Speculations on Armed Conflict

In a Time of Free Silicon




The inevitability and relevance of the Small and the Many may be challenged by several factors: weapons of mass destruction, wide-area electronic counter-measures, the repeated difficulties of making artificial intelligence work, and the simple persistence of legacy warfare systems.

Just in case the future wanders away from these predictions, readers can get a head start on Monday morning by counting, in advance, all the ways it can get lost.

Obstacles to the triumph of the small and the many are numerous. Developments in wholesale war -- nuclear, biological, or chemical agents -- may obviate any technologies that alter the calculus of retail war. Even in retail war, new technologies usable only in large complex systems might nullify or destroy the small and the many en masse. Much of what lets small chips replace big manned platforms assumes advances in artificial intelligence, whose progress is notoriously resistant to forecast. Old technologies and institutions have ways of fighting back against new technologies that promote confusion; thus, crossover points frequently recede.

The Irrelevance of Retail War: To resolve important issues, would nations worry about the art of grabbing or defending territory or might they instead reach for weapons of mass destruction first?

Weapons of mass destruction come in three types: chemical, biological, and nuclear. Chemical weapons are not likely to affect the dominance of the small and many. Because few chemicals can affect the Mesh, their tactical application is limited. Today's chemicals are just one more obstacle to manned warfare on the battlefield. The low chances of a breakthrough chemical weapon mean the current calculus of uncertain effectiveness and certain retaliation will persist. Hitting U.S. civilian targets requires the use of strategic delivery vehicles. The few who may get them would not waste them on chemical weapons if nuclear ones are available. Successful use invites nuclear retaliation--yielding little scope for chemical weaponry.

Biological weaponry is even harder than chemicals to use tactically. The ability of germs to multiply permits havoc disproportionate to their payload; yet their battlefield use has been rare. Germs are hard to control and may backfire as gas did when first used in World War I. Their effect on humans is extremely difficult to test. The more open the world, the harder it will be to hide errant tests. Greater sophistication could lead to greater disaster. Anthrax, for instance, is very potent, but infected areas are uninhabitable to friend or foe for eons. Germs that kill their host quickly will not spread quickly. Germs that kill slowly cannot be timed for tactical military advantage. As with chemical weapons, effective germ use may trigger nuclear retaliation.

The strategic use of biological warfare may be more effective but runs the risk that the induced disease crosses national boundaries and comes home. However, a trump card of biological warfare could be the virus that is made genetically specific to target hosts. For instance, a virus that attacks only yams that grow in an enemy country might reduce the latter to starvation. More sinister would be a virus that attacks people of a limited human genotype. It is not clear, however, where any of the zillion combinations of nucleic acids would yield such a virus (much less whether one might be discovered soon).

Would nuclear proliferation make retail war obsolete? During the Cold War, both sides took conventional and nuclear operations seriously. They conducted the former but never the latter thanks to the nuclear stalemate.

The advantage of nuclear weaponry against (as opposed to instead of) the small and the many may be that it can destroy or disable the millions without having to look for them. A field swept clean of such objects would let platforms march through, even if just temporarily.

Yet, tactical nuclear weapons (even discounting their potential for escalation to strategic ones) may be no more effective against Meshes than General Grant's use of explosives at the battle of Cold Harbor. Fields cleared by one side may be promptly reseeded by the other with more sensors; the respite in between may be too temporary to yield much advantage. Delivery vehicles for nuclear arms are subject to the same real- time tracking and targeting that conventional platforms are subject to.

Nuclear weapons might also be used to generate an electro- magnetic pulse (EMP) big enough to clear electronics from an area so large that it cannot be reseeded quickly. However, Mesh electronics should be less vulnerable to EMP than are large systems connected to long wires. The only wire associated with such items would be receiving antennae which could be fitted with protective diodes to keep a large induced wave from frying the chips, themselves. Hardening electronics is another possibility. However, if Mesh components diverge to far from their commercial counterparts, they would become too expensive to buy in the right quantities.

Powers in the Big: Might there arise militarily decisive technologies available only in very large sizes that could erode the logic of the Mesh?

EMP effects smaller than those generated by nuclear weapons can be provided by microwave weapons, for instance. Such weapons require considerable reserves of energy to be effective, but could fry weapons systems electronics at a considerable distance. By staying under the nuclear threshold, microwave weapons may be more usable. Yet if they are less powerful, they would have less effect. They could also be tracked in real time before being used.

Incoming missiles with electronic targeting may be useless if countered by microwaves in certain situations; without missiles all the data generated by the Mesh would avail naught. Yet must missiles be that vulnerable? Those inertially guided and mechanically fuzed, do not need electronics. Moreover, the cost of the microwave machine may be greater than the cost of saturating its defenses with enough not-very-smart rockets to destroy it.

Finally, many sensors may be simply unavailable in small form. In particular, those which require being bathed in very cold liquids to work well may only function if coupled with large expensive cryogenic devices. The latter include SQUIDS (superconducting quantum interference devices) and certain types of infrared detectors (otherwise confused by ambient heat).

Shortfalls in Artificial Intelligence: The hoary "if it works, it isn't artificial intelligence" retains a certain bitter truth after several decades. Both advocates and skeptics of artificial intelligence share a long history of bad predictions. Advocates have consistently underestimated how much horsepower is required for useful work. Predictions of easy automatic language translation were made in the 1960s but only now can such programs be purchased. Conversely, skeptics predicted that certain feats -- a computer beating a grand-master at chess -- were inherently impossible, but within the last two years, a computer has defeated a grand-master.

Each of the three relevant areas of artificial intelligence -- pattern recognition, machine learning, and synthetic logic -- has seen startling successes and dismal failures. For pattern recognition, in particular, the trend is away from linear logical approaches and toward imitating human neural techniques. The excitement that greeted the widespread introduction of neural net techniques in the mid-1980s has abated -- functional nodal architectures are more complex than first realized. On the other hand, companies are busy casting neural net chips, so there must be something there.

Broad analogues of the human brain -- notably the faculty of common sense -- are still eons away. The more limited a domain (e.g., if all dialogue concerns biochemistry), the faster the chances for success. Ironically, the persistence of domain limitation argues against robotic images of technology and toward complex networks of simple sensors. But it still leaves man as integral to command and control in warfighting.

The Persistence of Legacy Systems: The last barrier to the Mesh is that radical futures seem to take longer getting here than simple technological extrapolation would suggest. Picking broad trends is easy; solving the thousands of problems that must be faced before the broad trends are realized is not.

The new always faces the resistance of the old, aided by patterns of familiarity, sunk costs, well-tested habits, and a large supportive infrastructure -- hence the observation that a the new must improve over the old by a factor of ten if it is to overtake it. In the meantime, the old rarely stands still. Chips are still made with silicon even the same chips recast in gallium arsenide would run three to five times faster. Silicon technology has been pushed past hitherto disabling hurdles, even as the promise of gallium arsenide confronts problems not clearly understood at the outset.

Yet: Two major considerations still favor the Mesh. First, the commercial technologies continue to advance; as they do, the gap between existing military systems and new systems based on commercial components shrinks. Advanced economies that have yet to develop a large military-industrial complex (e.g., Japan or the collectivity of overseas Chinese) would find that this gap could be bridged quickly. The route to a superior military, which otherwise would retrace the path taken by other nations, could be shortened by flying through a technological worm-hole.

Second, military technology continues to be intensely competitive, thus success in one place would promote its spread elsewhere. True, an agreement among superpowers can suppress known lines of development. Arms control and non- proliferation treaties have worked. Used to suppress a speculative line of development in an era of great strategic uncertainty, however, their success is less certain.