CS 561: Artificial Intelligence
Instructor: Prof. Prof Hadi Moradi, Moradi
[email protected] Lectures: M-Th 09:00-10:40, OHE136 Office hours: MW 2:30 – 4:00 pm, SAL310, Or by O b appointment i TAs: Jeong-Yoon Lee
SAL 112 Office hours: TTH 1:00-2:30 Email:
[email protected]
CS 561: Artificial Intelligence
Course web page:
http://www-scf.usc.edu/~csci561a Up to date information, lecture notes Relevant dates, links, etc. Also you may check http://den.usc.edu
Class format: two sections of 45 minutes Course material:
[AIMA] Artificial Intelligence: A Modern Approach, by Stuart Russell and Peter Norvig. 2nd edition
1
CS 561: Artificial Intelligence
Course overview: foundations of symbolic intelligent systems. Agents, search, problem solving, logic, representation, reasoning, symbolic programming, probabilistic reasoning, and robotics. Prerequisites: CS 455x, i.e.,
programming principles, discrete mathematics for computing, software design and software engineering concepts. Some knowledge of C/C++ for some programming assignments.
CS 561: Artificial Intelligence
Grading: 25% for midterm 25% for final 40% for homeworks and projects 10% for f Quizzes Q i
2
Practical issues
Class list: use learn.usc.edu learn usc edu
Login with your USC username and password If CSCI561A is not listed as your courses, notify ot y the t e TA.
Submissions: See class web page under Assignments submit -user csci561 -tag HW3 HW3.tar.gz
Administrative Issues
Midterm 1: 7/26/10 9:00 - 10:40pm
Midterm 2: 8/10/10 9:00 - 10:40pm See also the class web page: http://den usc edu/ http://den.usc.edu/
3
Why study AI?
Search engines Science Medicine/ Diagnosis Labor Appliances
What else?
Humanoid Robots: From Honda to Sony
Walk
Turn http://world.honda.com/robot/
Stairs
4
Sony AIBO
movie1
http://www.aibo.com
Natural Language Question Answering
http://aimovie.warnerbros.com
http://www.ai.mit.edu/projects/infolab/
5
Robot Teams
USC robotics Lab
Modular robots self re-assembly.
What is AI? The exciting new effort to make “The The study of mental faculties computers thinks … machine through the use of computational with minds, in the full and models” (Charniak et al. 1985) literal sense” (Haugeland 1985) “The art of creating machines that perform functions that require intelligence when performed by people” (Kurzweil, 1990)
A field of study that seeks to explain and emulate intelligent behavior in terms of computational processes” (Schalkol, 1990)
6
AI – The Bigger Picture ?
Computer Science
Philosophy p y
Artificial Intelligence Cognitive Science (Psychology)
Robotics (Engineering)
Neuroscience (Biology)
?
Acting Humanly: The Turing Test
Alan Turing Turing'ss 1950 article Computing Machinery and Intelligence discussed conditions for considering a machine to be intelligent
7
Acting Humanly: The Turing Test
What tasks require AI?
“AI AI is the science and engineering of making intelligent machines which can perform tasks that require intelligence when performed by humans …”
What tasks require AI?
8
What tasks require AI?
Tasks that require q AI:
Solving a differential equation Brain surgery Inventing stuff Playing Jeopardy Playing Wheel of Fortune What about walking? What about grabbing stuff? What about pulling your hand away from fire? What about watching TV? What about day dreaming?
Acting Humanly: The Full Turing Test
• Problem:
9
What would a computer need to pass the Turing test?
Communication: Memory: Reasoning: Learning:
What would a computer need to pass the Turing test?
Sensing:
M t control Motor t l (total (t t l ttest): t)
10
Thinking Humanly: Cognitive Science
1960 “Cognitive Cognitive Revolution Revolution”:: information-processing psychology replaced behaviorism
Thinking Humanly: Cognitive Science
Cognitive science and modeling the activities of the brain
What level of abstraction? “Knowledge” or “Circuits”? How to validate models?
11
Thinking Rationally: Laws of Thought
Aristotle (~ ( 450 B.C.) attempted to codify “right thinking”
What are correct arguments/thought processes?
Thinking Rationally: Laws of Thought
Problems:
12
Acting Rationally: The Rational Agent
Rational behavior: Doing the right thing! Provides the most general view of AI because it includes:
Acting Rationally: The Rational Agent
Advantages:
13
How to achieve AI?
