intellectual property systems but cannot be counted on to stabilize relations between any given pair of nations; indeed, the ease of copying information may lead to increased tensions due to ‘theft’ of potential earnings.
Today, the destruction wrought by war is an incentive not to engage in it; a nation may very well make more profit by trading with an intact neighbour than by owning a shattered neighbour. But if molecular manufacturing enables rapid inexpensive manufacture, it might be possible to rebuild quickly enough that war’s destruction would represent less of an economic penalty for the victor.
Better sensors, effectors, communications, and computing systems – made at very low cost – may enable deployment of teleoperated robot ‘soldiers’ able to occupy territory and carry out missions without risk to human soldiers. There could be a wide variety of such ‘soldiers’, including ground-, water-, and airbased robots in a wide range of sizes from a few grams to many tons. Each robot might be directly teleoperated or partially autonomous (e.g., able to navigate to a programmed location). It is important to realize that, while stronger materials and greater power efficiency would increase the performance of the robot somewhat, computers many orders of magnitude, more compact and powerful than today’s, would enable algorithms that are barely in research today to be deployed in even the smallest robots. (Robots smaller than a gram might have difficulty in locomotion, and small robots would also be limited in the payloads they could carry.)
A consequence of removing humans from the battlefield is to make war potentially less costly to the aggressor, and therefore more likely. Removal of humans from the battlefield (at least on one side), and the likely advances in ‘less lethal’ weapons aided by advanced nanotechnology, could reduce the moral sanctions against wars of aggression. It is tempting to think that removal of humans from the battlefield would make wars less destructive, but history shows that this argument is at best overly simple. Automated or remotecontrolled weapons, rather than removing humans from the field of battle, instead may make it easier to take the battlefield to the humans. Although these new weapons might shift the focus of conflict away from conventional battlefields, new battlefields would likely develop, and many of them could overlap and overwhelm civilian populations.
Finally, big wars often arise from small wars: with trust factors rapidly declining, the potential for escalation of internecine or regional wars into larger conflagrations will be substantially higher. A significant (and sustainable) imbalance of power, wherein some government (whether national or international) has access to far more force than any of the combatants, could prevent small wars and thus their growth into big wars. However, recent history shows that it is quite difficult for even a powerful government to prevent local conflict or civil war, and it is far from certain that even the technological power of molecular manufacturing, as used by fallible political institutions, would be able to prevent small wars.
The desire to avoid unexpected destructive conflict would in theory provide a counterbalance to these destabilizing factors. But recognition of the danger probably will not be sufficient to avoid it, particularly when failing to develop and deploy advanced weapon systems may be tantamount to unilateral disarmament. An alternative, which might be especially attractive to an actor perceiving that it has a definite lead in nano-based military technology, could be consolidation of that lead by force. Attempts to do so might succeed, but if they did not, the outcome may well be exactly the destructive escalating conflict that the pre-emptive actor wanted to avoid.
21.2.3.2 Economic and social disruption
It is unclear at this point how rapidly molecular manufacturing might displace other kinds of manufacturing, and how rapidly its products might displace established infrastructures and sources of employment. A sufficiently general manufacturing technology, combined with advanced inexpensive robotics (even without major Artificial Intelligence advances), could in theory displace almost all manufacturing, extraction, and transportation jobs, and many service jobs as well. If this happened slowly, there might be time to find new sources of employment and new social systems. But if it happened quickly, large numbers of people might find themselves economically superfluous. This would tend to facilitate their oppression as well as reducing their motivation to produce and innovate. In an extreme scenario, the resulting loss of human potential might be considered catastrophic.
Another potential problem raised by distributed general-purpose manufacturing is a variety of new forms of crime. Even something as simple as a portable diamond saw capable of quickly cutting through concrete could facilitate breaking and entering. Medical devices might be used for illegal psychoactive purposes. Sensors could be used to violate privacy or gather passwords or other information. New kinds of weapons might enable new forms of terrorism. These are not new classes of problems, but they might be exacerbated if nanofactory technology were available to the general public or to organized crime. Although it seems unlikely that crime by itself could constitute a catastrophic risk, sufficient levels of crime could lead to social breakdown and/or oppressive governance, which might result in significant risk scenarios.
