Foundations of Artificial Intelligence

1 Foundations of Artificial IntelligenceAIMA Chapter 1 (after Russell and Norvig)AIMA Chapter 1 (after Russell and Norvig) 1 Outline Administration What is AI? the understanding and building of intelligent entitites Foundations A brief history The state of the artAIMA Chapter 1 (after Russell and Norvig) 2 What is AI? [The automation of] activities that weassociate with human thinking, activitiessuch as decision-making, problem solving,learning.. (Bellman, 1978) The study of mental faculties throughthe use of computational models (Charniak+McDermott, 1985) The study of how to make computers dothings at which, at the moment, peopleare better (Rich+Knight, 1991) The branch of computer science thatis concerned with the automation of in-telligent behavior (Luger+Stubblefield,1993)Views of AI fall into four categories:Thinking humanlyThinking rationallyActing humanlyActing rationallyExamining these, we will plump for acting rationally (sort of)AIMA Chapter 1 (after Russell and Norvig) 3 Acting humanly: The Turing testTuring (1950) Computing machinery and Intelligence : Can machines think?

2 Can machines behave intelligently? Operational test for intelligent behavior: the Imitation Game Predicted that by 2000, a machine might have a 30% chance offooling a lay person for 5 minutes Anticipated all major arguments against AI in following 50 years Suggested major components of AI: knowledge, reasoning, languageunderstanding, learningProblem: Turing test is not reproducible, constructive, oramenable to mathematical analysisLoebner Prize: The first formal instantiation of a Turing test Chapter 1 (after Russell and Norvig) 4 Thinking humanly: Cognitive Science1960s cognitive revolution : information-processing psychology replacedprevailing orthodoxy of behaviorismRequires scientific theories of internal activities of the brain What level of abstraction?

3 Knowledge or circuits ? How to validate? Requires1) Predicting and testing behavior of human subjects (top-down)or 2) Direct identification from neurological data (bottom-up)Both approaches (roughly, Cognitive Science and CognitiveNeuroscience)are now distinct from AIBoth share with AI the following characteristic and thus direction:the avail-able theroeis do not explain (or engender) anything resembling human-level general intelligenceAIMA Chapter 1 (after Russell and Norvig) 5 Thinking rationally: Laws of ThoughtNormative(or prescriptive) rather than descriptiveAristotle: what are correct arguments/thought processes?Several Greek schools developed various forms of logic:notationand rules of derivationfor thoughts;may or may not have proceeded to the idea of mechanizationDirect line through mathematics and philosophy to modern AIProblems:1) Not all intelligent behavior is mediated by logical deliberation2) What is the purpose of thinking?

4 What thoughts should I have?3) Computationally intractable4) Expressive inadequaciesAIMA Chapter 1 (after Russell and Norvig) 6 Acting rationallyRationalbehavior: doing the right thingThe right thing: that which is expected to maximize goal achievement,given the available informationDoesn t necessarily involve thinking , blinking reflex butthinking should be in the service of rational actionAristotle ( nicomachean ethics ):Every art and every inquiry, and similarly every actionand pursuit, is thought to aim at some goodBeyond programs: autonomy, perception, persistence, adaptive, goal adop-tionAIMA Chapter 1 (after Russell and Norvig) 7 Rational agentsAn agentis an entity that perceives and actsAbstractly, an agent is a function from percept histories toactions.

5 F:P AFor any given class of environments and tasks, we seek theagent (or class of agents) with the best performanceCaveat: computational limitations make perfect rationality unachievable design best programfor given machine resourcesAIMA Chapter 1 (after Russell and Norvig) 8 Course: Rational agentsThis course is about designing rational agentsThinking rationally is only one way of acting rationallyRationality is more clearly defined and general than human-centered ap-proachesAIMA Chapter 1 (after Russell and Norvig) 9AI Prehistory and HistoryMany disciplines (philosophy, mathematics, economics, psychology, linguis-tics, computer engineering, control theory, neuroscience, and more) havecontributed ideas, viewpoints, and techniques to AIThe history of AI has had cycles of success, misplaced optimism, and resultingretrenchments.

6 Cycles of new creativity and systematic refinement of bestapproachesAIMA Chapter 1 (after Russell and Norvig) 10 Potted history of AI1943-1955: Gestation1956: Birth1952-1969: Great Expectations1966-1973: Reality1969-1979: Knowledge is Power1980-present: AI and Industry1986-present: The Return of Neural Networks1987-present: AI Becomes a Science1995-present: Intelligent AgentsAIMA Chapter 1 (after Russell and Norvig) 11AI becomes a science, Intelligent Agents1987 Rapid increase in technical depthBuild on existing theoriesBase claims on theorems and/or experments rather than intuitionReal world applications rather than toy examplesReplication of experiments with data and code repositoriesLess isolationism1995 Whole agents rather than fragmentsSituated movementInternet environments ( bots )AIMA Chapter 1 (after Russell and Norvig) 12 State of the artAutonomous planning and scheduling (NASA)Game playing (Deep Blue)Automomous control (minivan steering)Diagnosis (medicine)Logistics planning (Gulf war)Robotics (surgery)Language understanding and generation (translation, dialogue)

7 Problem solving (crossword puzzles)AIMA Chapter 1 (after Russell and Norvig) 13 SummaryDifferent people think of AI differently (thinking or behavior, humans orideal)We will adopt the Rational Action view: an Intelligent Agenttakes the bestpossible action in a situationAI has its roots in many disciplinesThe history of AI has had various cyclesCurrently: greater use of the scientific method, progress inboth theory andpractice, integration within subfields and across disciplinesAIMA Chapter 1 (after Russell and Norvig) 14