There seems to be a tendency in the current SDI debate to fall into an old engineering fallacy: that systems scale up linearly. Everyone seems to avoid this trap when talking about cost and effort--it seems to be well accepted that a 10-million line program is much harder than 10 1-million line programs-- but (most) people are *not* avoiding the trap when they speak of SDI's effectiveness. A recurrent argument seems to be that "SDI will be 80% [to use a number currently being bandied about] effective against a Soviet attack of N missiles; thus the Soviets would have to build and launch 5N missiles in order to have N missiles reach their targets, which would be economically ruinous." The implicit assumption is that if SDI is x% effective against N, it will continue to be x% effective against N'. This is fallacious unless x is very close to 0 or 100%. Assuming 80% effectiveness and 1000 missiles, SDI stops 800. Using the reasoning above, against 2000 missiles, SDI would stop 1600; but this cannot be so. If 1000 missiles strains the system to the point that it can only stop 800, why would anyone think it could stop more when the number of missiles and decoys is doubled, straining the system's ability to identify, track, and destroy missiles at least twice as much? Or to put it another way, if SDI could stop 1600 out of 2000, shouldn't it be able to stop 1600 out of, say, 1800 (1800 is surely an easier problem than 2000!). Or turn the argument around: if SDI can stop 800 out of 1000--80% effectiveness--does this mean it can stop only 80 out of a 100-missile attack? Or 8 out of a 10-missile attack? When anyone says that SDI will have such-and-such effectiveness, they must be made to state the assumptions used to calculate that effectiveness. Otherwise the numbers are meaningless.
A lot of recent RISKS messages have discussed one kind or another of aircraft accident. Many of the reports have included things like "The pilot thought [X] but in fact [Y]" or "[X] occurred, though the indications were that [Y] had occurred" or "[X], though there was no way for the flight crew to know that". So, what's going on in the area of improving flight crew/control system awareness of the state of basic external structures? Is anyone considering whether the FAA should require external cameras or periscopes so that (for instance) the pilot could find out that her entire vertical stabilizer had fallen off or her starboard outboard engine exploded? While there are many cases where the pilot would not, in any case, have time to check, there are also cases like the Japan Airlines crash where the plane stayed up for some time but the pilot had no way to determine the gross condition of the control surfaces. Some reports have said that that plane might have been saved if the pilot had known what he had to compensate for. Given that we are depending more and more on automated controls, should we be spending more effort on sensors that can determine more basic kinds of information? Should the control surfaces be instrumented so that the flight controls can tell the captain "Oh, the starboard outboard engine is no longer on its pylon and the outer flaps on that wing seem to be missing." as opposed to current systems just recognizing the effects of that loss and trying to compensate, with the risk that the operator will be unaware of the magnitude of that compensation and forced to guess at the state of the aircraft by observing what the control system is doing to deal with the effects of that state ("Oh, I'm having to turn the rudder vigorously to port to maintain my heading; can't say why."). scott preece, gould/csd - urbana uucp: ihnp4!uiucdcs!ccvaxa!preece
Henry Spencer writes: > A probable contributing factor here is that the US Navy's submarine people > do not trust automation at all in crucial roles... That's how deep the > distrust of complexity runs. I'm not surprised that they have manually- > controlled reactors. Then, he observes: > The USN also has an outstanding reactor safety record -- no big accidents, > no serious radiation releases -- with a stable of reactors comparable in > numbers (although not in output) to the entire US nuclear-power industry. > They are very fussy about materials, assembly, and operator training. Perhaps we should suspect that the safety record follows directly from the suspicion? Brint
I generally concur with Henry Spencer's accessment. The USN is very conservative about its use of proven technologies and reliability (also notice all new Navy jets have two engines [exclude older A-4, A-7, and F-8s]). But, while the Navy's record is certainly outstanding, I must point out there is a question about "no big accidents." One of the major contending theories on the loss of the USS Thresher in 1964 was sudden loss of reactor power. We will never really if this is the case, but it cannot ignored. Excellent reading about the safety record, the conservativitism, and the development of the nuclear navy is found in the 700+ page unauthorized biography of Rickover. --eugene
There is another factor to consider here, redundancy. Submariners are ALL cross trained EXTENSIVELY (the ideal is that everyone can do everything, usually they come fairly close to the ideal). Why, you may ask, does the Navy go to such lengths? The answer is fairly simple; these are WARSHIPS, they need to be able to function even after suffering SEVERE damage and heavy casualties. Just for normal day to day operations there are at least 2 people for every job (watch on and watch off), usually there are 3, often there are 4 or more. The following from net.aviation may be of interest to you. (ESP the quote). You may be interested in the whole discussion there. [scw] >From: wanttaja@ssc-vax.UUCP (Ronald J Wanttaja) >Newsgroups: net.aviation >Subject: Re: Problems with flying by the book (a pithy comment) >Date: 14 Oct 86 15:58:15 GMT >Organization: Boeing Aerospace Co., Seattle, WA <> I understand and appreciate your comments in the mod.risks about nth party/ <> hearsay stuff. But, from the examples you gave, in case you are really <> looking for some aviation accidents partially due to obedience to the <> "book", here are two - both commercial accidents at Toronto International <> (Now Pearson International). Both from MOT (then DOT) accident <> investigations: > [...] >"Rule books are paper: They will not cushion a sudden meeting of stone and > metal." > - Earnest K. Gann
