From: Joscha Bach < Sent: Monday, January 21, 2013 1:58 PM To: Jeffrey Epstein Subject: Re: It might be a question of putting the parts together; =he minimal components would probably be the false belief task (the =epresentation that others may believe something that is false, which =as been demonstrated non-verbally in 15month olds) with basic verbal =bility and autonomous goal-seeking behavior. Each one has been done =lready, but that does not mean that the resulting system is generally =ntelligent. It merely learns how to deceive (ie necessary but not =ufficient►. Ron Arkin from Georgia Tech (who is mostly known =or discussing the ethics of military robots) has also built a bunch of =imple robots that compete for limited resources and develop deceptive =ehavior to gain an advantage. The =iterally-minded Turing Test people that crowd around the Loebner prize =or most human-like behavior are probably not in the game for deceptive =ehavior, though. Their systems are usually not intentional, i.e. they =ould fake their fakery, using cleverly pre-scripted dialog =trategies. I would probably start from the =ther direction, i.e. the premise that deceptive behavior is an emergent =uality of any system that has the explicit goal of changing the beliefs =f others to further its own aims. The idea that the induced beliefs are =actually correct would need to be added on top of that. As soon as a =ystem switches from signalling of internal states towards true =ommunication (the goal-directed creation of beliefs in others), we will =ave deceptive systems. Most Al research =pparently ignores this, because they start out with applications =hat benefit from total cooperation (all agents have compatible goals, =ike cars that should not bump into each other, robotic ants that =ransmit their respective states, soccer playing robots that broadcast =heir positions among team =ates). Cheers, Joscha</=iv>. I have seen =any proposals that would like to atttain