havens@ubc-vision.CDN (Bill Havens) (04/03/85)
In reply to henry@utzoo: Be careful in describing other people's analyses as hysteria. It is quite useful to concentrate on establishing the facts from which informed debate may proceed. However, you have gone beyond that goal in a personal way. The flurry of activity on this network is not a debating exercise. We are concerned about a very serious change in Canadian Science and an even more serious escalation of weapons for mass destruction. My abstract about the dangers of relying on computers to achieve reliable automatic strategic defense was not hysterical. It was factual. The information that I presented was taken from a number of recent papers published by Alan Borning, Severo Ornstein and others in such journals as "The Bulletin of the Atomic Scientists" and soon in the CACM. I quoted their findings and analyses correctly. In particular, you dismiss their assertions that we are inexorably moving towards "launch on warning" and the removal of human (read Presidential) decision making about starting nuclear war. Your reason for dismissing these frightening facts is that you have heard no suggestion that the US should change its public policy of Presidential command. To the contrary, I specifically outlined testimony to Congress where these facts were admitted by SDI representatives. To repeat - the Star Wars technology is only an effective deterrent if it can be used during the "boost phase" of the rising Russian missiles. The 60-second decision time required NECESSITATES automatic decision making and automatic "nuclear war fighting". Its not a matter of current US policy. It is dictated directly by the technology. In the second part of your response to my argument, you miss the point about Nuclear Power Systems. A technology is "safe" only if we are willing to accept the consequences of infrequent but inevitable system failures. In an accidental nuclear power plant "meltdown", society survives. In an accidental nuclear holocaust, society and possibly the planet both die! Bill Havens.....
henry@utzoo.UUCP (Henry Spencer) (04/03/85)
> In particular, you dismiss their assertions that we are inexorably moving > towards "launch on warning" and the removal of human (read Presidential) > decision making about starting nuclear war. My mistake; I am aware of the suggestions that "launch on warning" will be increasingly attractive due to steadily-reduced decision times. What I'd like to know is, what does "launch on warning" of offensive weapons have to do with automatic initiation of defensive weapons? > the Star Wars technology is only an effective deterrent if it > can be used during the "boost phase" of the rising Russian missiles. The Star Wars technology is a *defence*, not a *deterrent*. These are two very different animals. A defence protects against an actual attack; a deterrent attempts to avert attacks by frightening the opposition. (By the way, some of the suggested SDI methods do not rely on boost-phase interception, although it is the most attractive time to do it.) > The 60-second decision time required NECESSITATES automatic decision making > and automatic "nuclear war fighting". I'm willing to go along with the first half, but not the second. I see no reason why initiation of defensive systems, i.e. SDI, need have anything to do with initiation of offensive systems, i.e. nuclear weapons. They are two quite separate issues. In fact, it has been suggested that SDI would substantially increase the available decision time for launching offensive weapons, since it would interfere severely with any attempt to quickly destroy offensive systems. I agree with your point that a technology is "safe" only if we can live with occasional failures, although your example of nuclear power was singularly ill-chosen, since accidental failures are generally less dangerous there than almost anywhere else. (Nuclear power plants are better protected against accidents than almost any other technology, including many that handle dangerous chemicals or explosive fuels in large quantities.) Clearly, however, an accidental failure in a *defence* system is far less dangerous than an accidental failure in a *deterrent* system. SDI systems do not launch nuclear missiles; they shoot down missiles that have already been launched by someone else. I repeat, the worst consequence of accidental initiation of an SDI system is shooting down a manned space launch. This would be regrettable, but surely we can live with the risk. -- Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,linus,decvax}!utzoo!henry
lesperan@utai.UUCP (Yves Lesperance) (04/04/85)
In <5407@utzoo.UUCP>, Henry Spencer says: >My mistake; I am aware of the suggestions that "launch on warning" will >be increasingly attractive due to steadily-reduced decision times. What >I'd like to know is, what does "launch on warning" of offensive weapons >have to do with automatic initiation of defensive weapons? > ... I see no reason why initiation of defensive systems, i.e. SDI, need >have anything to do with initiation of offensive systems, i.e. nuclear >weapons. ... SDI systems do not launch nuclear missiles; >they shoot down missiles that have already been launched by someone >else. I repeat, the worst consequence of accidental initiation of an >SDI system is shooting down a manned space launch. This would be >regrettable, but surely we can live with the risk. This ignores the fact that offensive and defensive systems are likely to be linked in their operation. Moreover the opponent's perception of threat is based on the characteristics of the combined system. Consider the following senario: assume that both superpowers have SDI-type systems together with all the surveillance hardware that is necessary to make them work. The US system goes on alert as a result of some situation that was not anticipated in the system design; these false alarms are said to happen on a daily basis in current systems. The Soviet system gets reports of the US system going on alert and reacts by doing the same. This leads the US system to go on a higher level of alert, and so on. At some point the US system, which must make a decision while the incoming missiles are still in boost phase, that is in about one minute, decides to fire at the percieved missiles and maybe also at the Soviet surveillance satellites. All this may happen with next to no human involvement due to the short decision time. Now the Soviets know that the SDI system is poor protection against a first strike, but could work well against a second strike. They take the US action as the beginning of an attack (with good evidence), and so they fire their missiles. I view this kind of scenario as very plausible. The recent changes in the nuclear scene, that is, cruise missiles, short and medium range missiles, and stealth technology, have all decreased decision times and increased uncertainty; they have done nothing to increase our security. SDI will extend this mad trend. Because of these changes, our destiny will be put in the hands of ``expert'' computer systems. The Strategic Computing Initiative document makes clear that this is why they want these AI defence systems although it is less candid as to where the technology will be used. But nobody knows how to program common-sense either now or for the forseeable future. So the ``expert'' systems will follow their rigid rules to their very end, as well as ours. Yves Lesperance utcsri!utai!lesperan
henry@utzoo.UUCP (Henry Spencer) (04/05/85)
> .... At some point the US system, which must make a decision [fast] > decides to fire at the percieved missiles and *maybe* also at the Soviet > surveillance satellites. [emphasis added] Firing on surveillance satellites is an insane thing to do, because (as you point out) it's a very threatening move. I would be very surprised to see missile defence and antisatellite attack combined under the same automatic-response system; it's too dangerous. Furthermore, there is no terribly good reason for it: the missile-warning satellites tend to be in geostationary orbit, much too high for most proposed SDI systems to attack them effectively, and hence it isn't even the same hardware doing the job. You have constructed a frightening scenario, all right, but it's based on the same assumption I was attacking in the message you cited: that a system which *must* have super-fast and hence automatic response will also command other, much more dangerous, systems into action. This is definitely a possibility which needs to be guarded against, but there are enough false alarms in existing systems (as you point out) that it is *most* unlikely that anything which didn't *absolutely* *have* to have lightning response would be placed under fully automatic control. There is no need for 60-second decisions about attacking satellites. > ... Because of [shortening decision times], our destiny will be put > in the hands of ``expert'' computer systems. The Strategic Computing > Initiative document makes clear that this is why they want these AI defence > systems ... > But nobody knows how to program common-sense either now or for the forseeable > future. So the ``expert'' systems will follow their rigid rules to their > very end, as well as ours. All the more reason to support a system that lengthens decision times on the really bad weapons, by reducing the fear of sudden obliteration that motivates worrisome ideas like "launch on warning". -- Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,linus,decvax}!utzoo!henry