Copy Link
Add to Bookmark
Report
AIList Digest Volume 8 Issue 021
AIList Digest Friday, 22 Jul 1988 Volume 8 : Issue 21
Today's Topics:
Does AI kill? -- Fourth in a series ...
----------------------------------------------------------------------
Date: 19 Jul 88 16:47:02 GMT
From: mcvax!ukc!etive!aiva!ken@uunet.uu.net (Ken Johnson)
Subject: Re: does AI kill?
In article <12400014@iuvax> smythe@iuvax.cs.indiana.edu writes:
> I don't know which event involved the
> Sheffield, but there was no misidentification in either case.
I had also heard the story that the Exocet, having been sold to the
Argentinians by the Europeans, was not identified as a hostile missile.
Subsequently the computers were `reprogrammed' (media talk for giving
them a wee bit of new data.) Presumably if you sell arms to your enemies
this is what you must expect.
--
------------------------------------------------------------------------------
>From Ken Johnson, AI Applications Institute, The University, EDINBURGH
Phone 031-225 4464 ext 212
Email k.johnson@ed.ac.uk
------------------------------
Date: 20 Jul 88 06:31:00 EDT
From: "CUGINI, JOHN" <cugini@ecf.icst.nbs.gov>
Reply-to: "CUGINI, JOHN" <cugini@ecf.icst.nbs.gov>
Subject: Does AI kill?
[ please excuse if this is repetition - our mailer has been berserk lately ]
Two points I haven't seen made so far...
1. Are AI systems to be held to a standard of perfection? I don't know
of *any* kind of system, constructed by humans, that doesn't fail -
airplanes crash, walkways collapse, nuclear power plants explode.
And, yes, people die ... so the issue isn't whether AI/computer
systems will ever fail, causing loss of life - they will, be assured.
But so would the non-computer sustems which are the alternative.
Moreover, there will be instances (maybe the Vincennes was one, maybe not)
in which a non-AI system would've made a better choice. Big deal.
The serious question is, on average, will the performance of a system
be enhanced (wrt whatever criteria you like - saving lives, etc.)
or degraded by use of AI components.
The Vincennes critics should make the case (if they can) that the AEGIS
system caused the shoot-down, that it wouldn't have occurred otherwise,
and that AEGIS has no compensating effects (maybe it's already saved
291 lives by deterring Iranian attacks on American ships...).
2. For what it's worth, I think it's a cheap shot to use this incident
as a excuse to hold forth on one's political views (on AI list). The
AI-component debate is proper, but I'd urge some self-restraint before
we start pontificating on SDI, etc etc. The "larger lesson" of the
Vincennes, like many other "lessons" of recent history, seems to depend
much more on one's prior political views than on any unambiguous
interpretation of events.
John Cugini <Cugini@ecf.icst.nbs.gov>
------------------------------
Date: Wed, 20 Jul 88 10:02 EST
From: <INS_ATGE%JHUVMS.BITNET@mitvma.mit.edu>
Subject: Vincennes and AI
Lets examine the faults of the recent airbus downing in the gulf.
A commercial airbus supposedly carrying innocent civilians was shot
down by an Aegis cruiser. Evidently, there is reason to suspect that
the airbus did not carry proper IFF transponder identification.
The Aegis was not able to properly identify the aircraft as friendly,
and since it was recently involved in a skirmish, shot the aircraft down.
Did the Navy realize that a commercial airbus with incorrect IFF
transponders could be shot down in the Persian Gulf? If not, then
there was a severe problem in the Navy's understanding of how their
equipment operates. From what I have heard, the Aegis is designed to
shoot down large numbers of unfriendly aircraft, not neccessarily
designed to avoid shooting down nearby unidentified aircraft.
I don't see how the Navy could have placed such a killing system in the
Gulf without the realization that there would be danger to civilian
aircraft in the tight waters.
Furthermore, it has often been noted in the literature that in wars,
innocent civilians are killed. Allow me to define war as a state of
hostility between states which allows at least one of those states
to injure some part of the other state's population.
It would be nice if we could have nice clean wars, but unfortunately,
they do not exist.
But I have strayed off the subject. I find this incident a possible
example of mal-understanding of how a system works in a situation,
if this incident did indeed suprise anyone in the Pentagon.
The systems in the Persian Gulf are going to have a hard time identifying
friend and foe, and there may indeed be increased use of commercial
transponders on military aircraft in the region if this is possible.
--As we noted in Vietnam, even "real intelligence" cannot always
discover who is friend, and who is foe.
Thomas Edwards
ins_atge@jhuvms
------------------------------
Date: 20 Jul 88 22:38:03 GMT
From: smithj@marlin.nosc.mil (James Smith)
Subject: Re: does AI kill?
In article <1376@daisy.UUCP>, klee@daisy.UUCP (Ken Lee) writes:
> Computer-generated mistakes aboard the USS Vincennes may lie at the root
> of the downing of Iran Air Flight 655 last week, according to senior
> military officials being briefed on the disaster.
> ...
> The officials said Rogers believed the machines - which wrongly
> identified the approaching plane as hostile - and fired two missiles at
> the passenger plane, knocking it out of the sky over the Strait of
> Hormuz.
> ...
> Some obvious questions right now are:
> 1. is AI theory useful for meaningful real world applications?
> 2. is AI engineering capable of generating reliable applications?
> 3. should AI be used for life-and-death applications like this?
>The blame lies entirely on the captain of the
>USS Vincennes and his superiors for using the system in a zone where
>commercial flights are flying.
