140 likes | 237 Views
KR & R Knowledge Representation for Intelligent Agents. CIS548 November 8, 2006. KR for IA. Reaction alone insufficient Street crossing The feint :: war, sports Data, information, knowledge, wisdom D – I – K – W Components Abstraction Generalization Aggregation Individuation
E N D
KR & RKnowledge Representationfor Intelligent Agents CIS548 November 8, 2006
KR for IA • Reaction alone insufficient • Street crossing • The feint :: war, sports • Data, information, knowledge, wisdom • D – I – K – W Components • Abstraction • Generalization • Aggregation • Individuation • Structure • Maturation Kutztown University
Framing the Debate • Rodney Brooks • Intelligence Without Representation • Rep gets in way of simple level intel • Rep wrong unit of abstraction for intel • Build capability incrementally • Nothing succeeds like success • David Kirsh • Today the Earwig, Tomorrow Man • Insect ethologists ≠ cognitive scientists • Robotics requires theories of reasoning Kutztown University
Concept Using Creatures{Kirsh} • Capacity to predicate • Judgments of identity • New borns • Dogs • Chimps • Praying mantises • Role of thought in action Kutztown University
Intension/extension • Gottlob Frege • Venus = Venus {identity} • Morning star = evening star {fact} • Are these two statements equivalent? • Sense (Sinn) • Reference (Bedeutung) • Same extension (Venus) • Different intension (described object) Kutztown University
Action & Conceptualization{Kirsh, Frege} • Extension – arm movement • Intension – purpose • Wave hello • Wave goodbye • Swat flies • Differentiation – self-awareness of purpose Kutztown University
Conceptualization{Kirsh} • Ego-centric space • Agent at spatial-temporal origin • Agent-oriented concepts • In front of me; behind me; etc. • Public space • World viewed from nowhere • Generalized object-concepts Kutztown University
Knowledge-Rich Tasks{Kirsh} • Predict behavior of others • Project beyond sensory periphery • Gain objective perspective • Engage in problem solving • Counterfactual reasoning • When I get to Wyoming • Operate stimulus-free To make a prairie it takes a clover and one bee, One clover, and a bee, And revery. The revery alone will do, If bees are few. Kutztown University
Knowledge about Knowledge{McCarthy} • John says, “I don’t know Mike’s phone #.” • Jack says, “Yes, you do. Mike is Sam’s roommate.” • Jack says, “Yes, I do know it; it’s 555-1212.” • Did John know Mike’s phone number? • Define ‘know’ • Define “know that I know” • Define “know that I don’t know” • Define “don’t know whether I know” Kutztown University
Knowledge about Knowledge - II{McCarthy} • Two numbers, m & n, are chosen such that: 2 <= m <= n < = 99. • S is told their sum; P is told their product. • P: "I don't know the numbers." • S: "I knew you didn't know. I don't know either." • P: "Now I know the numbers." • S: "Now I know them too." • What are the numbers? Kutztown University
The Structure of Knowledge • The Case for case • Relations and graphs • Cohesiveness of concepts Kutztown University
The Case for Case • Jen gave Brianna a book. • Verb: give • Subject: Jen • Direct object: book • Indirect object: Brianna • Sentence can be parsed to create “knowledge graph” • Knowledge graph can be used to generate sentence Kutztown University
Relations • Express real world associations among discrete objects • Representable as graphs • Graphs yield variety of structures • Directed, undirected • Connected, not connected • Trees • Planar, non-planar • Graphs permit operations on structures • Traversal • Edge/node addition/deletion Kutztown University
To Come • Reasoning • Structural analysis • State Space Search (previous lectures) • Graph traversal • Node information retrieval • Expert Systems • Backward chaining • Forward chaining • Theorem Proving • Semantic tableaux • Resolution Kutztown University