1 cs b551: elements of artificial intelligence instructor: kris hauser hauserk

38
1 CS B551: Elements CS B551: Elements of Artificial of Artificial Intelligence Intelligence Instructor: Kris Hauser Instructor: Kris Hauser http://cs.indiana.edu/~hauserk http://cs.indiana.edu/~hauserk

Upload: amice-fleming

Post on 28-Dec-2015

216 views

Category:

Documents


0 download

TRANSCRIPT

1

CS B551: Elements of CS B551: Elements of Artificial IntelligenceArtificial Intelligence

Instructor: Kris HauserInstructor: Kris Hauserhttp://cs.indiana.edu/~hauserkhttp://cs.indiana.edu/~hauserk

22

BasicsBasics

Class web siteClass web site• http://cs.indiana.edu/courses/b551http://cs.indiana.edu/courses/b551

TextbookTextbook• S. Russell and P. NorvigS. Russell and P. Norvig• Artificial Intelligence: a Modern Artificial Intelligence: a Modern

ApproachApproach• 22ndnd edition edition

33

BasicsBasics

InstructorInstructor• Kris Hauser (Kris Hauser ([email protected]@indiana.edu))

AIsAIs• Ik Hyun Park (Ik Hyun Park ([email protected]@indiana.edu))• Mark Wilson (Mark Wilson ([email protected]@indiana.edu))

44

Office HoursOffice Hours

Kris HauserKris Hauser• M,Th 1-2 in Lindley 301FM,Th 1-2 in Lindley 301F

Ik Hyun ParkIk Hyun Park• Th 1:30-3:30 in TBATh 1:30-3:30 in TBA

Mark WilsonMark Wilson• M 10-12 in Lindley 406M 10-12 in Lindley 406

55

AgendaAgenda

Intro to AIIntro to AI Overview of class policiesOverview of class policies

6

Intro to AIIntro to AI

77

What is AI?What is AI?

AI is the reproduction of AI is the reproduction of human human reasoning and intelligent reasoning and intelligent behaviorbehavior by computational methods by computational methods

88

What is AI?What is AI?

AI is AI is an attempt ofan attempt of reproduction of reproduction of human reasoning and intelligent human reasoning and intelligent behaviorbehavior by computational methods by computational methods

99

What is AI?What is AI?

Discipline that systematizes and Discipline that systematizes and automates reasoning processes to automates reasoning processes to create machines that:create machines that:

Think like humansThink like humans Think rationallyThink rationally

Act like humansAct like humans Act rationallyAct rationally

1010

The goal of AI is: to build machines that The goal of AI is: to build machines that operate in the same way that humans operate in the same way that humans thinkthink• How do humans think?How do humans think?• Build machines according to theory, test how Build machines according to theory, test how

behavior matches mind’s behaviorbehavior matches mind’s behavior• Cognitive ScienceCognitive Science

Manipulation of symbolic knowledgeManipulation of symbolic knowledge How does hardware affect reasoning? How does hardware affect reasoning?

Discrete machines, analog mindsDiscrete machines, analog minds

Think like humansThink like humans Think rationallyThink rationally

Act like humansAct like humans Act rationallyAct rationally

1111

The goal of AI is: to build machines that perform The goal of AI is: to build machines that perform tasks that seem to require intelligence when tasks that seem to require intelligence when performed by humansperformed by humans

Take a task at which people are better, e.g.:Take a task at which people are better, e.g.:• Prove a theoremProve a theorem• Play chessPlay chess• Plan a surgical operationPlan a surgical operation• Diagnose a diseaseDiagnose a disease• Navigate in a buildingNavigate in a building

and build a computer system that does it and build a computer system that does it automaticallyautomatically

But do we want to duplicate human But do we want to duplicate human imperfections?imperfections?

Think like humansThink like humans Think rationallyThink rationally

Act like humansAct like humans Act rationallyAct rationally

1212

The goal of AI is: to build machines that The goal of AI is: to build machines that make the “best” decisions given current make the “best” decisions given current knowledge and resourcesknowledge and resources

““Best” depending on some Best” depending on some utilityutility function function• Influences from economics, control theoryInfluences from economics, control theory

How do self-consciousness, hopes, fears, How do self-consciousness, hopes, fears, compulsions, etc. impact intelligence?compulsions, etc. impact intelligence?

Where do utilities come from?Where do utilities come from?

Think like humansThink like humans Think rationallyThink rationally

Act like humansAct like humans Act rationallyAct rationally

1313

What is Intelligence?What is Intelligence?

““If there were machines which bore a resemblance to If there were machines which bore a resemblance to our bodies and imitated our actions as closely as our bodies and imitated our actions as closely as possible for all practical purposes, we should still possible for all practical purposes, we should still have two very certain means of recognizing that they have two very certain means of recognizing that they were not real men. The first is that they could never were not real men. The first is that they could never use words, or put together signs, as we do in order to use words, or put together signs, as we do in order to declare our thoughts to others… Secondly, even declare our thoughts to others… Secondly, even though some machines might do some things as well though some machines might do some things as well as we do them, or perhaps even better, they would as we do them, or perhaps even better, they would inevitably fail in others, which would reveal that they inevitably fail in others, which would reveal that they are acting not from understanding, …”are acting not from understanding, …”Discourse on the Method, by Descartes (1598-1650)Discourse on the Method, by Descartes (1598-1650)

1414

What is Intelligence?What is Intelligence?

