Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Knowledge Representation: Rules Knowledge Representation • formal logic • rules • concepts • analogies • images • connections Rule-based systems: example How to get from Kalamazoo to Chicago? IF you want to get to Chicago, and you are in Kalamazoo, and you have no car, THEN take the train. IF you want to take the tarin from Kalamazoo to Chicago, THEN go to the train station and buy a ticket. IF you want to buy a ticket, THEN get some money. IF you want to get some money, then go to the bank and withdraw it. IF you want to get to Chicago, and you have a car, THEN take highway I94 than I90. IF: conditions ; THEN: action What rules do you use to plan weekend entertainment? Rule-based systems: some historical remarks Newell and Simon, General problem Solver: GPS 1950s-60s well-defined problems: math. problems, chess ill-defined problems: how to get a job, how to bring economical decisions etc... GPS GPS needs well-defined problem • Representation of state space • Set of operators (actions) • Table of differences/operators Search is guided by Means-end analysis subgoals to reduce particular differences between current state and goal state EXAMPLE Tower of Hanoi http://www.cut-the-knot.org/recurrence/hanoi.shtml GPS Newel and Simon’s main research effort was aimed at extending the boundaries of artificial intelligence, with particular concern for the simulation of human thought processes, moving from the well-structured tasks addressed in early A.l. programs to wider ranges of tasks that call on substantial bodies of knowledge and that are relatively loosely structured. Among the important early programs were the General Problem Solver (GPS) The initial idea was to represent problems of some general class as problems of transforming one expression into another by means of a set of allowed rules. GPS was the first system to separate the problem solving structure of goals and subgoals from the particular domain. If GPS had worked out to be really general, perhaps the Newell and Simon predictions about rapid success for AI would have been realized. GPS: Limitations • lot of the work done in the original framing • part of the intelligence is recognizing the problem and relevant symbols • humans operate in very many domains • the general problem solver isn’t very general Production Systems • Cognitive skills are realized by production rules • Production rules are organized around a set of goals • Complex cognitive processes involve a sequence of production rules • production rules are an elaborate knowledge base • Rules are psychologically realistic, because they describe many aspects of skilled behavior, and predict the details of that behavior Production Systems and the ACT ”Architecture of Cognition” (John R. Anderson, CMU) The goal is to understand how people organize knowledge that they acquire from their diverse experiences to produce intelligent behavior. The concern is very much with how it is all put together and this has led to the focus on what are called ”unified theories of cognition.” A unified theory is a cognitive architecture that can perform in detail a full range of cognitive tasks. The theory is called ACT-R and takes the form of a computer simulation which is capable of performing and learning from the same tasks that subjects. ACT-R is also an instance of a hybrid cognitive architecture in that it represents knowledge symbolically as rules and facts but also has a neurally-based activation process that determines which facts and rules get deployed in which situations. http://act-r.psy.cmu.edu/people/ja/ja-interests SOAR Newell and his students, 1980s, John E. Laird (Univ. Michigan) ”Unified theory of cognition” Soar is a general cognitive architecture for developing systems that exhibit intelligent behavior. Researchers all over the world, both from the fields of artificial intelligence and cognitive science, are using Soar for a variety of tasks. It has been in use since 1983, evolving through many different versions to where it is now Soar, Version 8.5. John Laird http://sitemaker.umich.edu/soar ELIZA ELIZA emulates a Rogerian psychotherapist. ELIZA has almost no intelligence whatsoever, only tricks like string substitution and canned responses based on keywords. Yet when the original ELIZA first appeared in the 60’s, some people actually mistook her for human. The illusion of intelligence works best, however, if you limit your conversation to talking about yourself and your life. http://www.manifestation.com/neurotoys/eliza.php3 How different rule-based represenation from logic? • Note less informational power than logic: rule-based system may not have full quantifiers and rules of inference. • But it can nevertheless be more computationally powerful, just because it focuses on the task to be accomplished • Uses processes that are not inherently part of logic: subgoaling. Strengths of rule-based systems • Have been used in many commercial technical systems • Modular, easy to add to • Have modeled various kinds of psychological experiments • Lots of human knowledge and ability naturally described in terms of rules Disadvantages of rules • Natural language is much more flexible than formal logic: not easy to formalize • Restricted to verbal information • Much reasoning is nonmonotonic: you can’t just add more beliefs deductively, but must subtract as well • Potentially computationally explosive • Most interesting kinds of reasoning are non-deductive From Logic to Rules and Mental Models Taxonomy of thought A taxonomy of thought Goal? Yes No Daydream Deterministic ? No Yes Precise goal No Calculation Yes Creation Increase in semantic information ? Yes Induction No Deduction Language Basic questions: 1. What representations are required for our ability to understand and produce language? 2. How is language learned? Behaviorist answer: Language is based on a set of associations, learned by trial and error. Chomsky’s revolution Language Syntactic Structures 1957 Rejected associationism • grammars are complex, rule-like structures • universal grammar is innate • we are born with readiness to notice what kind of grammar our native language has Steven Pinker, ”Rules of Language”