How is AI research done?
Theoretical Experimental
How to achieve AI?
There are two main lines of research:
Biological, study humans and imitate their psychology or physiology. phenomenal, study and formalize common sense facts about the world and the problems that the world presents to the achievement of goals.
The two approaches interact to some extent, and both should eventually succeed. It is a race, but both racers seem to be walking. [John McCarthy]
14
Branches of AI
Logical AI Search Natural language processing pattern recognition Knowledge representation Inference From some facts,, others can be inferred. Automated reasoning Learning from experience Planning To generate a strategy for achieving some goal
AI Prehistory
15
Brief History of AI Thinking Rattionally:Laws of T Thought
Ancient Times
M iddle Age
384 B.C.
1200
- Aristotle - Logic: The science of knowing. Ramon Lull Ars Magnus: a rule-based device to model man's behavior and nature - Empiricism Explanation of processes - Gottfried Leibniz - 1st system of formal logic -
Renaissance 17 th Century
18 th Century
19 th Century
Next time implement links
Rene Descartes Dualism 1845
- Charles Babbage - Analytical Engine - George Boole - Formalization of the Laws of Logic -
1879-1903
Early 20th Century
1910-1912
-
Gottlob Frege First-order predicate calculus Russel-Whitehead Principia Mathematica Bertrand Russel
1931
- Kurt Godel - Incompleteness Theorem of Logic -
Roots of AI in Science:
Aristotle(b.384-): syllogism – formal reasoning Ramon Lull (b.1235): Ars Magna – a machine capable of answering all questions Rene Descartes (1596): mind / body separation (dualism); "cogito ergo sum“ Wilhelm Liebniz (1646-1716): a mechanical concept generator;; "materialism" g Charles Babbage(1792-1871), Ada Lovelace (1815-1860): Analytical Engine – a general-purpose calculator George Boole(1815-1864): logic algebras - logical encoding and calculation of thoughts Gottlob Frege(1848-1925): predicate calculus
16
Birth of Artificial Intelligence 1940-1956 1942
1943
1945
Greaat Expectations
1949
-
ENIAC :First digital computer
Mc Culloch and Pitts Artificial neural network J. Von Neumman Modern computer architecture
Claude Shannon Use of heuristics to solve complex problems
1950
- Alan M.Turing - Computing Machinery and - Intelligence: Turing Test
1955
- Herbert Simon,Alan Newell - 1st AI program:Logic Theorist -
1956
- Dartmouth Conference -
Herbert Simon
The Beginning of AI
McCulloch & Pitts developed theory of artificial neurons (precursor to ANN's) – 1943 Alan Turing – "Can Machines Think?" the turing test (1950) the turing machine Marvin Minsky & Dean Edmonds first ANN constructed, 1951 John McCarthy convened the Dartmouth conference that coined the term artificial intelligence (AI) (1956) and set the research agenda symbolic AI connectionism st AI language LISP (list processing) 1958 1
17
The Rise of AI 1957- 1960’s 1958
1960
1961
Growingg Disenchantment
1962
1965
- John McCarthy. - LISP -
Marvin Minsky Theory of Frames Herbert Simon,Alan Newell GPS:General Problem Solver Herbert Simon
Frank Rosenblatt Perceptron: Learning in Neural Networks
- L Lotfi tfi A. A Zadeh Zd h Fuzyy Logic Fuzzy Sets -
1968
Joseph Weizenbaum ELIZA: simulates diagnosis by a psychiatrist.
1969
- Marvin Minsky,Seymour Papert - Limitations of Perceptrons S. Papert
An Optimistic Start
In the 50's, 60's and early 70's, much exciting progress was being made in AI: Chess
The Logic Theorist
Feigenbaum, Buchanan, Lederberg, 1969
SHRDLU – NLP (Blocks World)
Joseph Weizenbaum, 1966
DENDRAL – Knowledge-Based System
Arthur Samuels, 1959
Eliza - NLP
Alan Newell, Cliff Shaw, Herb Simon, 1957
Checkers (Machine Learning)
Claude Shannon, 1950
Terry Winnograd, 1972
GPS (General Problem Solver)
Alan Newell & Herb Simon, 1972
18
The 70’s Bi th and Birth d Rise Ri of fE Expertt S Systems t 1970-mid 1980’s 1973
1974
-
1975
Alain Colmerauer PROLOG Paul Werbos Neural Networks Back Propagation Law E. Feigenbaum, R. Lindsay. Dendral E.FeigenBaum
Edward Shortliffe MYCIN 19761980
R. Duda, P.Hart, P. Barnett PROSPECTOR: The first commercial Expert System
1982
John McDermott XCON – "Expert Configurer
P.Hart
The Plateau
In the 70's, AI researchers began to discover that the problem wasn't as easy as it looked!
