Chapter1.pdf

EECS 492 (Satinder Singh) EECS 492 (AI) Satinder Singh Chapter 1 (R&N) EECS 492 (Satinder Singh) Outline ?Administrivia ?What is AI? ?A brief history ?The state of the art EECS 492 (Satinder Singh) Administrivia? ?(Lectures) 9-10:30 pm Tue-Thurs in 1109 FXB ?Discussions ?(Tuesday, 11:30-13:30) 104 EWRE [Yagil Engel] ?(Thursday, 12:30-1:30) 3427 EECS [Yagil Engel] ?(Book) Artificial Intelligence: A Modern Approach by Stuart Russell and Peter Norvig (2nd Edition) ?(Course Web Site) ?announcements, dynamic schedule, homework assignments, links,? EECS 492 (Satinder Singh) Administrivia? ?My Office: 176 ATL, 1101 Beal Ave. ?E-mail: baveja@umich.edu ?Please begin subject heading with EECS 492: ?Office Hours: Wednesday 4:00 to 5:30 pm or by appointment (setup by e-mail); not this week. ?Discussions ? examples, homeworks, questions, ? ?Attendance during discussions is important ?See website syllabus for details of grading, late policy, honor code, etc. ?6 Homeworks (60%); 1 Mid-term (20%), Final (20%) ?Bring ?policy? questions to discussion section after reading website. EECS 492 (Satinder Singh) Homeworks ?Homeworks will contain no real programming, though you may be asked to use some software as part of your assignment. ?First Homework will be out on Tuesday; each homework will have approximately 2 weeks before it is due ?Homeworks will be posted on ctools in pdf format. ?Homeworks will only be accepted on ctools ?Homeworks will be accepted late for up to 72 hours past due date and time with a 1% deduction for a delay of every hour or less; no exception except under the most extraordinary circumstances ?Homework points will be posted on ctools ?Regrade requests must be made by e-mail or in writing with an explanation of which question should be regraded and why. ?Regrade requests must be made within one week of the posting of the points on coursetools. EECS 492 (Satinder Singh) Exams ?Midterm: Date TBA (in class exam) ?20% of grade ?Location is in class ?Closed book (include all the material covered prior to the midterm) ?Final exam: Wednesday, Dec 21, 10:30 - 12:30 ?20% of grade ?Location: TBA ?Closed book (include all material since midterm) EECS 492 (Satinder Singh) How to tell your neighbor is intelligent? EECS 492 (Satinder Singh) What is AI? ?Playing chess ?Proving mathematical theorems ?Writing poetry ?Diagnosing diseases ?Interviewing a presidential candidate ?Flying a 747 ?Online search ?Robotic actors ?TV newsreaders EECS 492 (Satinder Singh) What is AI? Systems that act rationallySystems that act like humans Systems that think rationally Systems that think like humans human rational think act EECS 492 (Satinder Singh) Acting humanly: The Turing Test ?Alan Turing (1950) ?Computing machinery and intelligence? ?Operational definition of intelligence ?Several components of AI: knowledge, reasoning, learning, natural language understanding, etc. ?Predicted that by year 2000, a machine might have a 30% chance of fooling a lay person for 5 minutes. Problem: Turning test is not reproducible, constructive, or amenable to mathematical analysis EECS 492 (Satinder Singh) Thinking humanly: Cognitive Science ?1960s ?cognitive revolution?: information-processing psychology replaced prevailing behaviorism ?Requires scientific theories of internal activities and representations of the brain ?What level of abstraction? ?Knowledge? or ?circuits?? ?How to validate? Requires ?Predicting and testing behavior of human subjects (top-down) ?Direct identification from neurological data (bottom up) ?Both approaches (roughly, Cognitive Science and Cognitive Neuroscience) are now distinct from AI; though some areas, e.g., vision, benefit from integration ?All fail (at this point) to explain (or engender) anything resembling human-level general intelligence EECS 492 (Satinder Singh) Thinking rationally: Laws of Thought ?Aristotle: what are correct arguments/thought processes? irrefutable reasoning, right thinking? (Socrates is a man; all men are mortal; therefore, Socrates is mortal) ?Logicians derived notation and rules of derivation for thoughts ?Problems ?Not easy to represent informal, often uncertain, knowledge formally as required by logic ?Tractability issues even if we could do the above EECS 492 (Satinder Singh) Acting Rationally ?Rational behavior: doing the right thing ?The right thing: that which is expected to maximize goal achievement, given the available information ?Subsumes thinking rationally, when that serves the goal of acting rationally ?Does not necessarily involve thinking at all, e.g., blinking reflex. ?More amenable to scientific development than the acting humanly approach This is the point of view of the course! EECS 492 (Satinder Singh) Rational Agents ?An agent is an entity that perceives and acts ?This course is about designing rational agents ?Abstractly, an agent is a function from percept histories to actions ?For any given class of environments and tasks, we seek the agent (or class of agents) with the best performance ?Caveat: computational limitations make perfect rationality unachievable ?Design best program for given machine resources ? f:P * ?A EECS 492 (Satinder Singh) Course Overview ?Intelligent Agents ?Problem Solving ?Search; uninformed, informed, adversarial,? ?Knowledge and Reasoning ?propositional and first-order logic ?Planning ?Probablistic Reasoning and Decision Making ?Bayes Nets ?Learning ?Supervised, unsupervised, and reinforcement learning will get harder & more unfamiliar EECS 492 (Satinder Singh) AI prehistory Homeostatic systems, stability Simple optimal agent designs Control Theory Plastic physical substrate for mental activity Neuroscience Knowledge representation grammar Linguistics Formal theory of rational decision Economics Adaptation Phenomenon of perception and motor control Experimental techniques (psychophysics, etc.) Psychology Formal representation and proof Algorithms, computation (un)decidability, (in)tractability, probability, statistics Mathematics Logic, methods of reasoning Mind as physical system Foundations of learning, language, rationality Philosophy EECS 492 (Satinder Singh) AI history 1943McCulloch & Pitts: Boolean circuit model of brain 1950Turing?s ?Computing machinery and intelligence? 1952-69Look, Ma, no hands! 1950sEarly AI programs, including Samuel?s checkers program, Newell & Simon?s Logic Theorist, Gelernter?s Geometry Engine 1956 Dartmouth meeting: ?Artificial Intelligence? adopted 1965Robinson?s complete algorithm for logical reasoning 1966-74AI discovers computational complexity Neural network research almost dissapears 1969-79Early development of knowledge based systems 1980-88Expert systems industry booms 1988-93Expert systems industry busts: ?AI Winter? 1985-95Neural networks return to popularity 1988-Resurgence of probability; general increase in technical depth ?Novelle AI?: Alife, Gas 1995- Agents agents everywhere ?. EECS 492 (Satinder Singh) State of the art ?Which of the following can be done at present? ?Play a decent game of billiards/pool ?Drive along a curving mountain road ?Drive in the center of Cairo ?Buy a week?s worth of groceries at local supermarket ?Buy a week?s worth of groceries on the web ?Play a decent game of bridge ?Discover and prove a new mathematical theorem ?Write an intentionally funny story ?Give competent legal advice in a specialized are of law ?Translate spoken English into spoken Spanish in real time ?Perform a complex surgical operation EECS 492 (Satinder Singh) Read Chapter 2 (R&N) for next lecture.