Why have AI researchers not been able to give computers human-like 'common sense', the ability to think about ordinary things the way people can? In our view, the source of the difficulty is that they too often seek after types of cognitive architectures, kinds of representations, and methods of inference that are based on some single, simple process, theory, or principle. Despite their elegance, no single one of such techniques can capture the diversity of mechanisms needed to reason about the broad range of common sense domains for example, those that require reasoning about temporal, spatial, physical, psychological, social, and self-reflective matters. Here we describe aspects of an architecture that we are developing to support the construction of AI systems resourceful enough to combine the advantages of many different ways to think about things, by making use of many types of mechanisms for reasoning, representation, and reflection.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below