Nanotechnology and International Security

Mark Avrum Gubrud
Center for Superconductivity Research

University of Maryland
College Park, MD 20742-4111

This is a draft paper for a talk at the
Fifth Foresight Conference on Molecular Nanotechnology.

This paper is available on the web at
Updated versions and additional material available at


Recent U.S. planning and policy documents foretell "how wars will be fought in the future," and warn of new or re-emergent "global peer competitors" in the 2005-2025 time frame. It is generally appreciated that this period will be characterized by rapid progress in many areas of technology. However, assembler-based nanotechnology and artificial general intelligence have implications far beyond the Pentagon's current vision of a "revolution in military affairs."

Whereas the perfection of nuclear explosives established a strategic stalemate, advanced molecular manufacturing based on self-replicating systems, or any military production system fully automated by advanced artificial intelligence, would lead to instability in a confrontation between rough equals. Rivals would feel pressured to preempt, if possible, in initiating a full-scale military buildup, and certainly not to be caught behind. As the rearmament reached high levels, close contact between forces at sea and in space would give an advantage to the first to strike.

The greatest danger coincides with the emergence of these powerful technologies: A quickening succession of "revolutions" may spark a new arms race involving a number of potential competitors. Older systems, including nuclear weapons, would become vulnerable to novel forms of attack or neutralization. Rapidly evolving, untested, secret, and even "virtual" arsenals would undermine confidence in the ability to retaliate or resist aggression. Warning and decision times would shrink. Covert infiltration of intelligence and sabotage devices would blur the distinction between confrontation and war. Overt deployment of ultramodern weapons, perhaps on a massive scale, would alarm technological laggards. Actual and perceived power balances would shift dramatically and abruptly. Accompanied by economic upheaval, general uncertainty and disputes over the future of major resources and of humanity itself, such a runaway crisis would likely erupt into large-scale rearmament and warfare well before another technological plateau was reached.

International regimes combining arms control, verification and transparency, collective security and limited military capabilities, can be proposed in order to maintain stability. However, these would require unprecedented levels of cooperation and restraint, and would be prone to collapse if nations persist in challenging each other with threats of force.

If we believe that assemblers are feasible, perhaps the most important implication is this: Ultimately, we will need an integrated international security system. For the present, failure to consider alternatives to unilateral "peace through strength" puts us on a course toward the next world war.

Introduction: the setting

Among the world's leading industrial and military powers, essentially no current official thinking envisions, within the near future, any alternative to a world system in which the ability to bring violent force to bear in international conflicts [1] remains a key component of national power, and in which sovereign states still feel compelled to deploy independent means of countering credible threats to their security and interests.

Indeed, recent United States military planning and policy documents enthusiastically attempt to envision "how wars will be fought in the future," and propose an open-ended pursuit of military technological superiority in order to maintain global hegemony (Shalikashvili 1996). While U.S. diplomatic policy and doctrine projects the image of a world almost devoid of serious potential threats (apart from a short list of "rogue states," along with criminals, terrorists, small-time warlords and social disorder), the United States military has begun to look forward to the possible emergence or re-emergence of one or more "global peer competitors" who could challenge the U.S. in its assumed role as "the sole remaining superpower" sometime in the not-too-distant future (Cohen 1997).

Overview: the spectre

The possibility that assembler-based molecular nanotechnology (Drexler 1986, 1992) and advanced artificial general intelligence may be developed within the first few decades of the 21st century presages a potential for disruption and chaos in the world system. The key areas of concern are:

A perfect balance of technical and material resources is unrealistic in any case, which leads to a new type of strategic instability:

Thus, a runaway nanotechnic arms race may be a race to nowhere; there may be no further island of stable military balance out there, even if we could manage to avoid war along the way.

Cool new weapons

The possible applications of nanotechnology to advanced weaponry are fertile ground for fantasy. It is obvious that three-dimensional assembly of nanostructures in bulk can yield much better versions of most conventional (nonnuclear) weapons; e.g., guns can be lighter, carry more ammunition, fire self-guided bullets, incorporate multispectral gunsights or even fire themselves when an enemy is detected. Science fiction writers can and do have a lot of fun imagining such things.

Certainly, nanotechnology offers many colorful possibilities for creative mass murder. For example, for some reason one of the most frequent flights of fancy is the programmable genocide germ that replicates freely and kills people who have certain DNA patterns. Such a weapon is possible, but it is dangerous to its creators, probably easy to defend against, and is of no use in attacking and overcoming a nanotechnic foe who doesn't depend on people to run his war machine anyway.

It is easy to kill. What is hard is to kill with impunity when your enemy is well armed and numerous. War is a contest to suppress your enemy's capabilities before he can suppress yours. This doesn't leave much room for fancy swordplay or gothic revenge scenarios in serious combat. An actual nanotechnic war, if one ever occurs, is likely to be inhumanly fast and enormously destructive. Clever tactics and nifty gadgets are irrelevant if your enemy can simply blow you up.

Talking about a revolution

Of course, as the US and its allies demonstrated in 1991, if you enjoy technological and numerical superiority, you can have a one-sided "turkey shoot." Pentagon propaganda during and after the Kuwait war focused on a new generation of weaponry that had been demonstrated under nearly ideal conditions. Precision-guided long-range munitions, night vision and advanced sensors, satellite reconnaissance, stealth aircraft, high-mobility, rapid-fire tanks, flexible maneuver tactics, computerization of intelligence and logistics, were said to constitute a "revolution in military affairs" that could guarantee swift victories with minimal losses. What was overlooked was that we were fighting a third-world country. Heavily armed, but still outnumbered, with technology that was a generation behind, and with few indigenous resources and no allies to call upon, Iraq was no model of a "peer competitor," and the US victory there provides no clue as to the result if the same high-tech arsenal were to confront a roughly equally-constituted force.

Pentagon visionaries paint a picture of future battlefields in which the US military seems to possess unrivaled technological omnipotence. For example, Joint Vision 2010, a sketchy 1996 pamphlet widely touted as outlining a futuristic doctrine, speaks of "...dominant maneuver, precision engagement, full dimensional protection, and focused logistics... Full Spectrum Dominance... dominant battlespace awareness... information superiority... [which] will enable us to dominate the full range of military operations... into the highest intensity conflict." The assumption of unilateral advantage seems to be a common fallacy afflicting those who fantasize about the uses of future technology in warfare (Franklin 1988). Your fantasy superweapons will work without a hitch, the enemy will behave as expected, he will not attack you at your weakest point, will not take simple measures to defend himself, and won't be able to match your technology. 1980s cartoons of Soviet missiles rising gracefully into the path of American interceptors, as though nuclear war would be a friendly game of catch, provide the most lurid example in recent memory, but anyone who assumes that America can remain the global hegemon by staying ahead in the technology race makes essentially the same mistake.

