Given that you would also be simulating a lot of unintelligent behaviour, how would you discover whether or not the simulation has been successful? I mean how would you find which parts of your system are the intelligent parts?
(Yes, I did just recently watch 2001. Sue me ;p)