gardiner@umn-cs.CS.UMN.EDU (David Gardiner) (07/22/89)
From: gardiner@umn-cs.CS.UMN.EDU (David Gardiner) The recent discussions in various newsgroups (particularly rec.aviation) as well as the media on fly-by-wire lead to some general questions about computers and aircraft. As many sci.military readers know, there is much research into using intelligent computer systems to assist fighter pilots. The best known of these projects is the Pilot's Assistant being developed by Texas Instruments, I believe. (I do not know the exact status of this project but believe that it continues -- someone can undoubtedly correct any mistakes I make and add details). The question that I pose is: What are the comparative advantages of using a computer vs. a person? Also, how do fighter pilots feel about this issue, even assuming that the computer interface was good. The list of relative advantages that I have come up with so far is: 1) The computer saves space and weight, allowing a smaller aircraft and/or more fuel and/or weapons. 2) The computer will be more consistent than a person in doing the things it is supposed to do (i.e. if it is supposed to keep track of fuel state, it won't forget). 3) A person has eyes and, assuming s/he knows how to use them, is incomparably better for spotting than any computer vision system in existance. 4) While expert systems are neat gizmos, I have serious doubts about their ability to make life or death decisions in novel situations. Everything I have read about flying fighters indicates that there is no substitute for experience, and computer systems have limited learning capabilities at best. 5) Price. Which is cheaper, developing a high tech GIB (guy in back) or building bigger planes with two seats? I don't know. Consider also the price of training a human GIB. 6) If a plane gets shot down,crashes, etc. you only lose one person if the GIB is a computer. This has lots of economic, moral, political, etc. effects. 7) A computer won't leave the military to join the airlines (if they get _that_ good, we AI researchers will have made alot of progress :-) David Gardiner gardiner@umn-cs -- ========================================================== David A. Gardiner | "Lately there has been an University of Minnesota | alarming increase in things gardiner@umn-cs | about which I know nothing."
aws@vax3.iti.org (Allen W. Sherzer) (07/24/89)
From: aws@vax3.iti.org (Allen W. Sherzer) In article <8520@cbnews.ATT.COM> gardiner@umn-cs.CS.UMN.EDU (David Gardiner) writes: >The recent discussions in various newsgroups (particularly rec.aviation) >as well as the media on fly-by-wire lead to some general questions about >computers and aircraft. As many sci.military readers know, there is >much research into using intelligent computer systems to assist fighter >pilots. The best known of these projects is the Pilot's Assistant being >developed by Texas Instruments, I believe. It was a dual award. McDonald Douglas was the leader on one wiht TI the sub. The other team was Lockheed with about 7 subs including GE, Technowledge, and others. I worked on the MCAIR TI team and designed the Tactical Planning component. >(I do not know the exact >status of this project but believe that it continues -- someone can >undoubtedly correct any mistakes I make and add details). I no longer work at TI and believe they are no longer on the team. Phase two of the program ended last Feb and the Air Force is now looking to see what can go into ATF. >The question that I pose is: What are the comparative advantages of using >a computer vs. a person? Also, how do fighter pilots feel about this >issue, even assuming that the computer interface was good. The main advantage for a computer is that it is a LOT cheaper than adding a second person. Not only in terms of training and paying the new person but that two person aircraft are a lot more expensive and add a lot more weight. As to the pilots, most I have spoken with feel there is indeed a need for something like PA. They will however, be very paranoid aobut it for a while untill they have tested it (an understandable position). Their main concerns in development was that the pilot stay in control and that PA must have an off switch. >The list of relative advantages that I have come up with so far is: > >1) The computer saves space and weight, allowing a smaller aircraft >and/or more fuel and/or weapons. According to our prime (MCAIR), adding another person adds about 10,000 pounds of weight to the aircraft. >2) The computer will be more consistent than a person in doing the >things it is supposed to do (i.e. if it is supposed to keep track of fuel >state, it won't forget). Both true and important for emergency procedures. >3) A person has eyes and, assuming s/he knows how to use them, is >incomparably better for spotting than any computer vision system in >existance. true >4) While expert systems are neat gizmos, >I have serious doubts about their ability to make life or death >decisions in novel situations. Everything I have read about flying >fighters indicates that there is no substitute for experience, and >computer systems have limited learning capabilities at best. I agree. PA makes no decisions unless the polot authorizes it before. PA's main job (again, when I was on it so it may have changed) was to watch the pilot's 6 and let the pilot do his job. >5) Price. Which is cheaper, developing a high tech GIB (guy in back) or >building bigger planes with two seats? I don't know. Consider also the >price of training a human GIB. I think PA will be a lot cheaper. Allen ---------------------------------------------------------------------------- | Allen Sherzer | PEACE | | aws@iti.org | through superior firepower | ----------------------------------------------------------------------------