The current world situation feeds this illusion. Coming out of a half-century of global confrontation, the post-Cold War United States has inherited a military establishment whose power dwarfs that of any potential rival. While Russia has been forced to slash military spending, the United States continues to spend nearly as much on its military as the rest of the world combined. No other nation is eager to take such a burden upon itself. Resistance to global capitalism has evaporated; everywhere the powerful and the ambitious are interested first in making money. Most nations seem content to follow the US lead on intervention in places like Bosnia and Central Africa. Ironically, one of the few remaining sources of dissension with American leadership is the US insistence on drastic measures to isolate the few states it considers as "rogues," a policy that seems to be driven in part by the need to identify enemies worthy of a huge military budget (Klare 1995).

But it is chiefly the cost, as well as the lack of motivation, that keeps other nations complaining about the US running out ahead, instead of taking up the challenge to race. In the industrial technology base which underlies military technological capability, the US has at most a few-year lead over its nearest rivals, and even that is doubtful. Other countries may not wish to match American spending, but they are still very unlikely to freeze their arsenals indefinitely while the United States continues to modernize. Surely, if relatively slowly, they will replace and upgrade with current technology.

Revolution 2

With the advent of nanotechnology, the qualitative advances in weapons technology will be enormous and compelling; no country will want to maintain armies that are effectively impotent against a potential threat. Molecular manufacturing based on self-replicating systems, and superautomation by artificial intelligence, will also profoundly alter the issue of cost. A nation's military potential will depend first on its position in the technology race. A second factor will be its natural resource base, but most nations have access to sufficient natural resources to support an arsenal many times larger than any which has ever existed on Earth.

Currently the development of nanoelectronics and nanofabrication, biotechnology, supramolecular chemistry, and other steps toward molecular nanotechnology is a worldwide academic and industrial enterprise. No country can be said to have a lead in the race to develop assemblers, because there isn't any race and no one is even close. But when and if it becomes clear that something like molecular manufacturing based on self-replicating assemblers lies within reach of, say, a five-year effort, it is likely that a race will begin. Industry will be heavily involved, but national efforts will be stimulated and coordinated by government and military initiatives. The leading competitors will be those with the greatest concentration of advanced technology: the United States, Japan and Europe. However, the race will also be joined by countries such as Russia, China, India, Israel, and others that have a strong technology base, a lot of resources, or both.

A race to be the first to develop and apply assemblers could be expensive. Such a project could probably absorb the efforts of as many teams of designers, experimenters and theorists as could be mustered. Practical molecular manufacturing systems will be enormously complicated, compared with the modest accomplishments of contemporary molecular engineering. Although there is good reason to believe that our capabilities are expanding rapidly enough to be able to meet this task within a few decades, it would be foolish to think that the task is one that can be accomplished by a small team in the near term. A serious, focused effort to develop molecular nanotechnology would be funded at the multi-billion dollar level, and would expand to a major industry in the race to develop applications.

The cost of such an effort is probably a steep function of its earliness; thus it would cost, say, the United States, quite a large amount to gain an advantage of a few years over its closest rivals. If the potential significance of molecular manufacturing were well understood, a very large budget for its development would perhaps be warranted. But in the absence of a compelling threat, such as motivated the US atom-bomb project in its early stages, it is likely that skepticism will rule the day, and so nanotechnology research will have to prove its worth by providing immediate payoffs. The size of the research effort will be tied to the size of the industry it generates. That industry is likely to grow up simultaneously in several countries of the world. Another corollary of the steep cost function is that even countries that cannot afford a "Manhattan Project" may not be too far behind the leading powers in crossing the threshold of molecular manufacturing.

Thus if the United States, or another nation, decides to pull out all the stops and beat the rest of the world to advanced nanotechnology, and is successful in that effort, it will have at most a few years to decide whether to use its advantage to impose a world order that prevents the rise of a competitor, or to face the prospect of a nanotechnic confrontation.

However, it seems unlikely that any nation will ever gain such a decisive lead. Nuclear weapons will remain a powerful deterrent until nanosystems have reached a very refined state of development, and even then it is doubtful whether politicians would have the nerve to risk an attack on a large nuclear arsenal. Even the thought of initiating a non-nuclear war with a nuclear-disarmed but large and resourceful power, in the absence of any immediate provocation, would probably be anathema to decisionmakers in a democratic state. In the mean time, the potential targets will be hurrying to catch up.

It is most reasonable to expect the more or less simultaneous emergence of advanced nanotechnology in a number of industrial and potential military competitors, and its more or less simultaneous application by several states to military systems and to their manufacture. Even if one nation gains an early lead, it seems very likely that, under the assumption of peaceful armed coexistence, there would soon emerge a world in which several sovereign states or blocs possessed a molecular manufacturing capability.

A world exploded

With the emergence of molecular manufacturing, and still more with advanced artificial intelligence, the world economy will experience profound upheavals. International trade in both raw materials and finished goods will be progressively replaced by decentralized production for local consumption, using locally available materials. Wage labor, transnational capitalism and global markets will evaporate as the organizing principles of the world system. What will replace them? We don't know.

Capitalist industry has urbanized populations, concentrated land ownership, and generated enormous inequalities in wealth and power. The nanotechnic revolution would seem likely to freeze these conditions in place, leaving the urban populations stranded, without work and without a dynamic economy offering hope of upward mobility.

How will the welfare of former wage-earners be organized? What visions of the future will nations strive for? In democratic countries, assuming democracy is not undermined by inequality, the answers will be reached through mass politics. However, economic insecurity, and fears for the material and moral future of humankind may lead to the rise of demagogues. Undemocratic states, or feudal concentrations of wealth in the hands of technoindustrialists, suggest grimmer possibilities: mass deportations, genocide, or ironhanded paternalism toward "surplus" populations, and domination by cliques of unbalanced egomaniacs vying for absolute power.