21.2.3.3 Destructive global governance
Any structure of governance is limited in scope by the technology available to it. As technology becomes increasingly powerful, it raises the possibility of effective governance on a global scale. There are a number of possible reasons why such a thing might be tried, a number of forms it could take, and likewise a number of pathways leading up to it and a number of potential bad effects. Molecular manufacturing itself, in addition to supplying new tools of governance, may supply several new incentives which might tend to promote attempts at global governance.
As discussed above, molecular manufacturing may enable the creation of new forms of weapons and/or newly powerful weapons on an exceptionally large scale. Along with other kinds of nanotechnology and miniaturization, it also may produce small and capable sensor networks. It should be noted that a single unrestricted nanofactory, being fully automated, would be able to build any weapon or other device in its repertoire with no skill on the part of the user (use of the weapon might or might not require special skills). Building large weapons would simply require building a bigger nanofactory first. Most governments would have a strong incentive to keep such capabilities out of the hands of their citizens; in addition, most governments would not want potentially hostile foreign populations to have access to them.
In addition to keeping excessively destructive capabilities out of private hands, both at home and abroad, many governments will be concerned with avoiding the possibility of a high-tech attack by other governments. If an arms race is seen as unstable, the alternative would seem to be the disarming of rival governments. This in turn would imply a global power structure capable of imposing policy on nations.
The analysis to this point suggests a winner-take-all situation, in which the first nation (or other group) to get the upper hand has every incentive to consolidate its power by removing molecular manufacturing from everyone else. Gubrud (1997) argues that development of molecular manufacturing will not be so uneven and that no one will have a clear advantage in the race, but at this point that is unclear. Given the potential for rapid exponential growth of manufacturing capacity, and given foresighted preparation for rapid design, it seems plausible that even a few months’ advantage could be decisive. Unless some foolproof safeguard can be devised, a policy of ruthless pre-emption may be adopted by one or more parties. If this calculation were generally accepted, then the first to develop the capability would know that they were first, because otherwise they would have been pre-empted; this temporary certainty would encourage an immediate strike. The ensuing struggle, if it did not result in a terminal catastrophic war, could quite possibly produce a global dictator.
One pathway to global governance, then, is a pre-emptive strike to consolidate a temporary advantage. Another pathway, of course, is consolidation of advantage after winning a more symmetrical (and presumably more destructive) war. But there are also pathways that do not involve war. One is evolutionary: as the world becomes more tightly interconnected, international infrastructures may expand to the point that they influence or even supersede national policy. This does not seem catastrophically risky in and of itself; indeed, in some respects it is going on today. Deliberate molecular manufacturing-related policy also may be implemented, including arms control regimes; the closest present-day analogy is perhaps the International Atomic Energy Agency (IAEA).
Global governance appears to present two major risks of opposite character. The first is that it may create stagnation or oppression causing massive loss of human potential. For example, an arms-control regime might choose not to allow space access except under such tight restrictions as to prevent the colonization of space. The second risk is that, in attempting to create a safe and predictable (static) situation, policy inadequately planned or implemented may instead lead to instability and backlash. The effects of the backlash might be exacerbated by the loss of distributed experience in dealing with hostile uses of the technology that would have been a result of restrictive policies.
We cannot know in advance whether a dictatorial regime, in order to ensure security and stability, would choose to engage in genocide, but if the desire existed, the means would be there. Even if it were simply a matter of permanent ‘humane’ subjugation, the prospect of a world hopelessly enslaved can be regarded as a terminal catastrophe. (Chapter 22 from this volume has more on the threat of totalitarianism.)
21.2.3.4 Radically enhanced intelligences
Nanotechnology, together with bioengineering, may give us the ability to radically enhance our bodies and brains. Direct human-computer connections almost certainly will be employed, and it is possible to imagine a situation in which one or more enhanced humans continue the process of augmentation to an extreme conclusion. Considering the power that merely human demagogues, tycoons, and authorities can wield, it is worth considering the potential results of enhanced post-humans who retain human competitive drives. Fully artificial intelligences – a likely consequence of nano-built computers – may pose similar issues, and Eliezer Yudkowsky (Chapter 15, this volume) argues that they may also cause existential risks due to unwisely specified goals.
