Conference 7.286::digital

356.0. "And now, heeeere's AI!" by TELCOM::MCVAY (Pete McVay, VRO Telecom) Wed Aug 05 1987 07:15

    The following is extracted from SCIENCE magazine, the 31 July issue:
    
    "For software developers, the most important results of this research
    [into artificial intelligence] may not be the AI programs at all,
    but the AI programming style.
    
    "...Texas Instruments' W. Joe Watson made a very disconcerting
    statement: 'Most of us think that AI per se will lose its identity
    within about 5 years.'

    "Watson did not mean by this that the recent surge of interest in
    commercial AI applications is beginning to wane; if anything the
    AI industry is maturing and becoming better established.
    
    "...Watson told SCIENCE he meant that AI programming techniques
    are rapidly merging into the mainstream of computer science.
    
    "Watson is hardly alone in his assessment.  SCIENCE heard variations
    on the same theme from a number of companies at the [AAAI] conference:
    from a software developer's point of view, the most important single
    idea coming out of AI research is not a program per se, but a new
    approach to programming in general."
    
    Suspicions confirmed!  For years I've been bothered by AI: with
    the exception of very few programs (the VAX configuration program
    and SPEAR, to name a couple), most of the crap people have touted
    as being artificially intelligent were traditional programs with
    windows, graphics, and objects.
    
    Separating the method/tool from the results makes sense to me. 
    There are a lot of things about AI methods I liked, such as rapid
    prototyping, but I was always turned off by the snake-oil salesman
    approach.  I favor weeding out those programs that claim to be
    "artificially intelligent" or "expert systems" merely because they
    were written in LISP or some object-oriented language.

    Pete, it's  always  been  like that.  Each new approach has always been 
    the answer.  Consider the following:
    
         PL/1:  (Programing  Language One)--It married COBOL & FORTRAN, and 
         was supposed to be  the  best  of both.  How many PL/1 programmers 
         are there out there (that admit to it--other than moi?? ;-) )
         
         EDP:  Electronic Data Processing  (vs.  unit record, ie punch card 
         equipment).
         
         OA:  Office Automation, the cure  for  office  efficiency, or lack 
         thereof.  Hs it?
         
         DSS:  Decision Support Systems.  This was  the  "AI"  of  a decade 
         ago.  Seems to me I recall  a  company  that  produced  a  general 
         ledger  package  claim it was a DSS.   Yet,  DSS  are  man/machine 
         systems:    The  machine  supports  the  human  in judgement  call 
         situations.    I think it's stretching it a little to  classify  a 
         general  ledger  package  as  a  DSS.  Carried to its ultimate,  I 
         suppose AI would take the man (generic;  women included ;-) )  out 
         of the  equation,  and make it all machine.  Doubt that's possible 
         in its ultimate form, given today's technology.  Plenty of science 
         fiction around about future  scenarios, though.  Some of it pretty 
         believable.
         
    To me, it relates to "technology push" as  opposed  to  "demand  pull";  
    the  snake-oilers out there have something to sell, so  they  sell  it.  
    Whether it  meets  human  needs  or not is a different question.  Do we 
    focus too much  on  technology as a replacement for humanity?  Or, as a 
    cure-all for our(USA!) productivity declines in the world market?  And, 
    do  the snake-oilers latch on  to  that  as  a  selling  strategy?    I 
    wonder...