With almost two hundred sovereign nations, each struggling to create a new economic and social order, perhaps the most predictable outcome is chaos: shifting alignments, displaced populations, power struggles, ethnic conflicts inflamed by demagogues, class conflicts, land disputes, etc. Chaos in less-developed and smaller countries that have traditionally relied on the value of their labor or raw materials on the world market will provide opportunities for exploitation by the leading powers. Countries that cannot develop advanced technology indigenously, and that face the loss of their role in the global economy, will be more than ever dependent on more advanced nations for access to technology, and more than ever vulnerable to sophisticated forms of control or subversion. If the decision to provide technology is a political one, the price could be political as well. Ironically, then, the successor to global capitalism could be a partial reversion to nationalistic imperialism, as major powers compete for the loyalty of clients in the former "third world."

Probably the most compelling and intractable source of tension between the leading technological powers will be competition for control of newly-exploitable resources on Earth and in space. Molecular manufacturing will, within a few years of its inception, make possible large-scale resource extraction and-or construction in the oceans, in space, and in inhospitable environments such as Antarctica. The decline of world trade will remove a powerful source of common interest in a stable world order, precisely at the same time that populations idled from productive work will be asking what they have to look forward to in the future, and precisely at the time that large-scale expansion into disputable territory becomes possible. Some nations will want equitable division of the "common heritage of mankind," others will argue that it's first-come first-served, and may stake their claims in military hardware. This is the one issue most likely to generate serious dissension and hostility even between democracies, erstwhile friends and allies, while providing an overt rationale for deploying competitive military forces.

Perhaps it will all turn out just fine; a harmonious, golden age of plenty would seem to be possible. But we don't know how to organize one. If genocidal dictatorships seem an extreme possibility, universal brotherhood, generosity and tolerance seems equally difficult to arrange. Global capitalism, in its twilight, may serve to diffuse technology, even nanotechnology, worldwide, and underdeveloped countries may find their economic independence in industries such as tourism and hand-made goods. Advanced countries may find a way to share wealth and power internally, and avoid conflicts with one another. But it hardly seems likely that the transition will be smooth.

A world in arms

If economic upheaval and the creation of a new social order poses a challenge to democratic politics even in the most advanced nations, the potential for interstate conflict still presents the greatest danger of the nanotechnic revolution. The two arenas cannot be isolated, since domestic chaos can lead to the rise of intemperate leadership, and global chaos can lead to conflicts with other nations.

Interstate conflicts, confrontations and rivalries have a life of their own. Military confrontation can be dynamically stable or unstable simply with regard to possible military moves: rearmament, mobilization, readiness, forward deployment, preemptive seizure of territory, or full-scale attack. Military threats interact with political processes in cycles that can be escalatory or deescalatory. A country that is completely at the mercy of a stronger power may seek accomodation when threatened, yet the escalation of military threat generally leads to more hostile attitudes when the two sides are more or less equally matched, even when the cost of a war would be unacceptably high to both.

The history of the Cold War provides ample evidence of both sides of this paradox. It also shows that, in spite of intense rivalry, hostility, and covert warfare, nuclear confronters will be deterred from open combat and will eventually seek detente when both are completely at the mercy of a stronger power — nuclear weapons. But finally, the many crises of the Cold War, particularly the 1962 crisis, and the long human history of disastrous wars blundered into by combinations of accident, misunderstanding, miscalculation and hubris, provides ample warning that holocaust is possible. On the assumption of stochasticity, given enough time and circumstances, global holocaust is a likely eventuality, as long as nations confront each other with arms and with threats.

Even in the total absence of political conflict or ill-will, merely that fact that sovereign states maintain separate armed forces under separate command, within reach of each other and able to attack each other, contains the germ of a possible confrontation, arms race, and war. With the advent of molecular manufacturing, nations that possess the technology will be able to greatly increase the size and quality of their arsenals in a short period of time. Unless they are controlled from doing so under some system of international agreement, it is very likely that they will begin to build up more credible armed forces, perhaps slowly and cautiously at first — but others will note the development and respond with similar increases. Soon the nanotechnic powers can be doubling and redoubling the size of the threats they pose to non-nanotechnic neighbors while imposing very low costs on themselves. Given the very large potential for expansion of arsenals by the use of a self-replicating manufacturing base, nanotechnic powers which do not engage in a very dramatic buildup will be artificially restraining themselves. It seems very unlikely that a large (orders of magnitude) gap between potential (at low-cost) and actual military production will be sustained for long.

No doubt the potential for disaster will be well foreseen, but so was the potential for nuclear disaster, and yet a combination of distrust, arrogance, and rapid technological progress made it impossible to slow the nuclear arms race before it reached the level of thousands of missiles minutes from their targets, the geopolitical equivalent of a high-noon standoff, a "balance of terror" which exacted a vast and unaccounted cost in collective neurosis, and which remains in effect to this day, in spite of the much ballyhooed Cold War "victory." The failure of the Security Council "allies" to effect radical nuclear disarmament at a time when no conflicts of interest serious enough to engender a war, hot or cold, exist, is not encouraging with respect to the prospects for avoiding a nanotechnic arms race.

A race to develop early military applications of molecular manufacturing could yield sudden breakthroughs, leading to the abrupt emergence of new and unfamiliar threats, and provoking political and military reactions which further reinforce a cycle of competition and confrontation. A very rapid pace of technological change destabilizes the political-military balance. Revolutionary new types of weaponry, fear of what a competitor may be doing in secret, tense nerves and worst-case analyses, the complexity of technical issues, the unfamiliarity of new circumstances and resistance to the demands they make, may overwhelm the cumbersome processes of diplomacy and arms control, or even of intelligence gathering and assessment, formulation of measured responses and establishment of political consensus behind them. A runaway military technological revolution must at some point escape the grasp of even wise decisionmakers.

If history teaches us anything

The pace and magnitude of technological change that we expect with the introduction of assemblers may be unprecedented in history. But the thirteen years between the explosion of the first Soviet atomic bomb (1949) and the Cuban Missile Crisis (1962) may give us some warning of what to expect. Development of nuclear weapons by the Soviet Union fed paranoia and fueled a war mentality in the United States, including widespread repression of political dissent and mobilization for an all-out arms race. Throughout the 1950s, US military posture remained hinged on preparedness to deliver a massive preemptive nuclear strike against the Soviet Union, should it appear that a major war was inevitable (or already underway). The US expected no less than that the USSR would attempt to destroy its society with nuclear weapons as well, yet the US military remained in a first-strike posture even after it became clear that the Soviet Union possessed a credible deterrent.

