Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site ubc-vision.CDN
Path: utzoo!utcsri!ubc-vision!banner
From: banner@ubc-vision.CDN (Allen Banner)
Newsgroups: can.politics
Subject: better dead
Message-ID: <964@ubc-vision.CDN>
Date: Mon, 10-Jun-85 17:51:52 EDT
Article-I.D.: ubc-visi.964
Posted: Mon Jun 10 17:51:52 1985
Date-Received: Tue, 11-Jun-85 13:29:24 EDT
Organization: UBC Computational Vision Lab, Vancouver, B.C., Canada
Lines: 51


> nuclear war would only occur if the Russians attacked us.  

Brad, as somebody employed in the field of computer science, you must
be fully aware that one of the grave dangers with the current trend in the
arms race is increasing reliance upon computers for control of weapons
systems.  The projects being undertaken in the Strategic Computing
Initiative and statements from Administration and DoD people show that the
military is hoping to use artificial intelligence to allow machines to fight
their battles autonomously or to take over tasks which are too difficult and
"unpredictable" for humans to handle.  Star Wars will, of course, be
entirely (or close to) autonomous.  If this trend continues, we are
begging for a disaster.  The chances of a nuclear war occurring as a result
of computer error (or by escalation after an initial error) increase greatly
from what they are now.

One of the unfortunate things when people start talking about "the risk
of..." is that the risk is not easily quantified.  What is the risk...high
or low?  Even if we cannot agree on a specific figure, it should be evident
that the risk of nuclear war is not going down!  I suggest that instead of
continuing this trend and saying "we have to because of the Soviet threat",
we should be looking around (and not too slowly either) for some other way
to ensure that they (and others as well) act in an acceptable manner.  For
the mutual good of *all* countries and people, we should stop considering
the threat of nuclear force as a viable means to make people act the way we
want them to.  Sooner or later, somebody may call our bluff or we may find
out that we have pushed our luck just a bit too far! 


Now I can hear people yelling out "But they will never deploy systems which
are unreliable or could bring on such a disaster...".  Two comments to that:

-- they may have no choice but to deploy such a system.  The Soviets are not
going to sit back and do nothing.  Once both sides are developing (or both
sides *perceive* that the other side is developing) such systems, there will
be tremendous pressure to produce something to counter it...and it may end
up being flakey because that's the best that can be done at the time!

-- under predictable circumstances perhaps the deployed system *won't* make
any mistakes (let's be optimistic...no bugs) but that is not to say that it
will not make a disasterous decision in an unforeseen circumstance!  In this
month's "Atlantic" magazine there is a good article addressing this very
issue.  There are quotes given from high ranking officials (I don't have it
right here in front of me right now) saying that some systems in SDI could
be shut off under most circumstances and then activated only in times of
crisis to avoid possible accidents...they are admitting that the system can
make mistakes if left on its own (who would have thought that!) but then to
top it off they are going to turn these dangerous parts on in times of
crisis (exactly when care should be the order of the day!)
-- 
Al Banner -- UBC Laboratory for Computational Vision