Transcript Data Modeling - Hiram College

Introduction (Chapter 1)
CPSC 386 Artificial Intelligence
Ellen Walker
Hiram College
Goals of this Course
• Become familiar with AI techniques, including
implementation
– Be able to read and write AI programs in LISP,
and to a lesser extent, Prolog and CLIPS
• Understand the theory behind the techniques,
knowing which techniques to apply when
(and why)
• Become familiar with a range of applications
of AI, including “classic” and current systems.
What is AI?
• Not just studying intelligent systems, but
building them…
• Psychological approach: an intelligent
system is a model of human intelligence
• Engineering approach: an intelligent system
solves a sufficiently difficult problem in a
generalizable way
Four Categories Of AI Definitions
Thinking Humanly
“The exciting new effort to make
computers think… machines with
minds” (Haugeland, 1985)
Thinking Rationally
“The study of mental facilities
through the use of computational
models” (Charniak & McDermott,
1985)
Acting Humanly
“creating machines that perform
functions that require intelligence
when performed by people”
(Kurzweil, 1990)
Acting Rationally
“AI … is concerned with intelligent
behavior in artifacts (Nilsson, 1998)
(Turing test)
Turing Test
• Given a communication terminal, can an
observer determine whether the entity at the
other end is human or machine?
– Tests “acting like a human”
– Does not test “thinking like a human”
– Does not test “rational” acting or thinking
Foundations of AI (Sec. 1.2)
• Philosophy
– Rationality
– Mind vs. brain
– Knowledge and goals
• Mathematics
– Algorithms for reasoning (with uncertainty)
– Computability theory
• Economics
– Decision theory
– Game theory
More Foundations…
• Neuroscience
– Studying brains
• Psychology
– Studying behavior
– Cognitive modeling
• Computer science and engineering
– An “artifact” to make intelligent
• Control Theory & Cybernetics
• Linguistics
Eras of AI (sec. 1.3)
• Gestation (1943-1955)
– Early learning theory, first neural network, Turing test
• Birth (1956)
– Name coined by McCarthy
– Workshop at Dartmouth
• Early enthusiasm, great expectations (1952-1969)
– GPS, physical symbol system hypothesis
– Geometry Theorem Prover (Gelertner), Checkers (Samuels)
– Lisp (McCarthy), Theorem Proving (McCarthy), Microworlds
(Minsky et. al.)
– “neat” (McCarthy @ Stanford) vs. “scruffy” (Minsky @ MIT)
More Eras of AI
• Dose of Reality (1966-1973)
– Combinatorial explosion
• Knowledge-based systems (1969-1979)
– Weak methods vs. domain-specific knowledge
• AI Becomes an Industry (1980-present)
– Boom period 1980-88, then AI Winter
• Return of Neural Networks (1986-present)
• AI Adopts the Scientific Method (1987-present)
• Intelligent Agents (1995-present)
– SOAR, Internet as a domain
• Very Large Data Sets (2001-Present)
What Makes a Solution AI?
• Not just the problem, also the generality of
the solution
• Examples
– Tic Tac Toe
– Question Answering
– Speech understanding
Tic Tac Toe #1
• Precompiled
move table.
• For each input
board, a specific
move (output
board)
• Perfect play, but
is it AI?
move
table
encode
look
up
Tic Tac Toe #2
• Represent board as a magic square, one integer per
square (834, 159, 672)
• If 3 of my pieces sum to 15, I win
• Predefined strategy:
–
–
–
–
–
1. Win
2. Block
3. Take center
4. Take corner
5. Take any open square
Tic Tac Toe #3
• Given a board, consider all possible moves
(future boards) and pick the best one
• Look ahead (opponent’s best move, your best
move…) until end of game
• Functions needed:
– Next move generator
– Board evaluation function
• Change these 2 functions (only) to play a
different game!
Question Answering
• Answer based on pattern matching
– Works in restricted domain (e.g. local driving
directions, directory assistance)
– Knowledge stored as canned answers
• Match question to knowledge, then generate
answer
– Wider variety of questions can be accommodated
Speech Understanding
• Directly match digits to “1” through “9”
patterns
• Learn to recognize “1” through “9” patterns by
training (feature-based)
• Recognize numbers in context, e.g. phone
number area code must be valid, prefer
numbers in address book, …