With the emergence of intercontinental ballistic missiles (ICBMs) at the end of the decade, the technological competition reached a fever pitch and the confrontation ascended to a nerve-wracking level of tension. Military planners now had to assume that a first strike would commence with a missile attack on major air and missile bases; with only minutes of tactical warning, it was possible that nothing would get off the ground before being destroyed. This is why the issue of forward-deployed intermediate-range ballistic missiles (IRBMs) in Cuba was considered important enough for the US to make a direct military challenge to the Soviet Union, and why the price for peaceful removal of the Soviet missiles from Cuba was the symmetrical withdrawal of American missiles from forward positions in Turkey. We now know that the 1962 crisis came within a few accidents, or a few bad decisions, of nuclear holocaust. In particular, if Kennedy had heeded the advice of his top military advisers and launched an immediate assault on Cuba, the Soviet commander's standing orders were to respond with the tactical nuclear weapons he had on hand. In all likelihood, this would have been interpreted on both sides as the signal that a general, nuclear, war was underway, and both sides were prepared to execute preemptive nuclear strikes in that situation.

There were other serious crises both before and after 1962, but the Cuba crisis was in some sense the culmination of a more chronic crisis associated with the move to a pushbutton, short-time line ICBM confrontation, and simultaneously to acceptance of the reality of mutual assured destruction (MAD), away from the even madder logic of first-strike preemption. The introduction of ballistic missiles was the most dramatic phase of the technological nuclear arms race, and it generated the most serious destabilizing effects. In the three decades that followed, the numbers of missiles and warheads expanded, they were made more accurate, more ready, more survivable, and then again more lethal against each other, but the fundamental law of strategic deterrence, MAD, became a matter of common sense. First-strike thinking enjoyed periodic revivals, and first-strike weapons were developed and deployed, but hardly anyone was foolish enough to believe that a meaningful victory in nuclear war was possible. Many were foolish enough to entertain the notion, nonetheless.

Can nanotechnology break the nuclear stalemate?

Currently, the world's major nuclear arsenals are stable or declining. Three undeclared nuclear powers — India, Israel, Pakistan — may still be building up their nuclear inventories, and others may be working to acquire a nuclear capability or develop their first nuclear weapons. North Korea may have a nuclear weapons potential or even a few completed weapons, but is not believed to be currently adding to its capabilities. Among the five declared nuclear powers, none is currently engaged in a nuclear buildup commensurate with its potential, and the two largest, the US and Russia, are committed by treaty to limits and to continuing reductions in their strategic arsenals.

If nuclear weapons remain limited in number, advanced nanotechnology could eventually undermine their potency as deterrents. No material structure can survive the fireball of a nuclear blast, and nanosystems, especially early generations of them, are likely to be especially sensitive to ionizing radiation. Advanced nanotechnology, could, however, provide relatively effective means of mass civil defense against limited nuclear threats. Shelters would probably be located underground, in deep tunnels dug by assembler-built machinery, perhaps with the aid of nanodevices that pre-cut neat fissures in the rock to minimize energy consumption. Active devices could assist in the absorption of shock-wave energy, and closed-cycle life support systems could permit isolation of the shelters from surface contamination. The result would be a great reduction in the radius of lethality of nuclear explosives, and a great improvement in the likelihood of surviving a limited nuclear exchange. Moreover, the use of self-replicating systems as a manufacturing base implies that such protection could be afforded to ordinary citizens, potentially to the entire population. The system of shelters could be sufficiently dispersed that it would present no obvious targets for concentrated attack. A transportation infrastructure adequate to evacuate urban populations in a crisis (perhaps in as little as a few hours) could also be provided.

Construction of such a system by a major nuclear power would obviously be alarming to its potential rivals, and would perhaps be anathema to civilians in the absence of a tense confrontation. However, the construction could be undertaken in the context of a crisis buildup, and should take not much longer than the number of replicator generations required to build up a sufficient quantity of primary assemblers — a matter of days, perhaps. Then again, if a nation, such as Japan, should enter the nanotechnic age without a nuclear arsenal, it would probably feel free to construct a civil defense system whenever it liked, and by so doing might provide an example that would prove irresistible for nuclear-armed states to follow.

Large-scale civil defense, coupled with active defenses against various nuclear delivery vehicles and counterforce weapons designed to disarm a nuclear opponent in a preemptive strike, might undermine the "balance of terror" and create the appearance of a possibility of victory in a war between major powers. Certainly, such a possibility would be a very strong stimulus to a developing arms race.

Indeed, it is very plausible that a nation that gained a sufficient lead in the development of molecular manufacturing and its application to military preparations would at some point be in a position to disarm any and all potential rivals, perhaps with few casualties if nuclear weapons were not used. Even if some nuclear weapons were used, there are plausible strategies for victory if one side has a sufficient technological lead. What is not plausible, from the present perspective, is that any nation ever will have such a large unilateral advantage. But the fact that rivals will fear it and will want to ensure that it is not possible is ultimately the most compelling reason to expect a nanotechnic arms race.

A nation that remained technologically stagnant, perhaps relying on a large nuclear arsenal as a guarantee of security, could suddenly become vulnerable to novel and unknown modes of attack. For example, covertly infiltrated sabotage and intelligence devices, as small as tiny insects, could be used to coordinate and execute a disarming attack, perhaps by swarming onto obsolete weapons at the appointed time, perhaps by quietly disabling them at critical points. Merely sending an expeditionary force of such mighty mites into another country's territory would create an ambiguous zone between war and peace. Spying in peacetime has traditionally been tolerated, or at least not seen as provocation to launch a nuclear strike, but in the nanotechnic era the detection of such a nanoinvasion could be the last warning a technological laggard would have before it was irreversibly disarmed. [2]

Active defense against nuclear attack has traditionally suffered from a fundamental asymmetry: the defense must be nearly perfect, since a single "leaker" can do unacceptable damage. The enormous destructive potential of nuclear weapons can also be directed against the defense; a weapon that cannot get through a wall of interceptors can be "salvage fused" to blast a hole in the wall. Nevertheless, it must be expected that a sufficient mass of varied types of interceptors can provide significant attrition of a limited nuclear missile attack. Nanotechnology will make such interceptors small and highly capable, and replicating assemblers will permit their production in almost any required number. With the addition of nanotechnic civil defense and counterforce weaponry, it is likely that an advanced nanotechnic power could achieve the capability to disarm even the present United States with assured impunity.

The problem with all these Strangelovian scenarios is, of course, that they assume a large unilateral advantage. While one side has forged ahead to develop and deploy extremely powerful new technologies, the other has remained stagnant and complacent. An increase of nuclear arsenals, deployment in more secure, covert basing modes, and development of new delivery systems designed to penetrate defenses, would, however, prolong the reign of nuclear deterrence and postpone the day of possible vulnerability to nanotechnic aggression. Thus a nuclear power or potential nuclear power, especially one that was behind in the technology race, might want to retain its nuclear options or even expand its nuclear arsenal.

