1/29/18 Chapter Outline • Motivations to study AI… Artificial Intel l i gence • What is AI anyway? • A brief history of the field • The state of the art Ch a p t e r 1 (Some slides adapted from Stuart Russel, Dan Klein, and others. Thanks guys!) 1 2 So you wanna do AI? What is Artificial Intelligence? • Why are you interested in AI personally? Why are you • We could start by looking at (supposed) areas of here? application… • … and there are plenty… ü Because I think Artificial Insemination is key to our nation’s agricultural success? (fail) ü I need another class to graduate ü ??? Knowledge Representation Social Networks Deep Graph Analysis Q&A Analysis Audio/Speech systems Analytics IOT, Smart Simulation Visualization Virtual homes Physics and • Why you should be interested… Personal Modeling Assistants • Intelligence defines us , as human, as a species… Soft Natural Known Robotics Language Generation Image Machine • Opportunity… Analytics Learning Unknown • AI will be huge… Smart AI Vehicles • AI vs Physics as a field… Natural Deep Known Language Recommender Learning Processing Systems 1
1/29/18 What is Artificial Intelligence? What is Artificial Intelligence? • Or maybe see what sorts of companies are working on • Hmmm, or maybe we could boil this down to some key what… “topic areas”… Neural Networks Planning Robotics Artificial Machine Intelligence Knowledge Learning Representation Natural Computer Language Vision Processing Credit: http://www.techforkorea.com What is Artificial Intelligence? What is AI? • We’re still not at the heart of it… • At the core, it’s about software systems that… behave in a certain way. • WHAT IS INTELLIGENCE REALLY? • Historically, we can discern four different perspectives Similar functionally…but quite different philosophically • ü Uhh…well…ya know… thinking . Too vague! ü “performance” (fidelity) “competence” (fidelity) Systems that think like humans Systems that think rationally Reasoning Systems that act like humans Systems that act rationally Acting 8 2
1/29/18 Acting humanly: The Turing test Thinking humanly: Cognitive Science Turing (1950) “Computing machinery and intelligence”: • 1960s “cognitive revolution”: information-processing psychology • “Can machines think?” −→ “Can machines behave intelligently?” replaced prevailing orthodoxy of behaviorism Operational test for intelligent behavior: the Imitation Game • • Requires scientific theories of internal activities of the brain – What level of abstraction? “Knowledge” or “circuits”? HUMAN – How to validate? Requires either: HUMAN • Predicting and testing behavior of human subjects (top-down) ? INTERROGATOR • Direct identification from neurological data (bottom-up) AI SYSTEM … plus, ideally, modeling findings in software! • Both approaches (roughly, Cognitive Science and Cognitive Predicted that by 2000, a machine might have a 30% • Neuroscience) are now distinct from AI chance of fooling a lay person for 5 minutes • Anticipated all major arguments against AI in following 50 years • Both share with AI the following characteristic: • Suggested major components of AI: knowledge representation, the available theories do not explain (or engender) anything • reasoning, language understanding, learning resembling human-level general intelligence Hence, all three fields share one principal direction: focus on • Problem: Turing test is not reproducible, constructive, or understanding intelligent behavior amenable to mathematical analysis 9 10 Acting rationally: doing the “right” thing Thinking rationally: Laws of Thought • Rational behavior: doing "the right thing” • Normative (or prescriptive) rather than descriptive approach • Don’t worry about how humans perform it • Don’t worry about logical truth • Focus on results: • Aristotle: what are correct arguments/thought processes? The right thing: that which is expected to Several Greek schools developed various forms of logic: • maximize goal achievement, given the available notation and rules of derivation for thought • information. • Will see more when we look at reasoning agents. • may or may not have proceeded the idea of mechanization Doesn’t necessarily involve thinking • e.g., blinking reflex…but thinking should be in the • i.e. pure philosophy versus application orientation. • service of rational action • Direct line through mathematics and philosophy to modern AI This course in AI is about engineering • Problems: About how to build it, how to make it happen • • Not about philosophy…or even theory of cognition 1. Not all intelligent behavior is mediated by logical deliberation 2. Not goal driven. What is the purpose of thinking? What thoughts should I have • Thus: we will focus on this practical view of AI. out of all the thoughts (logical or otherwise) that I could have? • How can we make the machine intelligently solve problems? Formally: Designing agents that act rationally • 11 12 3
1/29/18 Rational agents AI prehistory: Influences Philosophy logic, methods of reasoning, mind as physical system • An agent is an entity that perceives and acts foundations of learning, language, rationality • This course is about designing rational agents • Defn: a software agent that acts to achieve best expected outcome modulo: Mathematics formal representation and proof, concept of algorithms, computation, (un)decidability, (in)tractability, probability • Available knowledge at that moment • Uncertainty of knowledge that it does have Psychology Adaptation, phenomena of perception and motor control • Or often, realistically: Limited rationality: take the most rational experimental techniques (psychophysics, etc.) action given some time limit to act. • Abstractly, an agent is a function from percept histories to actions: Economics formal theory of rational decisions (decision theory), Game theory, Max-Min strategies, Adversarial reasoning f : P ∗ → A Linguistics knowledge representation, grammar (for NLP) • For any given class of environments and tasks, we seek the agent (or class of agents) with the best performance Neuroscience Model of plastic physical substrate for mental activity Caveat: computational limitations make perfect rationality • unachievable Control Theory homeostatic systems, stability, simple optimal agent designs • → design best program for given machine/situational resources 13 14 Brief glance: history of AI Quick quiz: How much do you know about AI? 1943 McCulloch & Pitts: Boolean circuit model of brain 1950 Turing’s “Computing Machinery and Intelligence” Which of the following can be done at present? 1952-69 Look Ma, no hands! Early automatons… Play a decent game of table tennis • 1950s Early AI programs, including Samuel’s checkers program, Newell & • Drive safely along a curving mountain road Simon’s Logic Theorist, Gelernter’s Geometry Engine • Drive safely in rush downtown phoenix 1956 Dartmouth meeting: “Artificial Intelligence” adopted Buy a week’s worth of groceries on the web • 1965 Universal solver: Robinson’s complete algorithm for logical reasoning • Buy a week’s worth of groceries at Bashas 1966-74 AI discovers computational complexity • Play a decent game of bridge Neural network research almost disappears Beat world champions in GO • 1969-79 Early development of knowledge-based systems • Discover and prove a new mathematical theorem 1980-1988 Expert systems industry booms • Design and execute a research program in molecular biology 1988-93 Expert systems industry busts: “AI Winter” • Give competent legal advice in a specialized area of law 1985-95 Neural networks concepts resuscitated…a new way forward • Converse successfully with another person for an hour 1988- Resurgence of probability; general increase in technical depth Perform a complex surgical operation • “Nouvelle AI”: ALife, GAs, soft computing • Unload any dishwasher and put everything away 1995- Agents, agents, everywhere . . . • Write an intentionally funny story 2003- Human-level AI back on the agenda (the next bubble?) 15 16 4
Recommend
More recommend