The Frame Problem
L k of Lack f Common C Sense S Reasoning R i
Combinatorial Explosion
The Gap – "Toy" vs. "Real" worlds
Perceptrons, by Minsky & Papert (1969) – proved limitations of perceptron networks and acted to limit significant research in the 70 70'ss Lighthill Report – 1973: curtailed research funding in British Universities AI developed a reputation as "over-hyped" and unrealistic
19
1982
Rebirth of Arttificial Neural Netw works
Commercializa ation of Expert Systtems
The 80’s
- John Hopfield - Hopfield Networks -
1982
1986
Teuvo Kohonen self-organising feature maps for speech recognitizion
T Sejnowski S j ki - Terrence - NETTalk Rumerhalt,McMelland
Neural Networks Rediscovering of Back-Propagation Learning 1987
- Marvin Minsky - The Society of Minds -
Fuzzy Appliances
1989
- Dean Pomerleau - ALVINN -
Commercial Success Despite it's it s reputation as "over over-hyped hyped", certain AI applications became very successful during the 70's – 80's: •
Expert Systems
•
Industrial Robotics
•
Planning & Scheduling Applications
AI became a $2,000,000,000 industry by 1988
20
Nowadays… - Major advances in all areas of AI, with - significant demonstrations -
Early 90’s
Late 90’s
1995
Birth of Intelligent Systems
1997
The Deep Blue chess program beats Garry Kasparov
- Web crawlers - AI-based information extraction - programs Intelligent Room and Emotional Agents at MIT's AI Lab
2000-
Interactive robot pets The Nomad robot
The Gartner Hype Curve
Interest in AI followed this pattern pattern,
typical of the hype surrounding new technologies
21
AI State of the art
Have the following been achieved by AI?
World-class chess playing Playing table tennis Cross-country driving Solving mathematical problems Discover and prove mathematical theories Engage in a meaningful conversation Understand spoken language Observe and understand human emotions …
Types of expertise
(with examples)
Deep cognitive skills
Judgmental High-level skills social skills
Highly creative
Musician
Senior manager
Analytical
Mathemati i ician
Economist, Social programmer scientist i ti t
Typist Strictly procedural
Driver
Author, poet
Social worker
22
A driving example: Grand Challenge
Goal:
Artificial Intelligence Applications Artificial Intelligence
Cognitive Science Applications •Expert Systems •Fuzzy Logic •Genetic Algorithms •Neural Networks
Robotics Applications
•Visual Perceptions •Locomotion •Navigation •Tactility
Natural Interface Applications •Natural Language •Speech Recognition •Multisensory Interface •Virtual Reality
23
AI Application Areas in Business Neural Networks Fuzzy Logic Systems Genetic Algorithms Virtual Reality y AI Application Areas in Business
Intelligent Agents Expert Systems
Components of Expert Systems The Expert System Expert Advice
User
User IInterface t f Programs
Inference E i Engine Program
Knowledge K l d Base
Workstation
Expert System Development Knowledge Engineering
Knowledge Acquisition Program Workstation
Expert and/or Knowledge Engineer
24
Expert System Applications Decision Management Diagnostic/Troubleshooting
Maintenance/Scheduling
Design/Configuration
Major Application Categories of Expert Systems
Selection/Classification
Process Monitoring/Control
Course Overview General Introduction
Introduction. [AIMA Ch 1] Course Schedule. Homeworks, exams and grading. Course material, TAs and office hours. Why study AI? What is AI? The Turing test. Rationality. Branches of AI. Research disciplines connected to and at the foundation of AI. Brief history of AI. Challenges for the future. Overview of class syllabus.
25
Agent
effectors
sensors
Course Overview
General Introduction
Intelligent Agents. [AIMA Ch 2] What is
an intelligent agent? Examples. Doing the right thing (rational action). Performance measure. Autonomy. Environment and agent design. Structure of agents agents. Agent types types. Reflex agents agents. Reactive agents. Reflex agents with state. Goal-based agents. Utility-based agents. Mobile agents. Information agents.