About 125,000 nuclear explosives have been manufactured worldwide since 1945. Production rates peaked around 1980, and total stockpiles peaked in the mid-1980s at close to 50,000 intact devices. It should not be thought that these numbers represent the maximum capabilities of the world's nuclear powers during the Cold War. Nuclear weapons production could probably have been increased severalfold without imposing an intolerable burden on the economies of either the United States or the Soviet Union. But there was no reason to do so; even these high numbers of weapons could be said to reflect the relative cheapness of nuclear destruction, combined with a senseless obsession with guaranteeing sufficiency against the remote possibility of a first strike.

Any of the larger nuclear powers, including states such as China and India, should therefore be able to expand their nuclear arsenals to warhead counts in the tens of thousands, given a decade or so to meet a potential threat, while a state with the resources of the United States should be able to produce nuclear weapons in the hundred thousands, if necessary, using only conventional technology. Advanced nanotechnology can also be applied to nuclear weapons production. A swift buildup to warhead counts in the millions is easy to imagine.

"Doomsday machines," extremely large devices designed to produce high levels of radioactivity, could also be built. Such radiological weapons need not be deployed only as mutual suicide devices, but could be designed to blanket particular regions with highly radioactive materials. Nanosystems may be particularly susceptible to disablement by radiation damage. Advanced nanosystems could be designed to be more radiation-tolerant, but only at a substantial penalty in performance, and probably only up to levels well below what can be delivered by drastic radiological weapons.

Thus it is possible that the result of a nanotechnic arms race will be rampant nuclear proliferation and the expansion of major nuclear arsenals to warhead counts in the hundred thousands or millions — possibly including radiological weapons designed to poison environments against hardy nanotech and incidentally, life. A new "balance of terror," incomprehensibly more terrifying, and yet still perhaps unstable.

Can a nanotechnic confrontation be stable?

From a purely military perspective, in the absence of a "balance of terror" which inhibits action in spite of military logic that compels it, a confrontation between more or less equally advanced terrestrial nanotechnology powers could be unstable to preemption as a result of the special dynamics of production based on self-replicating systems, and the high levels of armament that it would be capable of producing. It might be impossible to maintain an armed peace at low levels of armament without a very strong arms control regime, including highly intrusive verification provisions; further, it might be impossible to constrain a runaway arms race from breaking out into a general war.

Strategic stability theory, as developed during the Cold War and applied particularly to discussions of the nuclear confrontation, has traditionally focused on two levels of stability in confrontation: arms race stability and crisis stability. Both relate to the presence or absence of incentives or compelling reasons for escalation in a confrontation or conflict. Briefly, an arms race is unstable if there is reason to think that surging ahead of a competitor in a quantitative buildup, or making a more aggressive pursuit of next-generation technology in a qualitative arms race, can yield a significant advantage. In contrast, if any increase in effort will only be matched by a competitor, and if the relative balance of power is not a sensitive function of relative position in the race, then there is little incentive to push beyond reasonable expectations of what the opposition may be able to do, and no compelling reason to fear the consequences of a slight miscalculation. Similarly, a crisis is unstable if preemptive moves such as mobilization, forward deployment, seizure of particular objectives, or wholesale attack is likely to give the preempter a significant advantage in any subsequent fighting. If first moves can be answered by the opposition in such a way as to neutralize any advantage gained by the aggressor, then a crisis or conflict is theoretically stable.

Molecular manufacturing based on self-replicating assemblers virtually guarantees quantitative arms race instability in a confrontation in which two or more competitors possess a roughly equal technology and resource base.

In a typical conceptual picture of molecular manufacturing, self-replicating primary systems could have a very short replication time, perhaps as little as fifteen minutes or so. Masses of such systems could be raised to resource-limited levels in a matter of as little as hours, assuming that any required special facilities were already available. The primary systems would be used to manufacture secondary production systems, again in resource-limited quantities and in a matter of minutes to hours. The secondary systems would self-assemble into finished products or would be used to carry out more specialized synthetic operations in production of the final products, again requiring only minutes to hours to complete their task. A buildup to quantities of finished product limited only by the total available quantities of primary resources (mainly first-row elements and energy) could require as little as a day to complete. Early systems would probably be slower and require special facilities, such as vats of fluid or some other controlled environment, as well as consuming more specialized feedstocks and larger energy inputs. More advanced systems would converge towards astonishingly fast speeds and high energy efficiency, would draw material feedstocks directly from the environment, and would construct any special facilities as needed. With the addition of advanced computation, say, equivalent to humanoid artificial intelligence, the process could be entirely automated.

If at least the first stage of molecular manufacturing would be an exponential process based on self-replication, a nation could not afford to lose any time if a potentially hostile rival had initiated an arms buildup by such means. Each lost generation of replicators would roughly correspond to a two-to-one quantitative disadvantage in force levels. A large imbalance in deployed hardware might allow one side to strike with impunity, and even a few-to-one imbalance might be enough to provide assurance of success in a preemptive strike. The danger of falling behind on an exponential curve, or the opportunity to trump, would create umprecedentedly strong pressures to initiate or join and to maintain or gain the lead in a quantitative arms race. The race would slow down only when each competitor had begun to exhaust the available supplies of matter and energy, and that limit might be approached in as little as a few hours. Unprecedentedly large masses of military hardware could be produced in an unprecedentedly short time.

The arms race probably would not end with resource exhaustion, assuming it did not immediately break out in general war. Each side would continue with dynamic rearrangement and improvement of its force, in response to its opponents' deployments and analysis of their technology. Each would also try to gain control of additional mass, from and in disputed zones such as space and the oceans, if a fixed division of such resources had not previously been arranged and stakes claimed in some persuasive form. If there were no resolution to the crisis, and if such a gargantuan confrontation could exist without exploding into violence, the rivals would continue building outward into space, inward below the Earth's surface, and forward to an ever more finely-tuned and lethal war machine, until...

Could such a confrontation remain "peaceful" for any length of time? It is likely that skirmishing, or violation of mutual respect for property rights and rights of action, would begin early. For example, each side would want to capture and analyze representative pieces of the others' hardware. If there were unresolved conflicts of ownership over territory and resources, each side would try to grab what it could and to block the others' access (under the assumption of nonviolence). Each would probably also probe the others capabilities to detect and counteract covert action, and each would feel compelled to attempt to gain strategic intelligence even if it meant a violation of the others' sovereign territory.