21.2.3.5 Environmental degradation
Extensive use of molecular manufacturing could produce environmental impacts in any of several ways. Some types of environmental impact might represent disastrous, even existential, risks. Deliberate destruction of the environment on a globally significant scale would probably happen only as a side effect of a war that would do greater damage in other ways. It is conceivable that unwise development could do enough damage to threaten planetary ecology, but this is already a risk today.
The ease and speed with which planet-scale engineering could be carried out may tempt huge well-intended projects that backfire in ecological catastrophe. If the earth’s climate exists at a delicate balance between a runaway greenhouse effect and a sudden ice age, as some experts think (FEASTA, 2004), then large-scale releases of waste heat from overuse of billions of nanofactories and other nanomachinescould inadvertently tip the scale. Or, a seemingly benign effort such as the creation of huge areas of land or sea covered with solar collecting material might change our planet’s albedo enough to drastically alter the climate. (Climate change risk is covered in more depth in Chapter 13, this volume.)
Another possibility is ecosystem collapse as a secondary effect of massive infrastructure change. Such change could be rapid, perhaps too rapid to notice its effects until they are well underway. Computer models (Sinclair and Arcese, 1995) have shown that declining biodiversity, for example, can reach a tipping point and slide rapidly into widespread devastation, affecting whole food chains, including perhaps our own.
21.2.3.6 Ecophagy
When nanotechnology-based manufacturing was first proposed (Drexler, 1986), a concern arose that tiny manufacturing systems might run amok and ‘eat’ the biosphere, reducing it to copies of themselves. However, current research by Drexler (Burch, 2005) and others (Phoenix, 2003) makes it clear that small self-contained ‘replicating assemblers’ will not be needed for manufacturing – nanofactories will be far more efficient at building products, and a nanofactory contains almost none of the functionality of a freerange replicating robot. Development and use of molecular manufacturing appears to pose no risk of creating free-range replicators by accident at any point.
Deliberately designing a functional self-replicating free-range nanobot would be no small task. In addition to making copies of itself, the robot also would have to survive in the environment, move around (either actively or by drifting – if it were small enough), find usable raw materials, and convert what it finds into feedstock and power, which entails sophisticated chemistry. The robot also would require a relatively large computer to store and process the full blueprint of such a complex device. A nanobot or nanomachine missing any part of this functionality could not function as a free-range replicator (Phoenix and Drexler, 2004). Despite this, there is no known reason why such a thing would be theoretically impossible.
Although free-range replicators have no apparent commercial value, no significant military value, and only limited terrorist value, they might someday be produced by irresponsible hobbyists or by sects with apocalyptic creeds (compare Chapter 4, this volume), or might even be used as a tool for largescale blackmail. Cleaning up an outbreak would likely be impossible with today’s technology, and would be difficult or perhaps impossible even with molecular manufacturing technology. At the least, it probably would require severe physical disruption in the area of the outbreak (airborne and waterborne devices deserve special concern for this reason).
Possible ways to cope with outbreaks include irradiating or heating the infested area, encapsulating the area, spreading some chemical that would gum up the replicator’s material intakes (if such a weakness can be discovered), or using a large population of robots for cleanup. Although cleanup robots would require advanced designs, they might be the least disruptive alternative, especially for infestations that are widespread by the time corrective measures are deployed. An exponentially replicating free-range replicator population would not require an exponentially replicating population of robots to combat it. Since the time to nullify a robot is dominated by the time to find it in the first place, which is inversely proportional to the concentration in the environment, a constant population of cleanup robots would exert a constant pressure on the replicator population regardless of replicator concentration (see Freitas, 2000 for further discussion of these issues).
