More Lessons from Afghanistan
Throughout the fall of 2002, a major fight between the Army and the Air Force raged over the conduct of Operation Anaconda, an anti-Taliban sweep conducted in spring 2002. The issues raised are relevant to the new kind of warfare being developed by the United States. Briefly described, the new style of war emphasizes intelligence or surveillance as a way of minimizing the need for massive numbers of weapons. In theory, a relatively small force equipped with excellent surveillance systems and the relevant communications—and with access to precision weapons—ought to be able to defeat a much larger but more primitive force. This capability is extremely important, because the United States may find itself forced to deploy relatively small forces to meet large numbers of nearly simultaneous crises.
Anaconda was a test of the current assumption that surveillance can defeat most types of simple cover. The operation was intended to deal with a few hundred Taliban in a mountain valley. By the time it was over, the estimated size of the enemy force was a thousand or more. An operation intended to last two days continued for more than a week. Many readers probably received an e-mail, apparently inspired by the U.S. experience, that described the Taliban and al Qaeda troops as surprisingly professional and adaptable.
The interservice fight is at least nominally about close air support. The Air Force tends to prefer planned strikes dealing with enemy forces not in contact with friendly ones. In response to what it saw as a neglect of its tactical requirements, the Army built up a force of attack helicopters under its own control. On the other hand, given precision weapons, Air Force aircraft can deliver weapons in close support—if the force requiring that support includes forward air controllers.
It seems that the Army unit involved in Anaconda, the 10th Mountain Division, made no attempt to avail itself of Air Force support while it planned the operation. Given that it expected light Taliban or al Qaeda opposition, the Army apparently considered its own troops plus Afghan supporters and four to six Apache gunships sufficient. Reportedly, its own plan was a seven-pronged assault in the valley. There was talk that the Army had deliberately avoided calling on Air Force assets because its own units had had no part in the war thus far, and it wanted to demonstrate its combat capability. Later, Army spokesmen would claim that the Air Force would have offered a pre-assault bombardment—which the Army rejected for fear of losing the element of surprise. This very natural claim carries an unfortunate echo of the planning for Omaha Beach at Normandy, where pre-landing bombardment was vetoed on much the same ground.
As for the element of surprise, something went very wrong. About a day before the scheduled attack, the Taliban and al Qaeda force went into the mountains above the valley—giving the enemy excellent positions from which to fire down on the advancing troops. That should not have been too surprising, since preparations for the assault would have been quite visible to villagers nearby, many of whom were thought to be Taliban supporters. The surprise may have been that the Taliban and al Qaeda troops were willing to fight at very high altitudes, in some cases about 14,000 feet.
It is not, moreover, clear that a pre-attack bombardment would have been effective, given that U.S. intelligence did not detect the massive enemy movement into the mountains. Later, it was stated that sensors had missed about half the enemy positions—again, no surprise, because the positions were mainly extemporized, and as such could not have been recognized. What is important is the way that Murphy's Law affected the assumptions inherent in the new kind of warfare. Many targets were not revealed until the enemy fired from them.
As it turned out, U.S. errors and limitations were anything but fatal. Despite accounts suggesting that Anaconda was disastrous, casualties were light. Eight Americans were lost, seven of them when a helicopter landed in a zone that had not been sanitized by fire. The Taliban lost opportunities to do much more serious damage. Overall, a combination of good troop training and very heavy firepower support—by aircraft of the Air Force, the Marine Corps, and the Navy—canceled whatever surprise the Taliban had planned. As for the Army, its Apache gunships did not operate after the first day, having been hit by enemy antitank weapons—a tactical surprise.
So what does all of this tell us about the new kind of warfare? First, numbers or weight of fire still are vital, because Murphy's Law can offset our apparent advantage in surveillance. The small numbers of precision weapons envisaged by the new warfare can work if they are supported by excellent surveillance, but that may not be available. True, most munitions used in Anaconda were delivered to precise coordinates, but those coordinates often were discovered only when enemy troops began firing from them. Preoperation surveillance is not enough.
Second, a ground force going it alone invites disaster. What saved Anaconda was excellent improvisation at all levels, much of it in the form of close air support. If the Army saw Anaconda as a way of demonstrating its own capability—a move in the endless Battle of Washington—it failed badly, and in doing so it underestimated its enemy.
Links Grow Between Commercial and Military Technology
In December 2002, the Air Force announced a major program to replace the radar used by the B-2 bomber—not because the radar had been compromised, or because it was too expensive to maintain, but because the frequency range it uses no longer is available, having been released for commercial purposes. The B-2 radar thus becomes one of a long list of military electronic systems put out of business by the expanding civilian market. As 2002 ended, the U.S. government was trying to restrain developers of new broadband wireless systems that seem poised to dominate the 12-14 gigahertz frequency range, at the high end of what the military calls X-band radar frequencies.
In the case of the B-2, it seems that the original radar was hard wired to use a fairly narrow frequency range, which has now become unavailable (presumably because it is wanted for next-generation cell phones). The new radar will use broadband electronics whose transmitted frequency can be set by software to miss any commercial range (which may vary from country to country). It probably also will use separate receiving/transmitting modules, which can be controlled separately to form beams.
The growth of commercial electronics seems to be relentless. For military planners and developers, the most troubling aspect is that anything designed right now to incorporate the latest possible computers likely will be obsolete as soon as it is built, at least from a computer point of view. According to Moore's Law, computer processor chips double in power every 18 months. That is why it is so embarrassing to read computer magazines a year or two after buying a top-of-the-line machine; by next season, equivalent machines often are twice as fast, or cost half as much, or, worse, both.
When will Moore's Law stop? Virtually everything in life, including technology, seems to follow S-shaped curves. At the beginning, progress is very slow. Then it picks up, and the curve looks geometric—like a curve doubling every 18 months. Nothing, however, rises forever. At the top of the S, the curve slows dramatically, turning into another period of very slow progress. Unfortunately, the mathematics of the curve is such that no one on the steeply rising part of the S can judge where the upper knee of the curve is. Nor can we guess that from the technology itself. We do know that as chips become faster—and smaller—they require new fabrication plants to make them. Companies making chips thus periodically have to invest increasing amounts of money simply to keep making faster chips. At some point, that investment may itself become impossible.
For military developers, there is a crucial difference between Moore's Law terminating in 6 years (4 doublings) or in 30 years (20 doublings). For example, all forms of stealth are, in effect, attacks on signal processing. Everything produces some signal, but below a certain level it becomes effectively impossible to separate from noise, or at least to separate quickly enough from noise for tactical purposes. That level is set by signal processing power, which in the end means computing power. A form of stealth that could handle 4 doublings of signal-processing power might not survive 10 or 20.
Perhaps this year's Christmas season offered a clue. Personal computer sales seemed to have peaked; new machines sold at discounts so steep that the manufacturers may have made almost no profits. If personal computer sales provide the capital for new chip plants, then the current disaster may well discourage that investment—and end the headlong growth in computer-chip capacity. On the other hand, if commercial applications such as workstations are the key component in paying for new chip fabrication plants, then it seems unlikely that the end of Moore's Law is close at hand.