lrnecgysiam.com
@decisionsTheory
we suggested that it would be helpful to view the AI task as that of designing rational agents—that is, agents whose actions maximize their expected utility given their percept histories. We showed that the design problem depends on the percepts and actions available to the agent, the utility function that the agent’s behavior should satisfy, and the nature of the environment. A variety of different agent designs are possible, ranging from reflex agents to fully deliberative, knowledge-based, decision-theoretic agents. Moreover, the components of these designs can have a number of different instantiations—for example, logical or probabilistic reasoning, and atomic, factored, or structured representations of states.