The assumption of exactly matched technologies is unrealistic to begin with. One side's replicators would reproduce more quickly, one side's computers would be smarter, one's weapons would be more effective, one would have sovereignty over a larger territory and resource base. If one party to a conflict of some type knew that its technical or natural resources were somewhat inferior, it might try to gain an early lead by initiating its buildup preemptively and in secret; once its enemies caught on, it might feel compelled to use its advantage before losing it. Alternatively, a "superpower" that had long maintained a large arsenal might see a come-from-behind competitor building up at a fast rate, and feel compelled to launch a preemptive attack.

However, even under the assumption of more or less equal technologies and equal resources, a nanotechnic arms buildup would progress toward a confrontation that would be unstable on critical fronts. The clearest case for first-strike instability can be made in the case of co-occupied environments, "no-man's land," such as the oceans, space, and any other unowned or disputed territories, and along the front lines of territorial division.

During the Cold War, the ability to hide a missile-carrying submarine in vast ocean waters was considered a profoundly stabilizing factor, while ironically, the vulnerability of surface ships to detection, monitoring and attack from close range was thought to constitute a likely route to rapid escalation in the event of open combat between the superpowers — a potential tinderbox for nuclear war. With the advent of molecular manufacturing, it will become impossible to hide anything as large as a Trident submarine in any environment from which highly miniaturized and massively proliferated sensor systems cannot be excluded, once an arms buildup has reached saturation levels.

In the advanced stages of a nanotechnic arms race, the oceans will be teeming with miniaturized and massively proliferated combatants; opposing forces will be in contact throughout the volume of the sea. They will come in a range of sizes and types with different roles to play in the pre-war jousting and in full-scale combat. Smaller and more numerous devices will have advantages of survivability and flexibility, but there will be some floor on the size of a useful device, and larger systems may have capabilities for defense against mites as well as more potential uses in general warfare. Analysis of the tradeoffs in designing a maximally-effective "fleet" will no doubt absorb the life's work of many future naval strategists, not all of them biological. But it is evident that two such interpenetrating forces will be mutually vulnerable to coordinated surprise attack. The order to attack can probably be securely encrypted [3] , and transmitted shortly in advance of "zero-microsecond." Even if the victim's force is programmed to automatically counterattack upon detection of an ambush, it can propagate signals at best at the speed of light, and in practice somewhat slower due to the delay in processing and relaying information. By attacking simultaneously everywhere, the aggressor can gain an advantage that grows as the characteristic strike time drops, which it will do as a buildup proceeds and the average distance between nearest opponents falls.

Precisely the same logic applies, even more strongly, to space fleets co-deployed in near-Earth orbit. Space weapons will typically have the shortest strike times of any weapons in any arena, and in space it is even likely that some types of directed-energy, lightspeed or near-lightspeed weapons can be employed. Again, as the numbers of opposing systems in close contact grows, the characteristic strike time will drop and the danger of a preemptive attack will become more critical.

Along territorial boundaries, forward-deployed weapons will have the shortest time to penetrate and destroy an enemy's stronghold once an overt attack is underway. As a corollary, such frontlines will become yet another arena for time-critical attack and response planning; but here the characteristic time will not drop toward zero as the buildup mounts. Nevertheless, it will become extremely short. Missile flight times can be considerably faster than those of current ballistic missiles, through the use of hypersonic powered flight, facilitated by nanotechnic materials and energy systems. Confronted by nanotechnic missile forces deployed at its borders, a nation would face the prospect of absorbing an enormous body-blow in the first minutes of an attack, unless it chose itself to attack across the border preemptively.

Even the bedrock under the land could become an arena of combat. In the absence of any deep-earth defense, an attacker could burrow like Guy Fawkes and attempt to blow up an enemy from below. To prevent this, nations would probably claim the earth below their land masses as sovereign territory and erect subterranean lines of defense. Any attempt to cross such a line could probably be detected and intercepted. Note also that it is unlikely that any device could burrow directly through molten rock.

The destructive potential of large numbers of nuclear weapons, such as could certainly be produced with the aid of nanotechnic systems, ought to serve as a sufficient deterrent to any adventure that would be likely to bring on their use. And yet it is unclear whether the vulnerability to preemption of forward-deployed and interpenetrating forces would permit military decisionmakers, human or otherwise, to remain cool in facing a mounting confrontation of planetary proportions. The fragility of any such "peace" should in any case be obvious. Again, a true balance of power is unlikely to exist at any time, and the possibility that advantage may be gained and lost is a powerful incentive to preemptive action. A commander might reason that warfare could be limited to contested zones, avoiding direct attack on home strongholds and core values, e.g. populations, and without triggering any "Doomsday machines," by mutual deterrence. More likely, in such a massive and intense confrontation, a war would find a way to start itself.

Large-scale development of space, beyond the near-Earth environment, would complicate the question of military instability by possibly providing refuge for survivable forces. It is unclear, however, whether this would fundamentally alter the picture drawn above. In the near-term, at least, and even after the development of molecular manufacturing, it is likely that space development will occur under national auspices and under claims of national sovereignty extended by otherwise earthbound nations. From this perspective, space settlements and space-based weapons deployments would seem only to extend an Earth-centered confrontation, and extend it into an arena of potentially lightspeed weapons and maximal military instability. In the longer term, however, far-flung settlements would be outside the reach of any reasonable first-strike order, so that the danger of such an attack would be somewhat defused.

Long before that time, the issue of territory in space is likely to be a powerful source of competition and conflict between the leading technological nations of the Earth. Would-be sooners who hope to escape into outer space before terrestrial stick-in-the-muds blow themselves up are engaging in unrealistic fantasy. Any attempt to seize control of an extraterrestrial empire can only provoke a war that will not fail to pursue the pioneer settlers; indeed, it may target them first.

Artificial intelligence as a factor

Most of the points made above in connection with assembler-based molecular manufacturing can be drawn as implications of advanced artificial intelligence (AI) independently of, or in conjunction with assemblers. Information technology alone may lead to essentially the same dilemmas, even if molecular assemblers turn out to be unworkable or very hard to develop. Assemblers combined with artificial general intelligence will lead to these dilemmas with breathtaking swiftness.

By advanced artificial general intelligence, I mean AI systems that rival or surpass the human brain in complexity and speed, that can acquire, manipulate and reason with general knowledge, and that are usable in essentially any phase of industrial or military operations where a human intelligence would otherwise be needed. Such systems may be modeled on the human brain, but they do not necessarily have to be, and they do not have to be "conscious" or possess any other competence that is not strictly relevant to their application. What matters is that such systems can be used to replace human brains in tasks ranging from organizing and running a mine or a factory to piloting an airplane, analyzing intelligence data or planning a battle.