The following is an op-ed piece that I wrote for the Rochester, NY, DEMOCRAT AND CHRONICLE. It appeared on page 4A on September 29, 1986. 'STAR WARS' CAN'T SUCCEED AS SHIELD, HAS OFFENSIVE CAPABILITY Can the Strategic Defense Initiative succeed? The answer depends critically on what you mean by success. Unfortunately, the public per- ception of the purpose of SDI differs dramatically from the actual goals of the program. In his original "Star Wars" speech, President Reagan called upon the scientific community to make nuclear weapons "impotent and obsolete." He has maintained ever since that this is the SDI goal: to develop an impenetrable defensive shield that would protect the American population from attack. With such a shield in place, nuclear missiles would be useless, and both the United States and the Soviet Union could disarm. Can such a shield be built? The most qualified minds in the coun- try say "no." In an unprecedented move, over 6,500 scientists and engineers at the nation's research Universities have signed a statement indicating that "Anti-ballistic missile defense of sufficient reliabil- ity to defend the population of the United States against a Soviet attack is not technically feasible." The signatures were drawn from over 110 campuses in 41 states, and include 15 Nobel Laureates in Phy- sics and Chemistry, and 57% of the combined faculties of the top 20 Phy- sics departments in the country. Given the usual political apathy of scientists and engineers, these numbers are absolutely staggering. The obstacles to population defense include a vast array of prob- lems in physics, optics, astronautics, computer science, economics, and logistics. Some of these problems can be solved with adequate funding for research; others cannot. Consider the single subject of software for "Star Wars" computers. As a researcher in parallel and distributed computing, I am in a position to speak on this subject with considerable confidence. The computer programs for population defense would span thousands of computers all over the planet and in space. They would constitute the single largest software system ever written. There is absolutely no way we could ever be sure that the software would work correctly. Why not? To begin with, we cannot anticipate every possible scenario in a Soviet attack. Human commanders cope with unexpected situations by drawing on their experience, their common sense, and their knack for military tactics. Computers have no such abilities. They can only deal with situations they were programmed in advance to expect. Before we can even start to write the programs for "Star Wars," we must predict every situation that might arise and every trick the Soviets might pull. Would you bet the future of the United States that the Rus- sians won't think of ANYTHING we haven't thought of first? Even if we could specify exactly what we want the computers to do, the task of translating that specification into flawless computer pro- grams would be beyond our capabilities for many, many years, possibly forever. Current and projected techniques for testing and quality con- trol may reduce the number of flaws in large computer systems, but actual use under real-life conditions will always uncover further "bugs." (For details on the software problem, see Dr. David Parnas's article in the October 1985 issue of AMERICAN SCIENTIST.) The only way to gain real confidence in "Star Wars" software would be to try it out in full-scale nuclear combat. Such testing is clearly not an option. But if effective population defense is impossible, why are we spending billions of dollars on SDI, and why are the Russians so upset about it? The answer is remarkably simple: because population defense is not the goal of SDI. The kinetic and directed energy devices being developed for the "Star Wars" program will have a tremendous range of uses in offensive weapons and in increasing the survivability of U.S. land-based missiles. The Soviets fear "Star Wars" for its first-strike capabilities. To make nuclear weapons impotent and obsolete, SDI would have to be perfect. To shoot down Soviet satellites, to thin out a pre-emptive strike on U.S. missile fields, or to develop exotic new weapons for the conventional battlefield, SDI will only need to succeed on a much more modest level. By focusing public attention on the unattainable goal of population defense, the Administration has managed to avoid discussion of the more practical, immediate consequences of SDI research. The weapons developed for "Star Wars" will have a profound impact on both our war- fighting strategy and our treaty obligations. That impact should be the subject of public and Congressional debate. By pretending to develop a defensive shield, the President has fooled the American people into funding a program that is far less clear-cut and benign. In effect, he has sold a system we cannot build in order to build a system he cannot sell. BYLINE: Michael L. Scott is an Assistant Professor of Computer Science at the University of Rochester. His article was co-signed by 10 other faculty members [almost the entire department] and 36 doctoral students and researchers. The views expressed should not be regarded as the official position of the University of Rochester or of its Computer Science Department. [We haven't had any RISKS mention of this topic in a long time. Perhaps it is time to dust it off again in the light of Reykjavik. The nature of the offensive capability is not a new issue, but is clearly an enormous potential RISK -- at least in the eyes of the Soviets. However, subsequent discussion on that issue probably belongs on ARMS-D. Let's once again try to stick to issues relevant to computers and related technologies. PGN]
Please report problems with the web pages to the maintainer