LEVEL 1 - 49 OF 65 STORIES
Copyright 1987 Computerworld, Inc.
Computerworld

November 30, 1987

SECTION: VIEWPOINT; Pg. 17

LENGTH: 1082 words

HEADLINE: Star Wars: When imperfect man strives to make perfect machines

BYLINE: By Harvey Newquist; Newquist writes and consults on artificial
intelligence and other advanced high-technology topics from his office in
Scottsdale, Ariz.

BODY:
	This is a story about machines and maintaining them.

	I have a washing machine that has been serviced six times in the last three months.

	My girlfriend's portable computer has had several hard disk crashes in the last few weeks. 
Dozens of innocent files were destroyed.

	My brother's compact disk player gets stuck on the eighth track of every CD he plays.

	My father's Audi 5000 is a story about vehicular machines unto itself.

	My country's president wants to put mechanical systems into space that will decide for 
themselves how to employ destructive missiles and lasers -- "Star Wars."

	Man has had a so-so history of making perfect machines. His performance in servicing 
them and keeping them in proper working order is even less impressive. As the Reagan 
administration draws to a close and computer technology and artificial intelligence capabilities rush 
forward at an astounding pace, the U.S. military is trying to combine them and figure out how to 
put the ultimate weapon into the atmosphere. Unmanned.

	I believe that a complete Star Wars system can be sent into outer space in the next 10 years; 
I'm just not so sure it will work.
 
	The inspiration for these thoughts comes from two sources: the everyday world, which 
suffers at the hands of malfunctioning machines more often than we would like, and a book 
entitled Computers In Battle -- Will They Work?  edited by David Bellin and Gary Chapman 
(Harcourt Brace Jovanovich).

	The book raises a lot of questions that would give you goose bumps on the warmest of 
days. It is especially disconcerting to those of us who have dealt with computer evolution in the 
last few years. Computers are wonderful machines,but they're never perfect. The Department of 
Defense plan calls for perfect computers. It doesn't know where the perfect computers will come 
from, but it is basing Star Wars on them.

	Bellin and Chapman present essays from noted researchers and military watchers, most of 
whom are deeply involved in both hardware and software development. Given the current state of 
computer technology, the book holds Star Wars as an unrealistic effort.

	And Star Wars plans to use the most advanced of the advanced: artificial intelligence, expert 
systems, pattern recognition, fault-tolerant and parallel processing, autonomous decision making 
and independent, self-guided operation. Part of the problem is that each of these computer 
techniques is still in its infancy, and many have not been commercially (or developmentally) 
proven.
 
Fixing bugs on the fly

This fact may not be a big deal in the industrial and corporate markets, where fixing bugs and 
glitches on the fly is a way of life after system installation. However, when your system operates 
several miles straight up in the air, with no possibility of repairmen making a service call, some of 
the drawbacks of a deployed space weapons system become apparent.

These systems aren't doing batch processing or inventory control, relatively routine functions by 
normal computing standards. They are designed to dispense "kill power" based on an ability to 
recognize incoming missiles in a matter of seconds and then decide whether to destroy, intercept or 
ignore them.  This capability presupposes the system can tell if the weapons it senses are friend, 
foe, neutral or missiles at all, as opposed to deactivated satellites or maybe large fires on the 
Earth's surface. Oh, yes, the machine must decide in the time it took you to read this paragraph.

	Star Wars systems cannot be tested until they are required for use. You can't beta test them 
half a mile over the New Mexico desert because the real-time environment in space is not the same. 
You can't fire thermonuclear warheads at them and see how they'll react in some test lab up in 
northern California.
 
You have to put them in place and hope for the best. And if someone forgot to stabilize a disk drive 
(as they did with my girlfriend's portable), it's pretty much tough luck. You don't just send 
someone out to repair an oversight somewhere in orbit over earth.

	In order to work in real time in space, the machines must be autonomous --make their own 
decisions. This ability is one of DOD's biggest pursuits in a number of areas, not just Star Wars. It 
is planning an  unmanned tank with decision-making capabilities called the Autonomous Land 
Vehicle for the U.S. Army. It is investing in a computerized co-pilot for the U.S. Air Force called
the Pilot's Associate, which will take some of the work load off the man in the cockpit. A Battle 
Management System that will plan strategy for the U.S. Navy faster than its human counterparts is 
in the works.

	The upshot is that humans are being taken out of the decision-making loop, yet they live on 
on the receiving end -- where the "kill power" goes. Because of the need for split-nanosecond 
timing, there is no time for human interaction to judge whether to activate weapons. We rely then 
on the perfect operation of a machine created by imperfect humans.

	Leaving such a machine unattended makes me nervous. I won't even let my printer go for 
five minutes without double-checking to make sure it hasn't screwed up.

	There is a great scene in the computer-fantasy movie Westworld when the lead robot states 
that computer operation in the facility is perfect and that, "Nothing can go wrong . . . go wrong . . 
. go wrong . . . go wrong . . ."

Californian threats

	Suppose for an instant that an autonomous Star Wars system decided most of southern 
California was actually a threat to mankind's survival, having mistaken some of the annual San 
Fernando Valley fires as heat from a full-scale weapons launch. A few well-placed laser and 
missile blasts could leave the U.S.'s playground looking like the Sahara. Unfortunately, this 
misplaced attack would probably happen so fast that there would be no time to enact surface 
defense systems.

 	I could bring to the argument other computerized weapons systems, like the infamous Sgt. 
York tank, but that's a whole other story about computers in battle. Suffice to say that the billions 
of lines of code and the sheer flawlessness required to make Star Wars a functioning reality is 
daunting, aside from the moral question of littering the sky with weapons.

	For a society that has almost forgotten what it is like to have machines work the first time 
every time, the prospect of a fully enabled computerized space weapon doesn't make it any easier 
to sleep at night.