1、11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,1,What is AI?,Chapter 1Quotes & ConceptsGoals & ApproachesApplication AreasFramework for AI SystemsFundamental Issues for AI Problems,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,2,What is AI?,“AI is the d
2、esign, study and construction of computer programs that behave intelligently.”Tom Dean,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,3,What is AI?,What is artificial intelligence (AI)?is a field concerned intelligent behavior in - machines humans animalsother attempts at nami
3、ng this field included machine intelligence complex information processing heuristic programming cognology,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,4,What is AI?,“AI is the study of complex information processing problems that often have their roots in some aspect of bio
4、logical information processing. The goal of the subject is to identify solvable and interesting information processing problems, and solve them.”David Marr (1945-1980),11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,5,What is AI?,“It is the science and engineering of making int
5、elligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable.”“Intelligence is the computational part of the ability to achieve go
6、als in the world.”,see more: http:/www-formal.stanford.edu/jmc/whatisai/whatisai.html,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,6,What is AI?,What is intelligent behavior? perceiving ones environment acting in complex environments learning and understanding from experienc
7、e using reasoning to solve problems and to discover “hidden” knowledge applying knowledge successfully in new situations thinking abstractly, using analogies communicating with others what about - creativity? ingenuity? expressiveness? curiosity?,11/20/2018,2001-2004 James D. Skrentny from notes by
8、C. Dyer, et. al.,7,What is AI?,“I believe that understanding intelligence involves understanding how knowledge is acquired, represented, and stored; how intelligent behavior is generated and learned; how motives, and emotions, and priorities are developed and used;,how sensory signals are transforme
9、d into symbols; how symbols are manipulated to perform logic, to reason about the past, and plan for the future; and how the mechanisms of intelligence produce the phenomena of illusion, belief, hope, fear, and dreams-and yes even kindness and love.”see more: http:/tommy.jsc.nasa.gov/er/er6/mrl/pape
10、rs/symposium/albus.txt,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,8,Goals of and Approaches to AI,What are goals of AI? to replicate human intelligence (still distant goal) to solve knowledge-intensive tasks to make an intelligent connection between perception and action t
11、o enhance human-human, human-computer and computer-computer interaction/communication,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,9,Goals of and Approaches to AI,Engineering Goal develop concepts, theory and practice of building intelligent machines emphasis on system build
12、ingScience Goal develop concepts, mechanisms and vocabulary to understand biological intelligent behavior emphasis on understanding intelligent behavior,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,10,Goals of and Approaches to AI,Choose the goals of the computational model
13、the basis for evaluating performance of the system,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,11,Goals of and Approaches to AI,Box 1: Cognitive Science Approach focus on behavior and I/O model reasoning processes computational model should reflect how results were obtained
14、Goal not just to produce human like behavior (box 3) but to produce a reasoning process similar to the steps used by humans,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,12,Goals of and Approaches to AI,Box 2: Laws of Thought focus on inference mechanisms that are provably co
15、rrect and guarantee an optimal solutionGoal formalize the reasoning process as a system of logical rules and procedures for inferencing develop systems of representation to allow inferences to be performed,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,13,Goals of and Approach
16、es to AI,Box 3: Behaviorist Approach focus on action in the world, intelligent behavior not centered around representation of the worldGoal develop systems that are human-likeTuring Test Alan Turing 1950 can a machine fool an interrogator into believing that it is a human?,11/20/2018,2001-2004 James
17、 D. Skrentny from notes by C. Dyer, et. al.,14,Goals of and Approaches to AI,Box 4: Satisficing Methods focus on systems that act sufficiently if not optimally in all situations its OK to have imperfect reasoning if the job gets doneGoal develop systems that are rational and sufficient,11/20/2018,20
18、01-2004 James D. Skrentny from notes by C. Dyer, et. al.,15,Goals of and Approaches to AI,There are two kinds of people in the AI world:classical AI, the symbol-processing approach top down, knowledge to symbol to implementation uses logical operations and declarative knowledge bases Newell and Simo
19、ns physical symbol system sub-symbolic AI, the animat approach bottom up, from signal to symbol design proceeds in steps of the evolutionary ladder intelligence is an emergent property without requiring centralized models,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,16,Some
20、Application Areas of AI,Game Playing Speech Recognition Computer Vision Expert Systems Heuristic Classification Natural Language Understanding Scheduling and Planning,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,17,Some Application Areas of AI,Game Playing $100 for a machine
21、 that significantly uses AI playstation, game cube, etc. Deep Blue beat world champion Gary Kasparov by looking at hundreds of millions positions per secondSpeech Recognition PEGASUS allows users to obtain flight info and make reservations by speaking over the phone significant advances in speech re
22、cognition (and processor speed) now make systems possible,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,18,Some Application Areas of AI,Computer Vision face recognition programs: banks, casinos, police CMUs ALVINN autonomously drove a van from Washington, D.C. to San Diego, a
23、veraging 63 mph day and night, in all weather conditions handwriting recognition, electronics and manufacturing inspection, photo-interpretation, baggage inspection, reverse engineering to automatically construct a 3D geometric model advances have been made, but humans are far better at understandin
24、g and representing 3D information,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,19,Some Application Areas of AI,Expert Systems “knowledge engineer” tries to embody the knowledge of human experts in a computer program that is used to carry out some task diagnostic systems (MYC
25、IN 1974) system configuration (DEC) financial decision makingHeuristic Classification categorize examples based on guidelines credit card purchases (Visa),11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,20,Some Application Areas of AI,Natural Language Understanding web page tra
26、nslators (AltaVista) Catepillars truck manuals translated into 20 languages by machine getting a sequence of words and parsing sentences isnt enough, understanding requires knowledge of the domain, so present systems are quite limited,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et.