Consider: A computer that can conceptualize a problem, apply general knowledge and reason through to a solution, can do so without distraction, taking breaks or getting tired, and with immediate, direct access to number-crunching processors, hard-fact databases, and even robotized prototyping systems and test benches. Merely to reproduce human intelligence is thus probably to surpass human capability. Such a machine and its software can also be copied without a 20-year educational process, in numbers limited only by manufacturing capabilities. Armies of them can be networked and harnessed to large, complex problems. Consider also, that the active elements of electronic computers already propagate signals a million times faster than our neurons, and can potentially be more densely packed and numerous as well.[4] These images only serve to suggest what may be possible. Realistically, it is likely that AI systems will be only roughly anthropomorphic and will evolve through varied and unexpected forms, bypassing human equivalence.

It is people, and not material resources or technology, that determine the scale of industrial production and military operations today, and that limited the pace of weapons development during the Cold War. People may be highly motivated, convinced that the enemy is evil and will come breathing fire if they don't work around the clock to develop and manufacture terrible weapons, yet they will still get tired and slack off, and eventually go home to rest, and they will make mistakes, and they will resist being herded into senseless wars or burdened with futile arms races.

Information technology multiplies the capabilities of human engineers, planners, commanders and fighters, and oils the wheels of industry and of war. Luddite predictions of technological unemployment and dystopian tales of robot soldiers have been around a long time, but the key ingredient that has always been missing is a general replacement for the human brain. Yet already computers replace brains in many lines of work, and simple extrapolations of the growth of computer power show that computers should come to equal, and then to surpass, human brains, on a per-cost basis, within a few decades (Moravec 1988).

At some point, even without nanotechnology, the replacement of brains by computers will lead to a military production and war machine that is fully automated and can be ordered into action without delay or resistance. When the war machine can run itself, there will still be limits to what it can do, but those limits are likely to be well beyond those of historical experience. Note that a fully-automated production system is by definition potentially self-replicating, although the replication cycle for bulk-process manufacturing systems is likely to be much slower than that for molecular nanotechnology.

Perhaps the worst-case scenario for a runaway arms race is the simultaneous emergence of molecular manufacturing and artificial intelligence, or the emergence of nanotechnology at a point where the architecture and software of advanced artificial intelligence have already been developed. Nanosystems will be the most complicated electromechanical systems ever devised, and the participation of human-equivalent AI in early nanosystem development could speed the pace of innovation and application enormously. Conversely, the use of assemblers to manufacture machines containing enormous numbers of active devices could yield a sudden increase in intellectual capabilities to levels well beyond human equivalence. If assemblers come early, the world may have a better chance to digest the prospect of a self-replicating capital base and understand its implications. If they come late, the progress from breakthrough to full-scale deployment will more likely be too swift to control. [5]

The likelihood that AI systems, once they reach human-equivalent levels of general knowledge and reasoning ability, can participate in their own improvement as engineers and as technical implementers, outperforming human capabilities and producing an explosion of intelligence that transcends our understanding, has led to the concept of a "technological singularity" (Vinge 1993), an acceleration of technological progress to an essentially infinite pace [6] , a horizon, beyond which it is impossible to see, beyond which whatever emerges will exceed our ability to imagine or to comprehend. It is argued that machines will inevitably escape human control, or that humans must transcend along with the machines into some post-human form if we wish to escape domination by our intellectual superiors and successors. [7]

If a technological singularity lies ahead, one can argue that we are already in its basin of attraction, and perhaps we have been since the beginnings of technology. But certainly, if we want to try to affect the outcome, to have a chance of surviving a plunge into a technological black hole, the best we can do is try to set initial conditions as we sail in. As frightening as it is to contemplate the transcendence of technology over human intelligence and will, an even more frightening prospect is the transcendence of a technology committed to warfare and divided against itself. If a transcendent technology wars against itself, we fear we will perish in the crossfire.

Extending a metaphor, we can imagine that as we approach a singularity, rising tidal forces tear at the economic, political and military structure of our world, so that if we approach in a jerry-rigged vessel it will tend to tear apart, even if, strictly speaking, we never reach a singular point. In gravitational physics, tidal forces arise because the freefall trajectories of different points on an extended body diverge; so the part that leads pulls ahead, the part to the right is thrown to the right, and so on. Similarly, an accelerating rate of technical progress can amplify inhomogeneities in the level of development of different nations and even different groups and classes within nations, upsetting haphazardly constructed military, economic and political balances, and amplifying conflicts of interest and ideology as well.

The single greatest weakness in the modern world order is, and has been, the division into sovereign states which confront each other with arms, with threats and with acts of violence. This is the division of technology against itself, intelligence against intelligence, purpose against purpose. How can a world crazed by such fault lines of reason survive an encounter with a technological singularity?

Arms control and international security

It is highly likely the world will continue to be divided into sovereign states with competing militaries as the nanotechnic revolution approaches, and that, certainly in the United States at least, much of the research leading to it will be sponsored through the military with an eye to military applications. Therefore, if we are to have any hope of avoiding a catastrophic arms race it is essential to consider possibilities for control of nanotechnic arms.

It is easy to dismiss the idea of an outright ban on the use of assemblers in weapons manufacture, the use of nanostructures in weapons, or any similar proposal, as impractical and unverifiable. What is often lost sight of, however, is that arms control always involves more than treaties and "national means of verification." Above all, it requires the will to control dangerous and undesirable weapons, and a willingness to cooperate in order to achieve such control. If such will is present among the parties who need to be involved, there is much that can be done.

The most threatening aspect of a possible nanotechnic arms buildup is the sheer mass of weaponry that may be produced. Not producing such masses of arms, nor preparing the facilities that may be required for their production, is not an unverifiable commitment. Voluntary transparency, by which nations allow their treaty partners to maintain prescribed monitoring capabilities on national territory, can permit asssurance that no such large-scale buildup is taking place. Such arrangements will only remain effective, however, as long as nations refrain from attempting to settle disputes by violence. Even the mere threat of violence, ordinary "peaceful" confrontation, is likely to provoke withdrawal from a weak control regime in the event of a serious crisis.