In theory, a replicator that was built of materials indigestible to biology (such as diamond), and was not combated by a technological cleanup method, might be able to sequester or destroy enough biomass to destroy the planetary ecology. A hypothetical airborne replicator might block significant amounts of sunlight, and a widespread replicator operating at full speed could produce destructive amounts of heat even before it consumed most of the biomass (Freitas, 2000). Although much more research will be needed to understand the possibilities and countermeasures, at this point the future possibility of a global catastrophe cannot be ruled out.
In the preceding discussion of nanotech-related global catastrophic risks, the greatest near-term threats appear to be war and dictatorship. As we have seen, the two are closely related, and attempts to prevent one may lead to the other. Although the focus of this chapter is on understanding risk, we will take a brief look at some potential mitigating solutions.
One of the biggest unanswered questions is the balance between offence and defence in a nano-enabled war. If it turns out that defence is relatively easy compared to offence, then multiple coexisting agents could be stable even if not all of them are reliable. Indeed, preparing to defend against unreliable or hostile states might create an infrastructure that could mitigate the impacts of private misuse of molecular manufacturing. Conversely, if it turns out that resisting offensive weaponry requires a lot more resources than deploying the weaponry, there will be a strong temptation to maintain lopsided concentrations of power, comparable to or even more extreme than today’s imbalance between states and citizens, or between nuclear superpowers and non-nuclear nations. Such power imbalances might increase the risk of oppression.
One possible solution to maintaining a stable defence in a situation where defence is difficult is to design new institutions that can maintain and administer centralized global power without becoming oppressive, and to plan and implement a safe evolution to those institutions. Another possibility would be to devise safe and predictable versions of today’s untrustworthy institutions, and transition to them. Each of these solutions has serious practical problems. Some kind of limits will have to be set in order to prevent runaway processes, including human economic activity, from doing excessive amounts of damage. Self-regulation may not be sufficient in all cases, implying at least some need for central coordination or administration.
A research organization that focused on developing defensive technologies and disseminated its work freely to everyone might help to improve stability. It is admittedly unusual in today’s military or commercial systems to have an organization that is completely transparent; however, transparency could lead to trust, which would be a key factor in the success of this approach. If defence could be made easier relative to offence, then stability may be increased. Groups that wanted to defend themselves would be able to do so without engaging in intensive military research. Some researchers who would have worked on offensive weaponry might be attracted to the defence organization instead. Ideally, most nations and other organizations should support the goals of such an organization, especially if the alternative is known to be instability leading to war, and this level of support might be sufficient to protect against efforts to subvert or destroy the organization.
For private or civilian use of molecular manufacturing, it might be a good idea to make more convenient and less powerful capabilities than nanofactories available. For example, nanofactories could be used to make micron-scale blocks containing advanced functionality that could be rapidly assembled into products. The block-assembly operation would likely be far faster and less energy-intensive than the manufacture of the blocks. Product design would also be simpler, since products would contain far fewer blocks than atoms, and the designers would not have to think about nanoscale physics. Meanwhile, especially dangerous capabilities could be omitted from the blocks.
Another unanswered question is how quickly the technological components of molecular manufacturing will be developed and adopted. Competition from other technologies and/or a widespread failure to recognize molecular manufacturing’s full potential may blunt development efforts. In the absence of targeted and energetic development, it seems likely that preparation of a supporting infrastructure for molecular manufacturing – from CAD software to feedstock supply – would be uneven, and this could limit the consequences of rapid development and deployment of new products. On the other hand, slower and more limited development would retard the ability of molecular manufacturing technology to mitigate risks, including counteracting ecophagic devices.
These suggestions are admittedly preliminary. The primary need at this point is for further studies to understand the technical capabilities that molecular manufacturing will enable, how rapidly those capabilities may be achieved, and what social and political systems can be stable with those capabilities in the world.
Vast parallel arrays of precise molecular tools using inexpensive, readily available feedstock and operating under computer control inside a nanofactory could be capable of manufacturing advanced, high-performance products of all sizes and in large quantity. It is this type of nanotechnology – molecular manufacturing – that offers the greatest potential benefits and also poses the worst dangers. None of the many other nanoscale technologies currently in development or use appear likely to present global catastrophic risks.