Turing Test (c. 1950)Turing Test (c. 1950)

1515

An Application of the Turing An Application of the Turing TestTest

CAPTCHA: Completely Automatic Public Turing CAPTCHA: Completely Automatic Public Turing tests to tell Computers and Humans Aparttests to tell Computers and Humans Apart

1616

Chinese Room (John Searle)Chinese Room (John Searle)

1717

Can Machines Act/Think Can Machines Act/Think Intelligently?Intelligently?

Yes, if intelligence is narrowly defined Yes, if intelligence is narrowly defined as information processingas information processingAI has made impressive achievements showing that AI has made impressive achievements showing that tasks initially assumed to require intelligence can be tasks initially assumed to require intelligence can be automatedautomated

Each success of AI seems to push further the limits Each success of AI seems to push further the limits of what we consider “intelligence”of what we consider “intelligence”

181818

Some AchievementsSome Achievements Computers have won over world Computers have won over world

champions in several games, champions in several games, including Checkers, Othello, and including Checkers, Othello, and Chess, but still do not do well in GoChess, but still do not do well in Go

AI techniques are used in many AI techniques are used in many systems: formal calculus, video systems: formal calculus, video games, route planning, logistics games, route planning, logistics planning, pharmaceutical drug planning, pharmaceutical drug design, medical diagnosis, hardware design, medical diagnosis, hardware and software trouble-shooting, and software trouble-shooting, speech speech recognition, traffic monitoring, recognition, traffic monitoring, facial recognition, facial recognition, medical image analysis, part medical image analysis, part inspection, etc... inspection, etc...

DARPA Grand Challenge: DARPA Grand Challenge: robotic car autonomously traversed robotic car autonomously traversed 132 miles of desert 132 miles of desert

Some industries (automobile, Some industries (automobile, electronics) are highly robotized, electronics) are highly robotized, while other robots perform brain while other robots perform brain and heart surgery, are rolling and heart surgery, are rolling on Mars, fly autonomously, …, on Mars, fly autonomously, …, but home robots still remain but home robots still remain a thing of the futurea thing of the future

1919

Can Machines Act/Think Can Machines Act/Think Intelligently?Intelligently?

Yes, if intelligence is narrowly defined as Yes, if intelligence is narrowly defined as information processinginformation processingAI has made impressive achievements showing that AI has made impressive achievements showing that tasks initially assumed to require intelligence can be tasks initially assumed to require intelligence can be automatedautomated

Maybe yes, maybe not, if intelligence Maybe yes, maybe not, if intelligence cannot be separated from consciousnesscannot be separated from consciousness Is the machine Is the machine experiencingexperiencing thought? thought? Strong vs. Weak AIStrong vs. Weak AI

2020

2121

Big Open QuestionsBig Open Questions Is intelligent behavior just information Is intelligent behavior just information

processing?processing?(Physical symbol system hypothesis)(Physical symbol system hypothesis)

If so, can the human brain solve problems If so, can the human brain solve problems that are inherently intractable for that are inherently intractable for computers? Will a general theory of computers? Will a general theory of intelligence emerge from neuroscience?intelligence emerge from neuroscience?

In a human being, where is the interface In a human being, where is the interface between “intelligence” and the rest of between “intelligence” and the rest of “human nature”“human nature”• Self-consciousness, emotions, compulsionsSelf-consciousness, emotions, compulsions

What is the role of the body?What is the role of the body?(Mind-body problem)(Mind-body problem)

2222

AI contributes to building an information AI contributes to building an information processing model of human beings, just as processing model of human beings, just as Biochemistry contributes to building a Biochemistry contributes to building a model of human beings based on bio-model of human beings based on bio-molecular interactionsmolecular interactions

Both try to explain how a human being Both try to explain how a human being operatesoperates

Both also explore ways to avoid human Both also explore ways to avoid human imperfections imperfections (in Biochemistry, by engineering (in Biochemistry, by engineering new proteins and drug molecules; in AI, by new proteins and drug molecules; in AI, by designing rational reasoning methods) designing rational reasoning methods)

Both try to produce new useful Both try to produce new useful technologiestechnologies

Neither explains (yet?) the true meaning Neither explains (yet?) the true meaning of being humanof being human

2323

Main Areas of AIMain Areas of AI Knowledge Knowledge

representation (including representation (including formal logic)formal logic)

Search, especially Search, especially heuristic search (puzzles, heuristic search (puzzles, games)games)

PlanningPlanning Reasoning under Reasoning under

uncertainty, including uncertainty, including probabilistic reasoningprobabilistic reasoning

