XPost: comp.ai.edu   
   From: kym@ukato.freeshell.org   
      
   In comp.ai anees.haider@gmail.com wrote:   
   > Hi,   
   > I am final year, computer science student of under graduation program.   
   > In this year i have to make a computer science project.   
   > I have decided to do my MS in Artificial Intelligence.   
   > So, I am interested in implementing a Multiagent system having   
   > Intelligent Agents. I dont want to create business application instead   
   > i want to solve some real world problem espacially Natural Science   
   > problem using Artificial Intelligence.   
      
      
   "Solving a probnlem using Artificial Intelligence" is on the face of it   
   a self-contradiction. If a problem has no known algorithm that can solve it   
   exactly, we might restort to "atrificial intelligence techniques"   
   which is another way of saying "using heurtistics or guessing"   
   to find an approximation to the solution, or maybe only find a method   
   that is "better than nothing".   
      
   To try to do such a thing as a (say) 4th year project is biting off much   
   more than the typical undergrad can chew. It is usually enough to   
   demonstrate some aspect of multi-agent systems, or create a much-simplified   
   version of a published work that re-creates some interesting (or maybe   
   even boring :) aspect of that published work.   
      
   As an illustration of the types of things I have in mind, consider   
   the following.   
      
   There is a muti-agent system that is used to solve a certain class of   
   problems. Each agent can be presented with a "query", and may or may not   
   provide an "answer". In solving the query, the agent may or may not   
   sub-divide the query into parts and consult other agents, re-combining   
   the results from those agents into a result for its original query.   
      
   There is a special agent -- let us call it "God" -- which can   
   provide feedback on the accuracy of the answers to its questions.   
   Maybe this agent is especially clever, or maybe it can test out   
   the answers in some kind of "real world". Whatever, it can "reward"   
   agents that provide good answers. Rewards might include (let us imagine)   
   extra "system resourecs", like more cpu time, greater disk space, etc.   
      
   The question now is -- given you can simulate such a simple setup --   
   are there any problems that you "see" (and I mean "observe" because   
   this may be a surprise and can't be foretold just by considering the   
   foregoing descriptions) that develops with such a agent system?   
      
   (Assuming the obvious) how can such a problem be overcome, or   
   maybe just reduced?   
      
   Finally, how does your results relate to "intellient systems"?   
      
   > Please provide me some project ideas in this regards but i have very   
   > little time left to submit project proposal, last date is 14th   
   > September '06, i will be very thankfull to you.   
   > Sincerely,   
   > Anees Haider   
      
   [ 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)