In all of this debating over whether-or-not AEGIS should be used in the gulf,
and whether-or-not Captain Rogers erred in his decision to shoot at
IAF 655, one crucial point has been overlooked - there is _no other_
combat direction system (in our, or any other navy) which can even begin
to cope with the volume of information that is efficiently processed
and displayed by AEGIS. The commanding officer of a pre-AEGIS ship would
have had far less time from target detection to shoot decision; he
would also have had a less-precise radar track and IFF information, and
would have had to make the shoot/no-shoot decision at a greater range
than Captain Rogers.
This is not a problem of applying AI but, rather, a problem requiring
an immediate life-or-death decision made, not in the laboratory or
the office, but in what Klausewitz referred to as the *fog of war*.
Jim Smith
UUCP: smithj!marlin!nosc
DDN: smithj@marlin.nosc.mil
If we weren't crazy, we'd all go insane
- Jimmy Buffet
------------------------------
Date: 21 Jul 88 01:14:44 GMT
From: uvaarpa!virginia!uvacs!cfh6r@umd5.umd.edu (Carl F. Huber)
Subject: Re: does AI kill?
In article <470@proxftl.UUCP> tomh@proxftl.UUCP (Tom Holroyd) writes:
> ...
>Now, before y'all start firing cruise missiles at me, I am *NOT*, I repeat
>NOT praising the system that killed 280 people.
Let's be careful. That system didn't kill anyone.
------------------------------
Date: 21 Jul 88 11:30:12 GMT
From: amelia!prandtl.nas.nasa.gov!msf@ames.arpa (Michael S.
Fischbein)
Subject: Re: does AI kill?
In article <1054@marlin.NOSC.MIL> James Smith writes:
>In all of this debating over whether-or-not AEGIS should be used in the gulf,
>and whether-or-not Captain Rogers erred in his decision to shoot at
>IAF 655, one crucial point has been overlooked - there is _no other_
>combat direction system (in our, or any other navy) which can even begin
>to cope with the volume of information that is efficiently processed
>and displayed by AEGIS.
Nonsense. The Navy Tactical Data System (NTDS), the pre-AEGIS computerized
sensor/fire control computer system had very similiar capabilities as
far as tracking and individual's displays go. AEGIS supports the large
`wallboard' displays which were not supported by NTDS until recently;
budget constraints have prevented retrofitting NTDS ships with the large
screens. AEGIS is only present on the Ticonderoga class AAW cruisers
with the SPY-1 radar; other primary AAW combatants use NTDS and other radar
systems, such as the SPS-48E and will continue to do so.
> The commanding officer of a pre-AEGIS ship would
>have had far less time from target detection to shoot decision; he
>would also have had a less-precise radar track and IFF information, and
IFF has little or nothing to do with AEGIS; I would appreciate any reference
that compares the SPY-1 to the SPS-48 (current models of each) and shows
significantly greater precision to either. The SPY-1 is faster as it
doesn't rotate, but from my (admittedly slightly out-of-date) personal
experience, they can offer comparable performance.
>would have had to make the shoot/no-shoot decision at a greater range
>than Captain Rogers.
>
>This is not a problem of applying AI but, rather, a problem requiring
>an immediate life-or-death decision made, not in the laboratory or
>the office, but in what Klausewitz referred to as the *fog of war*.
Absolutely true. If you haven't been there, people, try to think about
what it was like on the ship.
mike
Michael Fischbein msf@ames-nas.nas.nasa.gov
...!seismo!decuac!csmunix!icase!msf
These are my opinions and not necessarily official views of any
organization.
------------------------------
Date: 21 Jul 88 17:24:59 GMT
From: lakesys!tomk@csd1.milw.wisc.edu (Tom Kopp)
Subject: AI...Shoot/No Shoot
It seems a lot of this argument upon the AI systems (or whatever you wish to
call them....information sorters...AI routines...whatever) Stems from how the
Captain acted up on it in the situation.
Someone brought up the phrase "Shoot/No Shoot" and that reminded me that
Police officers in many areas go through a special shoot/no shoot test
regarding the use of their firearms. Does anybody know if Naval Command
Candidates go through similar testing on a simulated ship, or what?
Looked at in this light, I can't see where he had any choice BUT to shoot,
based upon what his computers were telling him. If it were indeed a loaded
passenger get w/ civilians on board, then I of course, regret the action, but
that doesn't change the situation. He couldn't possibly get a visual fix on
the target, his computers were warning him of a threat, and an unidentified
aircraft on a 100% direct course to his ship was descending at an angle that
would very soon bring it into attack position.
I still don't understand WHY the computers mis-lead the crew as to the type of
aircraft, especially at that range. I know that the Military has tested
heavily some proposed radar gear for the Tomcat (and possibly other planes)
that is capable of target identification based upon the radar signature, and
having the target head on only helps. It counts the number of blades on the
turbofan and thus knows what kind of engine it is, and thus narrows the
possibility down to a very few aircraft, often even to one. Either this is
not installed on the AEGIS ships or it was malfunctioning....It also may not
even be completed yet....I read about it in Aviation Week/Space Technology or
something a year or so ago....forgot just where I saw it...
------------------------------
Date: Thu 21 Jul 88 11:51:54-PDT
From: Conrad Bock <BOCK@INTELLICORP.ARPA>
Subject: Re: Does AI kill?
Overautomation can kill if the task is to decide when to kill. We have
all experienced programs that try to be too smart. Both industry and
military institutions would like to take the human out of the loop
(more control for the people at the top), hence overautomation. I
believe this is the critique of Dreyfus and Dreyfus.
Conrad Bock
------------------------------
End of AIList Digest
********************
