Forgot your password?
Cyhist Dec 29 1996 C
Document Actions
Date: Sun, 29 Dec 1996 14:57:14 -0500
Reply-To: "CYHIST Community Memory: Discussion list on the History of
Cyberspace" <CYHIST@SJUVM.STJOHNS.EDU>
Sender: "CYHIST Community Memory: Discussion list on the History of
Cyberspace" <CYHIST@SJUVM.STJOHNS.EDU>
From: "A. Padgett Peterson P.E. Information Security"
<PADGETT@hobbes.orl.mmc.com>
Subject: cm> AI and be careful what you ask for...
______________________________________________________________________
Community Memory: Discussion List on the History of Cyberspace ______________________________________________________________________
Dave rote:
>Has anyone on the list worked in AI research at some point? If so, your input would be much appreciated.
Should know by the cows that I've been just about everywhere and learned a long time ago that things always get better with the retelling.
AI & "Expert Systems" were both dreams of lazy people (the ones that usually accomplished the most) but both had their flaws:
"Expert Systems" were the most achievable since were static rules based programs that consisted of asking experts a lot of yes/no questions and designing programs based on the answers. The more aggresive used CASE statements but all were attempts to leverage expert knowlege by applying canned responses to the "normal" problems freeing up the experts for deep study of the smaller portion that was left.
We see quite a bit of this type of programming in machine controls today.
AI was a different type of solution, essentially self-modifying programs that would "learn" what to do in a given situation. I worked quite a bit on this kind of programming with automatic test equipment. The idea was to shorten test cycles by allowing the machine to pick the order of tests which seemed proper. This same work was intended to predict failures by recording previous test results and watching for drift.
Unfortunately, more often than not, the machines in the field would develop "their own way" of doing things and start doing tests without the proper setup (programs usually were designed so that on the completion of one test, the UUT (unit under test) would require minimal change for the next - this meant that if the order changed, the machine had to consider not only the test order but the setup time for each. They were not quite advanced enough for that back then.
Best case we would receive failed equipment back from the field and not be able to duplicate the failure, often because the equipment was not at fault at all but rather had failed due to a subtle order/timing/value problem. Wost case you got max smoke.
Eventually it was decided that having repeatable conditions was more important than minimum time to repair and the people in the depot had enough trouble without the field equipment deciding how to do things (and not telling anyone) - people do that enough as it was.
Bottom line: an interesting experiment and as long as there was a guru present to keep the machine on the shining path with occasional minor tweaks, all was well. Many in the field on their own was just an exercise in Chaos theory.
Warmly,
Padgett
______________________________________________________________________
Reply-To: "CYHIST Community Memory: Discussion list on the History of
Cyberspace" <CYHIST@SJUVM.STJOHNS.EDU>
Sender: "CYHIST Community Memory: Discussion list on the History of
Cyberspace" <CYHIST@SJUVM.STJOHNS.EDU>
From: "A. Padgett Peterson P.E. Information Security"
<PADGETT@hobbes.orl.mmc.com>
Subject: cm> AI and be careful what you ask for...
______________________________________________________________________
Community Memory: Discussion List on the History of Cyberspace ______________________________________________________________________
Dave rote:
>Has anyone on the list worked in AI research at some point? If so, your input would be much appreciated.
Should know by the cows that I've been just about everywhere and learned a long time ago that things always get better with the retelling.
AI & "Expert Systems" were both dreams of lazy people (the ones that usually accomplished the most) but both had their flaws:
"Expert Systems" were the most achievable since were static rules based programs that consisted of asking experts a lot of yes/no questions and designing programs based on the answers. The more aggresive used CASE statements but all were attempts to leverage expert knowlege by applying canned responses to the "normal" problems freeing up the experts for deep study of the smaller portion that was left.
We see quite a bit of this type of programming in machine controls today.
AI was a different type of solution, essentially self-modifying programs that would "learn" what to do in a given situation. I worked quite a bit on this kind of programming with automatic test equipment. The idea was to shorten test cycles by allowing the machine to pick the order of tests which seemed proper. This same work was intended to predict failures by recording previous test results and watching for drift.
Unfortunately, more often than not, the machines in the field would develop "their own way" of doing things and start doing tests without the proper setup (programs usually were designed so that on the completion of one test, the UUT (unit under test) would require minimal change for the next - this meant that if the order changed, the machine had to consider not only the test order but the setup time for each. They were not quite advanced enough for that back then.
Best case we would receive failed equipment back from the field and not be able to duplicate the failure, often because the equipment was not at fault at all but rather had failed due to a subtle order/timing/value problem. Wost case you got max smoke.
Eventually it was decided that having repeatable conditions was more important than minimum time to repair and the people in the depot had enough trouble without the field equipment deciding how to do things (and not telling anyone) - people do that enough as it was.
Bottom line: an interesting experiment and as long as there was a guru present to keep the machine on the shining path with occasional minor tweaks, all was well. Many in the field on their own was just an exercise in Chaos theory.
Warmly,
Padgett
______________________________________________________________________
