Layered architectures for intelligent agents

Discusses the implications of a single architecture principle for the design of autonomous intelligent agents. The principle is named the 'power vs generality tradeoff principle'. The structuring principle of layering is discussed as a potential solution to the power vs generality tradeoff problem. Layering applies the strategy of separation of concerns to power and generality by allocating separate resources to each. The author considers strategies for the enhancement of power and generality, along with an illustration of the layering principle in the effective area of agent functionality.