LearningLearning Robotics and perceptionRobotics and perception Natural language Natural language

processingprocessing

Search

Knowledgerep.Planning

Reasoning

Learning

Agent

RoboticsPerception

Naturallanguage

... ExpertSystems

Constraintsatisfaction

2424

Bits of HistoryBits of History 1956:1956: The name “Artificial Intelligence” is coined The name “Artificial Intelligence” is coined 60’s: 60’s: Search and games, formal logic and Search and games, formal logic and

theorem proving theorem proving 70’s:70’s: Robotics, perception, knowledge Robotics, perception, knowledge

representation, expert systemsrepresentation, expert systems 80’s:80’s: More expert systems, AI becomes an More expert systems, AI becomes an

industryindustry 90’s:90’s: Rational agents, probabilistic reasoning, Rational agents, probabilistic reasoning,

machine learningmachine learning 00’s:00’s: Systems integrating many AI methods, Systems integrating many AI methods,

machine learning, reasoning under uncertainty, machine learning, reasoning under uncertainty, robotics againrobotics again

2525

SyllabusSyllabus Introduction to AIIntroduction to AI

• Philosophy, history, agent frameworksPhilosophy, history, agent frameworks SearchSearch

• Uninformed search, heuristic search, heuristicsUninformed search, heuristic search, heuristics Search applications (and variants)Search applications (and variants)

• Constraint satisfaction, planning, game playing, motion Constraint satisfaction, planning, game playing, motion planningplanning

Reasoning under uncertaintyReasoning under uncertainty• Probability, planning under uncertainty, Bayesian Probability, planning under uncertainty, Bayesian

networks, probabilistic inference, dynamic modelingnetworks, probabilistic inference, dynamic modeling Intro to machine learningIntro to machine learning

• Neural nets, decision tree learning, support vector Neural nets, decision tree learning, support vector machines, etc.machines, etc.

2626

B335

Robotics

???

B552

B553

Knowledge representation and learning

B657

Computer Vision

Biologically-inspired computing

B659

I486

Q360B651

Natural Language Processing

E626

Game theory

Q570

Topics in AI

B551

S626 S675

2727

Careers in AICareers in AI

‘‘Pure’ AIPure’ AI• Academic, some labsAcademic, some labs

Applied AIApplied AI• Almost any area of CS!Almost any area of CS!• NLP, vision, roboticsNLP, vision, robotics• EconomicsEconomics

Cognitive ScienceCognitive Science

2828

AI ReferencesAI References

ConferencesConferences• IJCAI, ECAI, AAAI, NIPSIJCAI, ECAI, AAAI, NIPS

JournalsJournals• AI, Comp. I, IEEE Trans. Pattern Anal. AI, Comp. I, IEEE Trans. Pattern Anal.

Mach. Intel., IEEE Int. Sys., Journal of AIRMach. Intel., IEEE Int. Sys., Journal of AIR SocietiesSocieties

• AAAI, SIGART, AISBAAAI, SIGART, AISB AI Magazine (David Leake)AI Magazine (David Leake)

29

Class PoliciesClass Policies

3030

GradingGrading

60% Homework60% Homework• Lowest score will be droppedLowest score will be dropped

30% Final30% Final 10% Participation10% Participation

3131

Programming AssignmentsProgramming Assignments

Projects will be written in Projects will be written in PythonPython Great for scriptingGreat for scripting

• Peter Norvig, Director of Research at Peter Norvig, Director of Research at Google, and textbook authorGoogle, and textbook author

Easy to learnEasy to learn 2 weeks for each assignment2 weeks for each assignment

3232

Homework PolicyHomework Policy

Due at end of class on due dateDue at end of class on due date• Typically TuesdaysTypically Tuesdays

Extensions only granted in rare casesExtensions only granted in rare cases• Require advance notice except Require advance notice except

emergenciesemergencies

3333

Final ProjectFinal Project

Encouraged if you are intending to do Encouraged if you are intending to do research or coursework in AI, pursue research or coursework in AI, pursue higher degreehigher degree• Individual or small groups (up to 3)Individual or small groups (up to 3)• Counts for 20% of homework gradeCounts for 20% of homework grade

ContentContent• Software, new research, or technical reportSoftware, new research, or technical report• Mid-semester project proposalMid-semester project proposal• End-of-year report and in-class presentationEnd-of-year report and in-class presentation

3434

EnrollmentEnrollment

Add/drop deadlineAdd/drop deadline• No penalty: Sept 4No penalty: Sept 4• Late drop/add: Oct 28Late drop/add: Oct 28

Waitlist deadline: Sept 5Waitlist deadline: Sept 5

3535

Swine FluSwine Flu

3636

TakeawaysTakeaways

AI has many interpretationsAI has many interpretations• Act vs. think, human-like vs. rationalAct vs. think, human-like vs. rational• Concept has evolvedConcept has evolved

‘‘I’ has many interpretationsI’ has many interpretations• Turing testTuring test• Chinese roomChinese room

AI success stories from each AI success stories from each perspectiveperspective

3737

HomeworkHomework

RegisterRegister TextbookTextbook SurveySurvey http://cs.indiana.edu/classes/b551http://cs.indiana.edu/classes/b551 Readings: R&N Ch. 1, 2, 26Readings: R&N Ch. 1, 2, 26

3838

What is Intelligence?What is Intelligence?

Total Turing TestTotal Turing Test• Physical interactionPhysical interaction