Abstract

knowledge [9] describes the nature of knowledge as the medium of a system level that lies above the symbol or program level in the hierarchy of levels of computational systems. The unique character of the knowledge level is that its central law of behavior, the principle that the system will take whatever actions attain its goals given its knowledge, can be computed without positing any internal structure of the system. Put another way, the knowledge level abstracts completely from representation, structure, and process. That this is possible--that a system's behavior can be predicted based only on the content of its representations plus its knowledge of its goals--is the essence of being an intelligent system. The knowledge level finds its most common use in designing systems, where the designer specifies a component by positing its function (i.e., its goal) and the knowledge available to it. Systems can be described at many levels. A system described at the knowledge level can also be described at the symbol level, which is to say in terms of representations, data structures, and processes. The symbol-level description shows how the knowledge-level behavior is attained, just as the register-transfer description shows how the symbol-level system is realized by architectural mechanisms, and so on down to descriptions in terms of physical concepts. The paper claims these concepts of knowledge and knowledge-level system are used in practice by the AI and computer science community, hence understood implicitly. The paper itself is just making explicit to the community what it already knows in practice.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call