From: earl.dukerschein@doit.wisc.edu   
      
   Well,  I have crawled out on a limb (excellent support by Ted), I may   
   as well hang off the end.   
      
   Instead of the word "oracle" lets use the word "environment".   
      
   Instead of the word "boolean" lets use the word "predictable".   
      
   So, we sample the environment and learn rules.  A predictable rule   
   will be something   
   like: Sate + Action = new state.  An unpredictable rule will be   
   something like:   
   State + Action = state1 or state2 or ...   
      
   An action can be null, like letting gravity take its course.   
      
   When we plan to achieve a goal, we would like to use predictable   
   rules, thats obvious.   
      
   But when predictable rules fail us, we can use some unpredictable   
   rules.  If a   
   rule either does nothing or makes a desirable change, we can run the   
   rule repeatedly   
   until the desired change happens, or we get tired.   
      
   If a rule produces either a desired change or an undesired change, we   
   can seek a sub-loop   
   that recovers from the undesired change.   
      
   If we cannot plan a rule path to a desired goal (likely due to the   
   threat of combinatorial explosion),   
   we can plan a rule path to a state closer to the goal (closer being   
   defined in the posts above as   
   between two points on a Karnaugh map).  Once closer to a goal we can   
   try the planning process again.   
      
   Using the Karnaugh map paradigm again, we want to get from 0 (0000) to   
   F (1111).   We can only plan   
   to get to 0XX1, before a mini combinatorial explosion (at 0XX1 we may   
   need to consider rules for   
   for each of 4 possible points).  So we run the plan to get to 0XX1,   
   and find ourselves   
   at 3 (0011).  Now we try to plan to get from 3 to F.   
      
   [ comp.ai is moderated ... your article may take a while to appear. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)