It should be much easier to put a control regime in place before any large-scale confrontation develops, while nations are at peace, than to stop an arms race in forward gear and ask for intrusive monitoring so that it can be reversed. Unfortunately, arms control, unlike arms acquisition, has no natural constituency other than public anxiety about the danger of war, and this is at its lowest precisely when the greatest progress could and should be made towards eliminating weapons left over from the last arms race, and erecting a structure that can hold off the next one.

Probably the most important arms control step that could be taken now, other than continuing the process of nuclear disarmament, is to conclude a general and global treaty banning the placement or testing of destructive devices in orbit, and the targeting of objects in space by ground-based weapons. The possibility of a rapid buildup of space weapons is the most dangerous and destabilizing prospect of 21st century military confrontation. The recent US test of a ground-based antisatellite laser is appalling; let us hope it serves to demonstrate the real continuing danger of a future space arms race.

The vexing questions of outer space and sea law, and the division of resources that may become valuable in the future, must be addressed before these issues become a source of international conflict, that is, before the technology is developed which makes what has heretofore been viewed as a "commons" an attractive target for sovereign ownership.

In spite of the arguments made above that nuclear weapons might serve as a stabilizing factor in a nanotechnic confrontation, continuing and completing the job of nuclear disarmament is an urgent priority. It is inexcusable to leave our people and our civilization exposed to the danger of nuclear annihilation at a time when there are few sources of tension between the major nuclear powers and none that are considered remotely serious enough to occasion a crisis that could lead to war.

Nations must learn to trust one another enough to live without massive arsenals, by surrendering some of the prerogatives of sovereignty so as to permit intrusive verification of arms control agreements, and by engaging in cooperative military arrangements. Ultimately, the only way to avoid nanotechnic confrontation and the next world war is by evolving an integrated international security system, in effect a single global regime. World government that could become a global tyranny may be undesirable, but nations can evolve a system of international laws and norms by mutual agreement, while retaining the right to determine their own local laws and customs within their territorial jurisdictions.


The nanotechnic era will be fundamentally different from the era in which nuclear weapons were developed and came to dominate the possibilities for global violence.

The bombed-out cities of the Second World War, and the nuclear holocausts of our imagination, have persuaded rational minds that there can be no expectation of a meaningful victory in total war between states armed with hundreds of deliverable nuclear weapons. From that point of view, war is obsolete, at least direct and open war between great powers.

Nanotechnology will carry this evolution to the next step: deterrence will become obsolete, as it will not be possible to maintain a stable armed peace between nanotechnically-armed rivals. The implications of this statement stand in sharp contradiction to the traditions of a warrior culture and to the assumptions that currently guide policy in the United States and in its potential rivals.


The history of war in the modern technological era is a history of surprise. Time and again, technology proves its power against our vulnerable bodies. A vast destructive potential is kept sealed until the time of battle, and when the seals are broken, we are surprised by how much vaster the devastation is than even the last terrible war. Order an attack over the trench lines... Surprise! The guns and artillery turn your brave soldiers into hamburger as fast as you can feed them into the grinder. Unleash a war of conquest... Surprise! Fifty million dead, your great cities in ruin, the survivors cold and starving. Start a nuclear exchange... Well, we were warned.

It was technology, not policy, that forced the doctrine of deterrence on us, just as it was technology that determined the outlines of the nuclear arms race, once the decision to pursue nuclear confrontation had been made. The logic of military technology produced a confrontation so complex and unmanageable, and with such short time lines for decision and action, that it threatened to explode in spite of "assured destruction." Again, people were intelligent enough to recognize realities, and to place restraints on the offensive arms race while shelving futile dreams of defense.

If technological realities now demand that we go further, and give up the warrior tradition, the illusion of independence and the vanity of sovereign self-defense, will we heed these demands, or will we try to preserve the institutions and attitudes of an earlier epoch, until we are surprised by a disaster beyond even our worst nuclear nightmares? If it is impossible to maintain an armed confrontation between nanotechnology-armed and hostile nations, then this is exactly our dilemma.



  1. The following terms are often used loosely or as euphemisms for war, but will be given strict definitions here: Confrontation means the maintenance of armed forces, assertion of control over specific assets and assertion of specific rights, up to mutually-recognized limits, by two or more centers of power; Conflict means the assertion by two or more parties of mutually incompatible positions, seriously enough to imply the possibility of violence should they fail to resolve the dispute by peaceful means; Warfare means actual violence; Combat implies that the violence is an intense and mutual effort to suppress each other's ability or inclination to fight; War is the condition that exists when warfare is general and uncontrolled, although war need not escalate to the exhaustion of all possibilities for violence.
  2. Other examples of novel attack modes that could be used by a nanotechnic power to overcome even a well-armed nuclear but non-nanotechnic power include earth-burrowing weapons, space weapons, or again the programmable microbe, but one programmed to kill at a specified time. If you have a sufficiently bloodthirsty imagination, you can keep yourself amused in this vein from now until....
  3. This seems likely, but it is a critical point and there are two reasons to doubt that encryption of a general attack order can be kept secure. The less persuasive one is current speculation about the possible code-breaking potential of quantum computers, for which no plausible physical implementation has yet been proposed. The more persuasive argument is that the smallest combatants may not be able to support secure decryption, while the potential victim need only somehow "capture" a small number of "canary" combatants of the potential attacker in order to obtain a functional warning system.
  4. It may be that evolution has had little incentive to drastically increase the speed of signal propagation, given the size of our bodies and the time scale of events that affect our survival. A move to computerlike serial algorithms is biologically implausible. Note that while evolution has a clear incentive to make neurons smaller and more efficient, it might be up against hard limits for cellular processes, while a slight increase in speed might come only at the cost of an increase in energy consumption. One should not underestimate the neuron. It implements a highly nonlinear space-time integration of thousands of inputs, and its growth and tuning is probably governed by complicated biomolecular interactions. If one wants to build an artificial brain, the artificial neuron is likely to be a fairly complex cellular automaton, not a simple linear integrator and threshold trigger. It will be a challenge to build an artificial brain that is as complex and compact as the original. However, the artificial brain should easily be faster.
  5. Taking a best guess, it is likely that at least a first generation of assemblers will be developed before any computer is built that equals the human brain in quantitative measures of complexity and speed, let alone before artificial thinkers can replace and surpass the skills of human engineers, planners, and warriors. But there seems no reason to assume that human-equivalent AI could not be achieved in machines built from components made by top-down nanofabrication methods, without requiring bottom-up nanoassemblers.
  6. "Infinite rate of progress" means relative to a human scale. Imagine, for example, that a transformative amount of progress were to take place in a subjective instant.
  7. Note that this implies willing our own destruction and replacement by something nonhuman.