Introduction and Overview
What is Artificial Intelligence (AI)?
A field that focuses on developing techniques to enable computer systems to perform activities that are considered intelligent. The objective is to understand and build intelligent entities.
Associated questions
1) understand what constitutes intelligent behavior
2) build ``better'' computer systems
AI is one of the newest disciplines; it was
initiated in 1956. It is a dynamic and open field.
The study on intelligence is one of the oldest
disciplines; for over 2000 years philosophers
have tried to understand and define intelligence.
What kinds of behaviors count as intelligent?
- Everyday activities such as: recognizing
friends, having a conversation, planning and
cooking a dinner, interpreting a photograph,
walking to school.
-Formal tasks such as: proving theorems in
logic, playing chess or checkers.
-Expert tasks such as: engineering a design,
diagnosing an illness, doing financial analysis.
Which is harder and why?
It depends on the information that is accessible
and how amenable it is to formalization.
Expert skills are often easier than everyday
knowledge to formalize.
How to build a system?
Intelligent actions are typically complex, must
be able to decide what to leave out and why.
System design problems where there is no single
right answer. Part of the challenge of AI is
defining appropriate tasks.
History has shown that until the problem is
approached in the appropriate manner, limited
progress is made.
Agent Design
An agent perceives an environment through
sensors and acts on the environment through
effectors.
Must be able to model an agent doing something
for some purpose, i.e., taking actions toward
some goal.
The objective is to design an agent that performs
well, but what does that mean?
Design specifications are a particular challenge
for AI because of the analogy with human
behavior.
The approach taken by the text: systems that act
rationally---acting so as to achieve one's goals
Human-like | Rational | |
Think | (A) think like humans | (C) think rationally |
Act | (B) act like humans | (D) act rationally |
"Think" dimension: reasoning;
"Act" dimension: behavior
Columns contrast the human with the ideal
(A) equates to cognitive modeling
(B) equates to behaviorism
(C) equates to making correct inferences; there are problems with computational complexity
(D) the approach taken in this course
Acting rationally means acting so as to achieve
one's goals, given one's beliefs; the assessment
takes knowledge into account.
Correct inference is only sometimes a part of
being a rational agent.
More amenable to standard evaluation than is
comparison to human behavior.
Concentrate on the general principles of rational agents and the components of constructing them:
Goals of this course
OVERVIEW OF SYLLABUS
Why study AI?
It is a very exciting time because subfields have
matured to the extent that there are real
applications in commercial use. 1993 Commerce
Department survey estimated $900 million in
software sales.
Examples:
Check out the following web sites:
AI is both system and theory
AI is exciting because its potential for
contribution is tremendous and the basic
methodology is still under development.
AI draws on & contributes to many other fields.
Foundations of AI
Disciplines: philosophy, mathematics,
psychology, computer engineering and linguistics
Do our philosophical inclinations shape our
work?
The history, present and future of AI
History: Birth, high expectation, disillusionment, and commercial success
Present and Future: borrowing from other disciplines, taking scale seriously, hard experimental evidence