LEVEL 1 - 29 OF 104 STORIES
Copyright 1993 Information Access Company, a Thomson Corporation Company
ASAP
Copyright 1993 Massachusetts Institute of Technology Alumni Associati
Technology Review

April, 1993

SECTION: Vol. 96 ; No. 3 ; Pg. 75; ISSN: 0040-1692

LENGTH: 1413 words

HEADLINE: War in the Age of Intelligent Machines._book reviews

BYLINE: Warner, Will

BODY:
	SOON after helping to wrest Palestine from Britain, Menachem Begin wrote: "When a 
nation reawakens, its finest sons are prepared to give their lives for liberation. When empires are 
threatened with collapse, they are prepared to sacrifice their noncommissioned officers." It's an 
ancient pattern. But current military research suggests that technically advanced empires of the 
future may have another option: spare even the noncoms and throw robot weapons into the breach.

	Will this be possible? I hope so. By now everybody should know what General Sherman 
knew even before the world wars set new standards for carnage: that "war is cruelty, and cannot be 
alloyed." Understandably, then, people will get machines to do their fighting for them if they can, 
and in fact have always sought technological Excaliburs with which to defeat their enemies at 
reduced risk to themselves. If robot weapons will afford some protection from the slaughter of 
modern warfare, and can be produced, they will and should be deployed.

	Manuel De Landa, a film critic and computer programmer, disagrees. In his book War in 
the Age of Intelligent Machines, De Landa depicts a military future that includes  autonomous 
weapons built around computers. The Pentagon's plans for AI research convince him that the 
military has "the will to endow machines with predatory capabilities," and certain R&D projects 
convince him that the machines are on the way, much to his distress.

	De Landa calls the new weapons "intelligent" to distinguish them from merely "smart." 
Smart weapons made their debut in Vietnam as bombs that could follow a laser beam to a target. 
Two years ago we saw films of smart bombs devastating targets in Kuwait and Iraq. A cruise 
missile is more sophisticated, though still only "smart." It uses its own radar to watch the terrain 
over which it flies, comparing what is below with radar-generated maps in its memory and 
adjusting its course accordingly. Cruise missiles and laser-guided bombs are merely smart because 
they do not choose their own targets and they remain under human control.

	The weapons De Landa writes about will possess much more sophisticated artificial 
perception and intelligence. These robots will attempt on their own to assess threats, identify 
targets, and decide whether, when, and how to attack. They will come with instincts for self-
preservation and, more important,with the means to blow up structures, vehicles, and people. 
"These weapons," De Landa writes, "will be the first killing machines that are actually predatory,
that are designed to hunt human beings and destroy them."

Awaiting Breakthroughs

	The book cites two first attempts at creating such weapons that have reached at least the 
prototype stage. BRAVE 3000 is an airborne drone that "operates largely autonomously, 
penetrating enemy airspace to trigger a radar signal, then homing in on it to eliminate its source." 
The drone "searches for . . . its targets, in a sense 'deciding' to destroy a particular radar station on 
its own."

	The other example is PROWLER, an unmanned tank "endowed with lethal capabilities" 
(pictured in the book with small-caliber cannon and apparently a machine gun or two), "homing 
instincts" (informed by cameras and possibly other sensing apparatus), and capable of viewing 
humans as its "prey." Unlike similar remotely piloted tanks, which remain under the control of 
absentee human drivers, PROWLER is self-directed and could be given "far ranging 
reconnaissance and attack missions."

	De Landa understands that the perfection and deployment of such weapons must await 
"radical new breakthroughs" in artificial intelligence, the means by which the machines could gain 
true independence. Consequently, they may be "far in the future."

	Far in the future indeed. For AI applications like these, "radical breakthroughs" means 
solving the problems of artificial perception--the main reason AI has not attained many of its goals. 
The machines have to "see" well enough to sort out the things in their field of vision, and must not 
only identify threats but also rank them. Given a description of a situation, computers can reason 
fairly well about it. Trying to apprehend the environment on their own is the stumbling block.

	The problems of artificial perception make BRAVE a surer bet than PROWLER. The 
former has to be perceptive enough to avoid flying into mountains, as well as to discern military 
radar signals and make out the direction of their source--all in all, a much easier job than trying to 
pick out important features in video images of battlefields.

	Despite the difficulties, I will not be surprised to see a version of both weapons fielded. 
Success will come the way it always does in ground-breaking engineering projects: by narrowing 
the definition of success. Engineers discover what is possible, and managers eventually settle for 
that, provided it is still useful. PROWLER may never master entire battlefields, but it might patrol
perimeters or find paths through minefields.

	Although work on BRAVE and PROWLER--along with a 1984 document entitled 
"Strategic Computing," in which the Pentagon spells out its plans for AI research--alarm De 
Landa, it's not always clear whether he is troubled by the technology or by the military itself. The 
author can hardly refer to robot weapons without the deliberately sensational "predatory," and it 
gets old. Sure, the prospect of armed robots (in whom a little learning will truly be a dangerous 
thing) ought to provoke a reasonable disquietude. And the ability to engage in relatively bloodless 
war (on our side at least) might lead us to exercise it recklessly. Those are risks. But the author 
neglects to balance them against the even greater risks of sending our flesh-and-blood soldiers over 
the top. For De Landa, the alternative to high-tech war is no war. While it is sad that so much time, 
talent, and money go to producing sophisticated instruments of destruction, experience has shown 
that the alternative is low-tech war or no-tech war--war with spears or clubs or fists.

	Knee-jerk antipathy toward the military is not the book's only defect. Promotional material 
from the publisher describes De Landa's style as free-association; stream-of-consciousness is more 
like it. The writing is larded with jargon, and vast tracts left me wondering what the point might 
be. Hoping to impart an understanding of how computers and robots work and how they arose, 
the author takes us on wide orbits through history. But all the background produces more 
frustration than understanding.

Nature's Arms Race

	On one of these excursions, De Landa alights briefly on zoologist Richard Dawkins's view 
of evolution as an "arms race" among species. Pursued more fully, this concept is a powerful 
antidote to the author's railings against the military, as it implies that weapons may be simply a 
continuation of biological evolution by other means.

	Over millions of years, according to Dawkins, the genes of different species compete to 
evolve more effective offensive and defensive armaments. "The mutual stimulation of pairs like 
armor/claw or visual acuity/camouflage is what accounts for the advanced and complex machinery 
that animals and plants possess," De Landa quotes Dawkins as saying. Despite numerous and 
complex symbiotic relationships, most organisms do not peacefully coexist in a Bambi-esque 
paradise. Plants vie for sunlight, growing taller in successive generations. Animals struggle for 
food and mates, growing faster and stronger--or smarter--over the eons.

 	In this view, we are ourselves, along with all other plants and animals, armaments for our 
genes. Compared with other animals, humans are slow and weak, our "claws" hardly worth 
mentioning, our senses dull. But we compensate. Having evolved the ultimate natural weapons of 
big brains and opposable thumbs, we use them to fabricate tools from materials not part of our 
bodies. Swords and plowshares are both tools for life's work, both weapons in the struggle to
escape extinction.

I'm not sure it's important to distinguish between evolved and invented weaponry. In a sense they 
are equally "natural." Similar forces fashion both from the world's elements and work to boost the 
sophistication of both. The intelligent weapons De Landa decries may thus be inevitable.