A nanofactory, as currently conceived, would be able to produce another nanofactory on command, leading to rapid exponential growth of the means of production. In a large project requiring scale-up to large amounts of fabrication capacity, the nanofactories could still weigh a fraction of the final product, thanks to their high throughput. This implies that the resources required to produce the product – not the nanofactories – would be the limiting factor. With nanofactory-aided rapid prototyping, the development and production of revolutionary, transformative products could appear very quickly. If solar cells and chemical processing plants are within the range of nanofactory products, then energy and feedstock could be gathered from abundant sources; with no shortage of manufacturing capital, projects of almost any scale could be undertaken.
As a general-purpose fully automated manufacturing technology, molecular manufacturing could become extremely pervasive and significant. Among the risks examined in this chapter, two stand out well above the others: global war fought with nano-weapons, and domination by nano-enabled governments. The probability of one or the other of these (or both) occurring before the end of the twenty-first century appears to be high, since they seem to follow directly from the military potential of molecular manufacturing. (The argument for instability of a nano-driven arms race in particular deserves careful scrutiny, since if it is incorrect but accepted as plausible, it may lead to unnecessary and destructive attempts at pre-emption.)
In the absence of some type of preventive or protective force, the power of molecular manufacturing products could allow a large number of actors of varying types – including individuals, groups, corporations, and nations – to obtain sufficient capability to destroy all unprotected humans. The likelihood of at least one powerful actor being insane is not small. The likelihood that devastating weapons will be built and released accidentally (possibly through overly sensitive automated systems) is also considerable. Finally, the likelihood of a conflict between two MAD-enabled powers escalating until one feels compelled to exercise a doomsday option is also non-zero. This indicates that unless adequate defences can be prepared against weapons intended to be ultimately destructive – a point that urgently needs research – the number of actors trying to possess such weapons must be minimized.
In assessing the above risks, a major variable is how long it will take to develop nanotechnology to the point of exponential molecular manufacturing (nanofactories building nanofactories). Opinions vary widely, but nanotechnologists who have studied molecular manufacturing most closely tend to have the shortest estimates. It appears technically plausible to us that molecular manufacturing might be developed prior to 2020, assuming that increasing recognition of its near-term feasibility and value leads to a substantial ($100-1000 million) programme starting around 2010.
If molecular manufacturing were to be delayed until 2040 or 2050, many of its strengths would be achieved by other technologies, and it would not have as much impact. However, this much delay does not seem plausible. Well before 2030, rapid advances in nanoscale technologies and computer modelling will likely make molecular manufacturing development relatively straightforward and inexpensive. It should be noted that once the first nanofactory is developed, the cost of further development may drop sharply while the payback may increase dramatically, thanks to the convergence of high performance, atomic precision, exponential manufacturing, and full automation. As the cost of nanofactory development falls, the number of potential investors increases very rapidly. This implies that the development of capabilities and products based on molecular manufacturing may come quite soon. Given the present lack of interest in related policy research, it seems likely that risks which might be mitigated by wise policy could in fact be faced unprepared.
Elementary level
Engines of Creation by Drexler, K.E. (1986) – While outdated in some respects, this book remains a classic introduction to the potential of nanotechnology. The manufacturing mechanism described, a collection of semi-autonomous cooperating ‘assemblers’, has been superseded by a model far more similar to conventional factories; this development obsoletes the book’s most famous warning, about runaway self-replicators or ‘grey goo’. Although Engines was written in the political context of the US-Soviet conflict, its broader warnings about military implications and the need for wise policy still have great value. Many of the author’s predictions and recommendations are relevant and forward-looking two decades after publication (entire book online at http://www.e-drexler.com/d/06/00/EOC/EOC_Table_of_Contents.html).
Nanofuture by Hall, J.S. (2005)- Nanofuture combines a layman’s introduction to molecular manufacturing with a description of some of its possible applications. The discussion of implications is scanty and rather optimistic, but the reader will get a good sense of how and why molecular manufactured products could transform – and often revolutionize – many of the technologies we use in daily life.
