On Artifical Intelligence

Discussion in 'Intelligence & Machines' started by geistkiesel, Jun 27, 2007.

Thread Status:
Not open for further replies.
  1. geistkiesel Valued Senior Member

    Messages:
    2,471
    When working for a major Defense contractor I went through a 6 months 8 hour per day class, 15 of us, in the first AI.class. There are many sides to the generic reference to AI. All sides have one common element - computers. Like the old saying goes, "garbage in, garbage out."

    Computers replacing humnan cognition ain't gonna be seen by any of us.

    Another, facet, and to me the most interesting, waqs developing "expert systems". This simply means one takes an "extert" interview them and codes their expertice. Brain surgeons and fighter pilots are not good candidates.

    An example: Campbell soup has this soup can washing facility. The cans go in one end and progress through a series of washing cycles. The mainteenance expert was about to retiire so the "coded him" Typically, when at hone he would get phone calls when something failed. He always had a well structured way he went about explainingh wha to do. The AI folks, got about 97% of his expertise, and he retired in peace.The computer team had his expertise boiled down to appoximately 20 rules, where each might interact with one or more of the others..

    The expert then is one whose expertise is is a limited discipline.limited

    I worked on a program for analyzing U-2 spy plane data of Soviet bloc SAM sites. Each Soviet Reinforced Infabntry Division had a regiment of SAM batteries attached. The SAM batteries were constantly being tested using microwave crystals in their radar systems. A particlular regiment might have 50 SAMN batteries. Each missile launcher has an assigned set of frequencies they used in their radars. The U2 would fly along the border and detect the testing of Soviet bloc rardar and were able to record two things, signal frequency and location.

    From many flights the Western Forces were able to ibuild a data base and identify the unit and where is was located. Army intelligence analysist did this by hand going through reams of data. One flight would take half a day to analyze by a dozen analysts.

    My program used an expert, working for my company. I got about 92 % of him coded. He had been an arny analkyst before recruioted by the defense firm.

    One example was a case where we couldn't determine if one regiment had moved to Poland or was in Ruman ia. We knew how long a regiment took to move and we had it located at one time, but we couldn't be sur so we were lost.

    Finally, my expert came with the answer and we located the rgiment. I asked him what frequency combination and travel speed scenario he used. None of the above he said. He received some information from some secret agency in DC who received a reposrt from a "friend of ours " on the ground in Eastern Europe. THis source on the ground saw the regiment when it was moving and told his contact by radio.

    So expert systems have a number of aspects.

    The super computer systems are best utilized with somne combination of expert systems and algorithm code.

    I asked my expert once how did he know that tehSoviets wouldn't secretly install a bunch of ubdetected crystals and confuce the shit oiut of maybe even laubch a secrfet attactk One very important aspect of the ceystals the Soviuets used was avoidiung one crystal frequency interferring with another. This is whuy they were constantly testing different combinations.The bottom line was " An anrmy fights like it trains.

    Please Register or Log in to view the hidden image!



    Geistkiesel
     
Thread Status:
Not open for further replies.

Share This Page