Newsgroups: comp.ai
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!nntp.sei.cmu.edu!cis.ohio-state.edu!math.ohio-state.edu!usc!chi-news.cic.net!nntp.coast.net!swidir.switch.ch!swsbe6.switch.ch!news.unige.ch!usenet
From: squire@cui.unige.ch (David SQUIRE)
Subject: Re: Emotion-Mediated Artificial Intelligence  -  Is it intelligent?
Message-ID: <1995Dec20.094600.22404@news.unige.ch>
Sender: usenet@news.unige.ch
Reply-To: squire@cui.unige.ch
Organization: University of Geneva, Switzerland
References: <bmarshal.3.00024606@agt.net>
Date: Wed, 20 Dec 1995 09:46:00 GMT
Lines: 41


In article <bmarshal.3.00024606@agt.net>, bmarshal@agt.net (Brian Marshall) writes:
>I have developed an AI technique to find solutions to "dirty hard"
>problems like navigating a car, packing a knapsack, picking a chess
>move or scheduling staff.   It is based on my theory that people and 
>other animals use a technique that is controlled by hope (the
>emotional response to the evaluation of how well something is
>going to turn out).
>
>The AI technique is described briefly below.  For much more...
>
>   Web-Site:  http://www.agt.net/public/bmarshal/homepage.htm
>
>I have implemented the technique in a program that schedules
>staff.  It sort of seems intelligent.  It uses the same basic
>approach that a mouse might use trying to find a way across
>a creek on stepping stones.  Is the mouse acting intelligently?
>I would like to suggest that a program that implements the
>technique IS intelligent.  Comments?
>
>The technique in a nutshell  (for one-object problems like 
>navigating a car)...
>
>   Repeatedly try to complete a solution, remembering the best
>   try.  Complete a solution by always picking the option with the
>   most hope.  
>
>   Hope is the arithmetic sum of:
>    -  static goodness (ex. how good the road is)
>    - "Grass may be greener on the other side of the fence" factor
>    - "Been problem" factor
>    - "Works well" factor
>
>    In each try, only the most hopeful options are picked.  A chaotic
>    interaction between simple factors (hopefully) results in many
>    different solutions being found.
>
>The staff-scheduling program does find good schedules.