Our Molecular Future by Mulhall, D. (2002) – This book explores some of the implications and applications of molecular manufacturing, including its applicability to mitigating a variety of natural disasters. The proposal to deal with threatening asteroids by breaking them apart with self-replicating robots appears both unworkable and unwise, but aside from this, the book provides interesting food for thought as to how molecular manufacturing might be used.
Productive Nanosystems: From Molecules to Superproducts by Burch, J. and Drexler, K.E. (2005) – This remarkable animation shows clearly how a nanofactory might work to combine vast numbers of atoms into a single integrated product. The molecular manipulation steps shown have been simulated with advanced quantum chemistry methods (online at http://tinyurl.com/9xgs4, http://ourmolecularfuture.com).
‘Safe utilization of advanced nanotechnology’ by Phoenix, C. and Treder, M. (2003) – Early descriptions of molecular manufacturing raised fears that the manufacturing systems might somehow get loose and start converting valuable biomass into pointless copies of themselves. This paper describes a manufacturing system made of components that would all be entirely inert if separated from the factory. In addition, the paper describes some ways to make the factory system less prone to deliberate misuse or abuse (online at http://crnano.org/safe.htm).
Intermediate level
Kinematic Self-replicating Machines by Freitas, R.A., Jr and Merkle, R.C. (2004)-In its broadest sense, ‘replication’ means simply the building of a copy; the replicator need not be autonomous or even physical (it could be simulated). KSRM is worth reading for several reasons. First, it contains a wide-ranging and near-encyclopedic survey of replicators of all types. Second, it contains a taxonomy of 12 design dimensions containing 137 design properties, most of which can be varied independently to describe new classes of replicators. Third, it includes a discussion on the design and function of several molecular manufacturing-style replicators (again, in the broad sense; these are not the dangerous autonomous kind). The book puts molecular manufacturing in a broader context and is a rich source of insights (entire book online at http://molecularassembler.com/KSRM.htm).
‘Design of a primitive nanofactory’ by Phoenix, C. (2003) – This lengthy paper explores many of the design issues that will have to be solved in order to build a nanofactory. Design concepts have improved somewhat since the paper was written, but this only strengthens its conclusion that nanofactories will be able to be pressed into service rapidly in a wide variety of applications (online at http://jetpress.org/volume13/Nanofactory.htm).
‘Three systems of action: a proposed application for effective administration of molecular nanotechnology’ by Phoenix, C. and Treder, M. (2003) – Different types of issues need to be addressed in different ways. As explored in this paper, the approaches required for security issues, economic activity, and easily copied information are so dissimilar that an organization would have a very hard time encompassing them all without ethical conflicts. Molecular manufacturing will create issues of all three types, which will require a delicate and carefully planned interplay of the three approaches (online at http://crnano.org/systems.htm).
‘Thirty essential studies’ by Phoenix, C. (2003) – Molecular manufacturing needs to be studied from many different angles, including technological capability, strategies for governance, and interactions with existing world systems. These studies raise a broad spectrum of necessary questions, and provide preliminary answers – which are not especially reassuring (online at http://crnano.org/studies.htm).
Advanced level
Nanomedicine, Volume I: Basic Capabilities by Freitas, R.A., Jr (1999)-This book was written to lay a foundation for subsequent volumes applying molecular manufacturing to medicine. As a fortunate side-effect, it also lays the foundation for many other applications of the technology. Although some of the chapters are heavily medical, much of the book deals with technical capabilities such as sensing, power transmission, and molecular sorting. Scattered throughout are many useful physics formulas applied to real-world nanoscale, microscale, and macroscale problems. As such, it is useful on its own or as a companion to Nanosystems (entire book online at http://nanomedicine.com/NMI.htm).
Nanosystems: Molecular Machinery, Manufacturing and Computation by Drexler, K.E. (1992) – This book is the foundational textbook and reference book of the field. Although not accessible to non-technical readers, it makes a clear technical case, grounded in well-established physics and chemistry, that molecular manufacturing can work as claimed. Several of its key theoretical extrapolations, such as superlubricity, have since been demonstrated by experiment. Although others remain to be confirmed, no substantial error has yet been found (partially available online at www.edrexler.com/d/06/00/Nanosystems/toc.html).
