The Knowledge Level Paper written by Allen Newell published in - PowerPoint PPT Presentation

The Knowledge Level � Paper written by Allen Newell published in IT3706 Knowledge Representation Journal of Artificial Intelligence in 1982 � Concerns the question of what knowledge Lesson 6: The Knowledge Level is in AI. � Introduces a high-level view for characterizing intelligent behavior in agents. Agents: A Functional Architecture Two views on AI � An agent is embedded Affect Recognize The Components The Tasks in a task environment World input � A task statement enters via the perceptual � Perceptual system component Internal representation � Puzzle solving � The task statement is � Memory component � Natural language processing encoded in an internal � Reasoning system representation Change � Planning Apply Rep. � The method (to use to method � Acting system � Diagnosis accomplish the task) is � ... identified � ... Select � The method uses method general world knowledge to General Method accomplish the task. knowledge store

So, what’s the problem? Covering the Chessboard Task is to cover � Questionnaire given to Knowledge chessboard with Representation researchers show that there are dominos covering very little consensus . two squares. With normal � Representation has been given a ”magical” role board, this is easy. in intelligence. What if the top left What if the top left � Seen as the locus of square is and bottom right real intelligence (Turing machine) squares are removed? � The dominos-on-chessboard removed? example (next slide). Looking at ”Knowledge” and Computer System Levels ”Representation” � A computer system can � Representation: Quite specific meaning – almost be analyzed and a technical term in AI. understood on different Symbol Level � Knowledge: No agreement of what knowledge levels. � Each level is is. Register Transfer Level implemented in the level � Newell wants to investigate: below. Logic Circuit Level � What is the nature of knowledge? � A higher level is Circuit Level autonomous � How does it relate to representation? � Example: A programmer � What does a system have when it has knowledge? Device Level does not need to know how logic circuits work)

Computer System Levels Computer System Levels � A level is characterized by: Medium � A medium (that which is processed) Symbols and expressions (programs) Symbol Level Symbol Level � Components (provides the primitive processing) Register Transfer Level Bit vectors Register Transfer Level � Laws of composition (how components may be Logic Circuit Level Single bits assembled to a system) Logic Circuit Level � Laws of behavior (determines how the system behaves Circuit Level Current and voltage Circuit Level given how the components behave and how they are Device Level Electrons and magnetic domains Device Level composed) Computer System Levels The Knowledge Level Hypothesis ”There exists a distinct Symbol Level Register Transfer Level computer systems level, lying Knowledge Level immediately above the symbol level, which is characterized by Symbol Level System Computers Digital Systems knowledge as the medium and Medium Symbols (programs) Bit Vectors the principle of rationality as the Register Transfer Level behavior law. ” (p. 99) Components Memories, Operations Registers, Functional units Logic Circuit Level Composition Designations, Associations Transfer Path Principle of Rationality: Laws Circuit Level ”If an agent has knowledge that Behavior Sequential Interpretations Logical Operators one of its actions will lead to Device Level Laws one of its goals, then the agent will select that action.” (p. 102)

The Knowledge Level The Knowledge Level Agent Observer � The Knowledge Level provides the means to ’rationalize’ the behavior of a system from the standpoint of an external observer. Knowledge � ”He is running because he knows the store will close at 9.” � This observer treats the system as a ’black box’ but maintains that it acts as if it possesses certain knowledge about the world and uses this knowledge in a perfectly rational way toward reaching its goals. Goals � ”He will be here, this is the last showing and he really wanted to watch the movie.” � We can specify competence on a high level by specifying goals and Principle of knowledge before going into implementation details. Rationality � ”The agent should be able to play chess well, so it needs to know some good openings and what pieces positions on the board are important.” ”He will be able to solve the problem – he knows how and wants to do it.” The Knowledge Level Defined Knowledge Level and Symbol Level � System: Agent � The Knowledge Level can be ”reduced” i.e. Implemented in the Symbol Level. � Components: Goals, Actions, Bodies � Medium: Knowledge � Knowledge: What is stored in the Knowledge Representation system (frame-based system, rules- � Behavioral law: The principle of rationality based systems, logic ...) and all possible inferred � Compositional law: None – no structure! facts � Goal: Special kind of knowledge, or implicit. ”To treat a system at the KL is to treat it as having some knowledge and � Agent: The complete Symbol Level system some goals, and believing it will do whatever is within its power to attain its goals, in so far as its knowledge indicates” (p. 98) � Principle of Rationality: The Reasoning Strategy

Consequences Knowledge Level and Symbol Level � Knowledge is intimately linked to rationality – ’rational � The crucial difference between KL and SL is that systems can be said to have knowledge’ the Symbol Level is system oriented, whereas � Knowledge is a competence-like notion, having the the Knowledge Level is world oriented. ability to generate action � The knowledge level is an approximation � Knowledge serves as a specification of what a symbol � A Knowledge Level model is in terms of structure should be able to do knowledge about the world and the environment � Representations exist at the symbol level and realizes/mechanizes a body of knowledge. in which the agent's problems and actions occur. � Components (actions, goals, bodies of knowledge) of KL and its medium (knowledge) can be defined in terms of systems at the symbol level. Summary � A knowledge level model is a model of behavior in terms of knowledge, just like a symbol level model is a model of interaction in terms of symbols and representations. � Representation = Knowledge + Access � Emphasis on knowledge instead of representation and implementation allow allows one to make meaningful statements about system behavior without reference to the structures and mechanisms within the agent that realize that behavior.