From: bob.bechtel@comcast.net   
      
   Paolo Amoroso wrote:   
   > What happened to expert systems in industry within the past decade or   
   > so?  Are they still used?  Are they just another one of those "once it   
   > works, it's not AI" technologies?  Have expert systems been abandoned,   
   > or replaced by something else?   
   >   
   > Some context about my question.  I have been reading about, and   
   > experimenting with, expert systems for the past few months.  I bought   
   > from Amazon half a dozen well know books such as "Expert Systems -   
   > Principles and Programming" by Giarratano, "Programming Expert Systems   
   > in OPS5" by Brownston et al., "Building Expert Systems" by Hayes-Roth   
   > et al., and a few more.   
   >   
   > Although I have a basic knowledge of computing and AI history, I did   
   > not closely follow recent AI research and applications.  Many early   
   > expert systems books contain a lot of what might now be considered   
   > hype.  But one of the books I bought, Feigenbaum's "The Rise of the   
   > Expert Company", published in 1989, seem to tell several stories of   
   > how this technology improved the bottom line of many companies.   
   >   
   > To someone not deeply familiar with AI like me, all this seems to stop   
   > around the AI winter.  So, I always wondered what happened to expert   
   > systems since then.  I got Feigenbaum's book from Amazon at 0.01$ plus   
   > shipping: is this any indication of the fate of expert systems? :)   
   > Are expert systems alive and well?   
   >   
   >   
   > Paolo   
      
   Grumble. If you'll allow a quick definitional move, I'll suggest that   
   what was called "expert systems" in the books you've been reading would   
   more correctly be termed "rule-based systems." Now, that's not   
   completely true (for example, Carl Engelman insisted that Macsyma is an   
   expert system, even though it doesn't have rules in the conventional   
   sense, but it does have expertise), but I think that it may suffice for   
   your current purposes.   
      
   Given that, I'll claim that it's partly "once it works, it's not AI,"   
   partly a name change ("rule-based" rather than "expert"), partly that   
   (like almost every other technology) rules alone usually aren't enough,   
   so the tools have gotten more complex and may be described differently.   
      
   There are numerous tools available, both commercial and non-commercial.   
     Some of them can be traced directly back to the giants of AI summer -   
   Gensym Corporation makes G2;  Haley Systems has an OPS-descendant; as I   
   recall, there's still a descendant of ART*Enterprise going strong as   
   well, mostly in banking and insurance (I think).   
      
   A separate family holds CLIPS (originally NASA), JESS, a Java descendant   
   of CLIPS from Sandia, and LISA for Common Lispers.   
      
   In another renaming, rule-based technology has also been recast as   
   "business rules" - see BPEL, workflow, process control, etc.   
      
   You might look at comp.ai.shells (as in expert system shells or   
   rule-based shells) - while not super active, there is occasional useful   
   information.   
      
   There are also those that would argue that (if you squint a bit) logic   
   programming is just rules, opening up Prolog, constraints, and all those   
   wonderful technologies.   
      
   Rules are definitely alive. In the past year my company delivered a   
   custom application that used CLIPS to reason about information in   
   military messages - the use of rules made it possible for the customer   
   to tweak things to accommodate new message types without having to do   
   recompiles or have a full software development environment (Visual   
   Studio - they're a Microsoft shop) around.   
      
   bob bechtel   
      
   P.S. "The Rise of the Expert Company," being more a business book than a   
   technology book, has suffered the fate of most business books - after a   
   year or so, there's a new fad and a new book to read. I suspect his   
   Fifth Generation book is equally inexpensive.   
      
   [ comp.ai is moderated.  To submit, just post and be patient, or if ]   
   [ that fails mail your article to , and ]   
   [ ask your news administrator to fix the problems with your system. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)