Altmann, J. (2006). Military Nanotechnology: Potential Applications and Preventive Arms Control (Hampshire: Andover).
Bailey, R. (2003). SARS Wars. ReasonOnline, Apr. 30. http://www.reason.com/news/show/34803.html
Burch, J. (Lizard Fire Studios) and Drexler, E. (Nanorex) (2005). Productive Nanosystems: From Molecules to Superproducts (animated film). http://lizardfire.com/html_nano/ nano.html
Dienwiebel et al. (2004). Superlubricity of graphite. Physical Review Letters, 92, 126101.
Drexler, K.E. (1986). Engines of Creation (New York, NY: Anchor Books).
Drexler, K.E. (1992). Nanosystems: Molecular Machinery, Manufacturing, and Computation (New York: John Wiley &Sons).
FEASTA (Foundation for the Economics of Sustainability) (2004). World climate liable to sudden, rapid change. http://www.feasta.org/documents/feastareview/climate changepanel.htm
Freitas, R.A (1998). Exploratory design in medical nanotechnology: a mechanical artificial red cell. Artificial Cells, Blood Substitutes, and Immobilization Biotechnology, 26, 411–430. http://www.foresight. org/Nanomedicine/Respirocytes.html
Freitas, R.A. (1999). Nanomedicine, Volume I: Basic Capabilities (Austin, TX: Landes Bioscience).
Freitas, R.A (2000). Some limits to global ecophagy by biovorous nanoreplicators, with public policy recommendations. http://www.rfreitas.com/Nano/Ecophagy.htm
Freitas, R.A. (2005). Microbivores: artificial mechanical phagocytes using digest and discharge protocol. J. Evol. Technol., 14. http://www.jetpress.org/volume14/freitas.html
Freitas, R.A. (2007). The ideal gene delivery vector: chromallocytes, cell repair nanorobots for chromosome replacement therapy. J.Evol. Technol., 16(1), 1–97. http://jetpress.org/volume16/freitas.html
Gubrud, M. (1997). Nanotechnology and international security. http://www.foresight. org/conference/MNT05/Papers/Gubrud/
Hall, J.S. (1996). Utility fog: the stuff that dreams are made of. In Crandall, B.C. (ed.), Nanotechnology: Molecular Speculations on Global Abundance, pp. 161–184 (Cambridge, MA: MIT Press).
ISN (Institute for Soldier Nanotechnologies, Massachusetts Institute of Technology) (2005). Biomaterials and nanodevices for soldier medical technology. http://web.mit.edu/isn/research/team04/index.html
Joy, B. (2000). Why the future doesn’t need us. Wired Magazine, April. http://www.wired.com/wired/archive/8.04/joy.html
Lake, A. (1996). Defining Missions, Setting Deadlines. Defense Issues, 11(14), 1–4.
McCarthy, T. (2005). Molecular nanotechnology and the world system. www.mccarthy.cx/WorldSystem/
Orr, S.D. (2003). Globalization and democracy. A World Connected. http://aworld connected.org/article.php/527.html
Phoenix, C. (2003). Design of a primitive nanofactory. J. Evol. Technol., 13. http://www.jetpress.org/volume13/Nanofactory.htm
Phoenix, C. (2005). Molecular manufacturing: what, why and how. http://wisenano.org/w/Doing_MM
Phoenix, C. (2006). Preventing errors in molecular manufacturing. http://www.crnano.org/essays06.htm#11, Nov
Phoenix, C. and Drexler, E. (2004). Safe exponential manufacturing. Nanotechnology, 15, 869–872.
Sinclair, A.R.E. and Arcese P. (eds.) (1995). Serengeti II: Dynamics, Management and Conservation ofan Ecosystem (Chicago, IL: Chicago University Press).
Treder, M. (2005). War, interdependence, and nanotechnology. http://www.futurebrief.com/miketrederwar002.asp