27、 al.,21,Some Application Areas of AI,Scheduling and Planning automatic scheduling for manufacturing American Airlines rerouting contingency planner European Space Agency planning and scheduling of spacecraft assembly, integration and verification,11/20/2018,2001-2004 James D. Skrentny from notes by
28、C. Dyer, et. al.,22,Some AI “Grand Challenge“ Problems,Intelligent Agents Smart Clothes Aids for the Disabled Tutors Accident-Avoiding Vehicles Self-Organizing Systems Translating Telephone Conversations,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,23,A Framework for Buildin
29、g AI Systems,Perception Reasoning Action,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,24,A Framework for Building AI Systems,Perception biological systems experience the world through their sensesWhat perceptions would be needed by: autonomous vehicle? camera images and rang
30、e data medical diagnosis system? symptoms and test results,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,25,A Framework for Building AI Systems,Includes areas of vision speech processing natural language processing signal processing, eg: market data acoustic data,11/20/2018,2001-2004 Ja
31、mes D. Skrentny from notes by C. Dyer, et. al.,26,A Framework for Building AI Systems,Reasoning includes inferencing, decision-making, classifying from what is sensed and state of internal “model“ of the world AI systems use: heuristic search in a problem space logical deduction system neural networ
32、ks genetic algorithms hidden Markov model induction Bayes network inferencing,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,27,A Framework for Building AI Systems,Includes areas of: knowledge representation problem solving decision theory planning game theory machine learning
33、 uncertainty reasoning,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,28,A Framework for Building AI Systems,Action biological systems interact with the world through their movements, sounds, other behaviors What actions are needed by: autonomous vehicle? steering and speed co
34、ntrol, sensor positioning, medical diagnosis system? make prescriptions, suggest further tests, ,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,29,A Framework for Building AI Systems,Includes areas of: robot actuation natural language generation speech synthesis computer graph
35、ics sound synthesis,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,30,Fundamental Issues for AI Problems,Representation Search Inference Learning Planning,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,31,Fundamental Issues for AI Problems,Representation
36、 facts about the world are remembered How do we represent facts? What should we store? How do we structure this knowledge? What is explicit? What is inferred? How are inference rules encoded? How should inconsistent, incomplete, and probabilistic knowledge be dealt with?,11/20/2018,2001-2004 James D
37、. Skrentny from notes by C. Dyer, et. al.,32,Fundamental Issues for AI Problems,Representation example: “The fly buzzed irritatingly on the window pane. Jill quickly picked up a newspaper.“What is the inference? Jill is going to start a fire? Jill is going to start a papiermache project? Jill is goi
38、ng to exterminate the fly?,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,33,Fundamental Issues for AI Problems,Representation example: “Given 12 sticks in a 2 by 2 grid, move 3 to leave exactly 3 boxes.“,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,34
39、,Fundamental Issues for AI Problems,Search a problem space is searched for a solution Checkers: 1040 states Chess: 10120 states How do limit the search space? How do we find an optimal solution? How are heuristics and constraints used?,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et
40、. al.,35,Fundamental Issues for AI Problems,Inference new facts are determined from a set of existing facts deduction abduction non-monotonic reasoning reasoning under uncertaintyexample: All elephants have trunks. Clyde is an elephant. Does Clyde have a trunk? Willy has a trunk. Is Willy an elephan
41、t?,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,36,Fundamental Issues for AI Problems,Learning new knowledge is acquired inductive inference neural networks genetic algorithms artificial life evolutionary approaches,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dye
42、r, et. al.,37,Fundamental Issues for AI Problems,Planning a strategy for achieving a goal in terms of a sequence of primitive actions is generated What general facts about the world are needed? What facts about the specific situation are needed? What facts are needed about the effects of actions? Ho
43、w do you state the goal? How do you know the goal has been reached?,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,38,Can machines think?,AI has always been surrounded by controversy:Can machines think? Depends on how you define thingsCan? Can someday? Or can today? Can in pri
44、nciple? Or can in practice?,11/20/2018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,39,Can machines think?,Machine? computers have expanded notion of what a machine can be, no longer just gears and grinding, clanking and hissing simple organisms are starting to be explained as very com
45、plex machines (existence proof!) maybe machines made of proteins can do things that machines made of silicon cant Searle: thinking requires a special machine Newell and Simon: type of machine is irrelevant so long as it can process symbols complexity requires massively parallel architecture?,11/20/2
46、018,2001-2004 James D. Skrentny from notes by C. Dyer, et. al.,40,Can machines think?,Think? difficult to define, what abilities are required before we label a machine intelligent? maybe a test (Turing test) should be used Weizenbaums ELIZA fooled many Maudlins JULIA is even better intelligence may best be viewed as a spectrum of abilities of increasing complexity,