Course Overview (cont.)
Problem solving and search. [AIMA Ch 3]
measuring problem. Types of problems. More examples. Basic idea behind search algorithms. Complexity. Combinatorial explosion and NP completeness. Polynomial hierarchy.
3l
5l
9l
Using these 3 buckets, measure 7 liters of water.
Traveling salesperson problem
26
Course Overview (cont.) How can we solve complex problems?
Uninformed search. [AIMA Ch 3]
Depth-first. Breadth-first. Uniform-cost. Depth-limited. Iterative deepening. Examples. Properties.
3l
5l
9l
Using these 3 buckets, measure 7 liters of water.
Traveling salesperson problem
Course Overview (cont.) How can we solve complex p p problems?
Informed search. [AIMA Ch 4]
Best-first. A* search. Heuristics. Hill climbing. Problem of local extrema. Simulated annealing.
Traveling salesperson problem
27
Course Overview (cont.) Practical applications of search
Constraint Satisfaction
[AIMA Ch 5]
Backtracking g Local search
Course Overview (cont.) Practical applications of search
Game playing
[AIMA Ch 6]
The minimax algorithm. g Resource limitations. Aplha-beta pruning. Elements of chance and non-deterministic games.
tic-tac-toe
28
Course Overview (cont.) Towards intelligent agents
Agents that reason logically 1
[AIMA Ch 7]
Knowledge-based agents. Logic and representation. Propositional (boolean) logic.
wumpus world
Course Overview (cont.) Towards intelligent agents
Agents that reason logically 2.
[AIMA Ch 7]
Inference in propositional ii l logic. l i Syntax. Semantics. Examples.
wumpus world
29
Course Overview (cont.) Building u d g knowledge-based o edge ased agents: 1st Order Logic
First-order logic 1. [AIMA Ch 8]
Syntax. Semantics. Atomic sentences. sentences Complex sentences. Quantifiers. FOL knowledge base. Situation calculus.
Course Overview (cont.) Building knowledge knowledgebased agents: 1st Order Logic
First-order logic 2.
[AIMA Ch 9]
Describing actions. Planning. Action sequences.
30
Course Overview (cont.) Reasoning Logically
Inference in first-order logic.
[AIMA Ch 9]
Proofs. Unification Unification. Generalized modus ponens. Forward and backward chaining. Example of backward chaining
Course Overview (cont.) Representing and Organizing Knowledge
Building a knowledge base.
[AIMA Ch 10]
Knowledge bases. Vocabulary and rules. Ontologies Organizing knowledge. An ontology for the sports domain
31
Course Overview (cont.) Systems y that can Plan Future Behavior
Planning.[AIMA Ch 11]
Definition and goals. Basic representations for planning. l i Situation space and plan space. Examples.
Course Overview (cont.) Learning g from Observation
Decision Trees [AIMA 18]
Introduction to decision trees. Information theory. Constructing DT. Examples.
32
Course Overview (cont.) Expert p Systems y
Probabilities + Bayesian Networks [AIMA 13 + 14]
Basics of probability theory Bayesian rule. Conditional d l reasoning. Bayesian Networks. Reasoning under uncertainty
Course Overview (cont.) Statistical Learning g Methods
Neural Networks. [AIMA 20]
Human brain structure Neuron and activation function. Forward and backward propagations. Examples.
33
Course Overview (cont.) Logical g Reasoning g in the Presence of Uncertainty
Fuzzy logic [Handout]
Center of gravity
Introduction to fuzzyy logic. g Linguistic Hedges. Fuzzy inference. Examples.
Center of largest area
Course Overview (cont.) Machine Learning g
Genetic Algorithms [Handout + AIMA 4]
Genetic algorithm approach. Mutation, Crossover, Fitness function. Examples.
34
Course Overview (cont.) What challenges g remain?
Towards intelligent machines. [AIMA Ch 25]
The challenge of robots:
with what we have learned, what hard problems remain to be solved? Different types of robots. Tasks that robots are for. Parts of robots. Architectures. Configuration spaces.
robotics@USC
Course Overview (cont.) What challenges remain?
Overview and summary. [all of the above]
What have we learned. learned Where do we go from here?
robotics@USC
35
Outlook
AI is a very exciting area right now. now This course will teach you the foundations.
36