From: garyfrankNOSPAM@itol.com   
      
   "Fred"  wrote in message news:bnd5p7$3ba$1@mulga.cs.mu.OZ.AU...   
      
      
   > This is an odd question so I'll assume I am misinterpreting it.  The   
   reason   
   > your question is odd is because it seems obvious that all AI programs or   
   AI   
   > systems are going to be rule based.   
      
      
   > The biggest problem in AI seems to be trying to recognise image and sound   
   > data. I reckon that lots of AI programmers realise that it is pointless   
   > making rule finding systems when you don't have the data thats worth   
   finding   
   > rules on. The actual rule building process should be relatively easy and   
   > will probably get the spotlight later when the image and sound recogntion   
   > problem has been solved.   
      
   You brought together 2 important items - that rules and sensed data both   
   seem to be integral to AI.  I also think that advances in  image and sound   
   recognition will improve AI.  The reason is that sensed data is captured as   
   a group of small pieces, such as the group of pixels that constitute an   
   image.  A collection of small pieces is a necessary way to represent all   
   concepts, both the immediately sensed inputs and the more abstract concepts   
   that have been derived from those inputs.   
      
   Perhaps "match processing" would be more significant than "rule processing".   
   It seems obvious that object recognition is a matter of matching a group of   
   inputs such as the pixels in an image with the name of an object.  But other   
   cognitive functions such as those that rule based processing attempts to   
   perform are also matching functions.  If / then processing matches one thing   
   with another.  That can be accomplished by a mechanism that processes groups   
   of small parts, such as a pattern of charges on a group of neurons that was   
   charged by a corresponding group of pixels, as well as by the traditional   
   expert system rule processor.  The advantage to an ANN based rule processor   
   is that it would not be subject to the brittleness that is the undoing of   
   text based rule processors.  The fact that "A" consists of a pattern of   
   charges that represents a bird instead of the word "bird" allows matches to   
   occur to a certain degree, not just all or nothing.   
      
   Gary Frank   
      
   [ 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)