From: kym@sdf.lonestar.org   
      
   Randolph M. Jones  wrote:   
   > Gooseman wrote:   
   >> I have a quote from Van Parunak that says "...as agents get smarter,   
   >> their functionality goes down".   
   >> Does any one have a direct reference to this - something that backs   
   >> this up? It is an interesting thought.   
   > To my mind, this is one of those statements that can be either true or   
   > false, depending on how you choose to define "smart" and "functional".   
   [...]   
      
   Isn't that just part of the fun? :)   
      
   I was involved in a little  study a few years (mid 90s, I think) back relate to   
   game theory and agents.   
      
   In a typical scenario, different heuristics/agents "compete" for a reward   
   of some kind (e.g. increased amount of cpu time) in the service of a client.   
      
   In basic game theoric terms, this can result in an n-way Iterated Prisoners   
   Dilemma situation. Basicically, in an IPD the reward to an agent for   
   achieving a goal co-operatively is outweighed by the reward for   
   achieving the goal exclusively.   
      
   For intelligent agents (e.g. as modelled by a population of all   
   possible bounded algorithms) it's not unexpected that "lying" and   
   "stealing" become important strategies. In certain popualtions of agents,   
   "screwing up the other guy" tends to dominate after it becomes established   
   in a significant number of the pop'n.   
      
   When I delivered a paper on the subject to an Agent Group, there was   
   little comment.   
      
   [ 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)