Transcript CSE 471/598 Introduction to AI

CSE 471/598
Introduction to
Artificial Intelligence
http://www.public.asu.edu/~huanliu/AI09F/cse471-598.htm
Fall 2009
Introduction
You: a future AI Expert
TA: Ali Abbasi (mabbasi2 at asu.edu)
Time and Place: Please see our course
web page
Me: Huan Liu, huan.liu at asu.edu
(http://www.public.asu.edu/~huanliu)


My office hours (can be changed upon req)
Slides are updated periodically
CSE 471/598 H. Liu
2
Course Introduction
What is AI (many definitions of AI)


One definition: a field to enable computers
with human-level intelligence with attempts
to understand intelligent entities.
We will evaluate many definitions later.
What is this course about (or why should
everyone learn AI?)



understand ourselves better
build automated intelligent agents to
advance research
improve problem solving skills
CSE 471/598 H. Liu
3
Course workload and evaluation

We will work hard together - “No pain, no
gain!”
 Projects (30%, 2-3) – all in Lisp
 Exam(s) (2*25%)
 Homework (~20%)
 Quizzes and class participation (~10% extra)



Which grading system you prefer (w/wo +/-)
Late penalty, YES and exponentially increased
Academic integrity
(http://www.public.asu.edu/~huanliu/conduct.html)
CSE 471/598 H. Liu
4
Course plan
Text Book: AI - A Modern Approach

2nd Edition in green (3rd Edition will be out Nov. 09)
Reading assignment: chapters covered
About 13-15 chapters
Our lofty goal:
“to finish all the 27 chapters”
To be realistic,
Try to keep up and
avoid catch up
one major subject per week
TIP
CSE 471/598 H. Liu
5
Major Topics
Intelligent agents
Problem solving
Knowledge and reasoning
Acting logically
Learning
Uncertainty
CSE 471/598 H. Liu
TIP
Comprehend the topics
with your common sense
6
Welcome to this class!
We will work together and hard throughout
this semester and your active participation is
crucial for the success of the class – the REAL
shortcut to your success

Who don’t want shortcut? Apparent vs. true
shortcuts
Questions and suggestions are always
welcome.

E.g., if you find anything interesting to share,
incorrect or unclear, send an email or talk to me, or
discuss it in class
You get feedback from us (TA and me), and I
expect feedback from you, too 
Use myASU to send email and for discussions
CSE 471/598 H. Liu
7
Introduction of AI
- Gearing up for a fun semester
about intelligent agents
- What is an intelligent agent in your
view?
What is AI
About thinking and acting

We are not alone, but … (Homo (genus))
http://en.wikipedia.org/wiki/Homo_(genus)
Acting humanly: The Turing test (by Turing 1950)



Its original purpose
What do we need to pass the test?
http://www.loebner.net/Prizef/loebner-prize.html
Does that serve our purpose of developing AI?
Thinking humanly: Cognitive modeling


“Think-aloud” to learn from human and recreate in computer
programs (GPS)
What the Eyes see, a camera cannot
http://www.topcharoen.co.th/web/illusion/illusion-a19.gif
CSE 471/598 H. Liu
9
What is AI (2)
Thinking rationally: Syllogisms, Logic



What would you act on the $50 iBooks incident?
Unable to deal with uncertainty
Some paradoxes: Liar, Barber
 Gödel's incompleteness and Turing's undecidability
Acting rationally: A rational agent (something that acts)
to achieve best or best expected outcomes

Some rational actions do not involve inference
 An example – a reflex doe not need inference
A set of definitions (Figure 1.1)
CSE 471/598 H. Liu
10
Foundations of AI
Philosophy (428 B.C. - Present) –
reasoning and learning




Can formal rules be used to draw valid
conclusions?
How does the mental mind arise from a physical
brain?
Where does knowledge come from?
How does knowledge lead to action?
CSE 471/598 H. Liu
11
Mathematics (c. 800 - Present) - logic, probability,
decision making, computation



What are the formal rules to draw conclusions?
What can be computed?
How do we reason with uncertain information?
Economics (1776-present)



How should we make decisions so as to maximize
payoff?
How should we do this when others may not go along?
How should we do this when the payoff may be far in
the future?
CSE 471/598 H. Liu
12
Neuroscience (1861-present)

How do brains process information
 Processing speed, memory size in a computer
(Figure 1.3)
Psychology (1879 - Present) investigating human mind

How do humans and animals think and act?
 Mind Wide Open (the use of fMRI)
Computer engineering (1940 - Present) ever improving tools

How can we build an efficient computer?
 Moors Law, Raptures for the Geeks
CSE 471/598 H. Liu
13
Control theory and Cybernetics (1948present)


How can artifacts operate under their own
control?
Feedback and adapt
Linguistics (1957 - Present) - the structure
and meaning of language


How does language relate to thought?
Computational linguistics
CSE 471/598 H. Liu
14
Brief History of AI
Gestation of AI (1943 -1955)



McCulloch and Pitts’s model of artificial neurons
Minsky’s 40-neuron network
Alan Turing’s Computing Machinary and Intelligence
Birth of AI (1956)



A 2-month Dartmouth workshop of 10 attendees – the
name of AI
Newell and Simon’s Logic Theorist
Should another name like `computational rationality’ be
used? Any suggestion?
Early enthusiasm, great expectations (1952 1969)

GPS by Newell and Simon, Lisp by McCarthy, Blockworld
by Minsky
CSE 471/598 H. Liu
15
AI facing reality (1966 - 1973)

Many predictions of AI’s coming successes
 A computer would be a chess champion in 10 years (1957)



Machine translation – Syntax is not enough
Intractability of the problems attempted by AI
“What computers cannot do” in 76
Knowledge-based systems (1969 - 1979)


Knowledge is power, acquiring knowledge from experts
Expert systems (MYCIN)
AI - an industry (1980 - present)

Many AI systems help companies to save money and
increase productivity (Cyc)
CSE 471/598 H. Liu
16
The return of neural networks (1986 – present)


PDP books by Rumelhart and McClelland
Connectionist models vs. symbolic models
AI – a science (1987 – present)



Build on existing theories vs. propose brand new ones
Rigorous empirical experiments
Learn from data – machine learning, data mining
AI – intelligent agents (1995 – present)

Working agents embedded in real environments
with continuous sensory inputs
CSE 471/598 H. Liu
17
Some examples of AI
applications
Smart bombs
Deep Blue, and others
E-Game industry
E-Business
Intelligent houses
Intelligent appliances
RoboCup
Mars rovers
Biometrics
Communications (email,
word processor, social
media)
Auto driving from E to
W (98% vs. 2%)
Consumer protection
Social Networking Sites
CSE 471/598 H. Liu
18
Concluding remarks
“The real value of the discipline, Mr. Lazowska
said, is less in acquiring a skill with technology
tools - the usual definition of computer literacy than in teaching students to manage complexity;
to navigate and assess information; to master
modeling and abstraction; and to think
analytically in terms of algorithms, or step-bystep procedures.”
from
http://www.nytimes.com/2005/08/23/technology/23geeks.html
What is AI about?
CSE 471/598 H. Liu
19
Refresher for LISP
What is it?

ANSI Common Lisp, Paul Graham, Prentice
Hall
Input (e.g., terminal, files)
Output (e.g., files, printing)
Processing (various operations)
How to run it?
CSE 471/598 H. Liu
20