Transcript Introduction to AI ( slides)

ARTIFICIAL INTELLIGENCE
CSCI/PHIL-4550/6550
(IT’S FOR REAL)
DON POTTER
Institute for Artificial Intelligence
and
Computer Science Department
UGA
AI @ UGA
* - Originated around 1985.
* - First MS degree awarded: 1988.
* - We follow an interdisciplinary
approach based on logic
programming.
Participants: Computer Science,
Philosophy, Psychology,
Linguistics, Engineering,
Business, Forestry
What is Artificial Intelligence anyway?
“The science of making machines do
things that would require intelligence if
done by people” Marvin Minsky
I like: “the science of making machines
exhibit intelligent behavior”
Neither is an attempt to make a human
nor some superior being.
INTELLIGENT BEHAVIOR
(or stuff people are good at)
* - Problem Solving
* - Learning
* - Planning
* - Perception
* - Language Processing
* - Collecting Stuff
* - Independent Action
We’re scheduling a single elimination
tennis tournament with 200 players.
How many matches will we have?
COOL DUDES
Charles Babbage considered intelligent
devices long ago. Lady Lovelace?
Alan Turing brought the notion up to
date with some math foundations and a
test (called the TURING TEST).
John McCarthy coined the name
Artificial Intelligence.
TURING TEST
Interrogator
Guy
Girl
Replace the guy with a machine. If the interrogator can’t tell,
then the machine has exhibited intelligence.
Theoretical Computer Science
•- Automata Theory
•- Complexity Theory
•- Computability Theory
AUTOMATA THEORY
•Finite Automatons
•Pushdown Automatons
•Linear Bounded Automatons
•Unbounded Automatons (aka Turning
Machines, a math model of a computer)
COMPLEXITY THEORY
•Solvable Problems
•Unsolvable Problems
COMPUTABILITY THEORY
•Decidable Problems
•Undecidable Problems
Can a problem be solved (or can I prove
that it is unsolvable)?
If it can be solved, is it easy to solve or
hard to solve?
If it is easy, then develop the algorithm
and solve it.
If it is hard to solve then try using
artificial intelligence techniques.
HARD PROBLEMS
Search Space too big to be searched in
a reasonable time by a typical (good)
algorithm.
In AI, we use heuristics (rules of thumb
learned via experience).
E.g., Medical Diagnosis
From PHILOSOPHY
* Logic
* Knowledge
* lots more neat stuff
From PSYCHOLOGY
* Learning
* Comprehension
* sure, more neat stuff
From LINGUISTICS
* Language
* Language Processing
* yea, more neat stuff
PHYSICAL SYMBOL
SYSTEM HYPOTHESIS
Using symbol manipulation, we can
achieve intelligent behavior in
machines/devices.
Newell & Simon
15-Puzzle
Water Jug Puzzle (9 & 4 want 6)
Farmer, Fox, Goat, Grain
Pick up sticks (two player, go 2
nd)
Lily Pond problem
Counterfeit Coins (81, 12)
Fast Falcon (45mph)
WHAT DO WE NEED?
•Start State
•Goal State
•Representation
•Operators (recall PSSH)
* Heuristics, the good stuff
Water Jug Problem
9-Gallon Jug
4-Gallon Jug
Problem Specs:
infinite water supply,
no markings on the jugs
can fill, transfer, and empty
Start State: Both Jugs Empty (9,0) & (4,0)
Water Jug Problem
Start State: Both Jugs Empty (9,0) & (4,0)
Goal: Six Gallons in 9-Gallon Jug (9,6) (4,_)
Representation: (Jug ID , Gallons)
Operators:
fill 9-gallon jug, empty 9-gallon jug
fill 4-gallon jug, empty 4-gallon jug
transfer contents (no overflow)
from 9-gall to 4-gall
from 4-gall to 9-gall
Step 0: (9,0) (4,0)
Step 1: (9,9) (4,0)
Step 2: (9,5) (4,4)
Step 3: (9,5) (4,0)
Step 4: (9,1) (4,4)
Step 5: (9,1) (4,0)
Step 6: (9,0) (4,1)
Step 7: (9,9) (4,1)
Step 8: (9,6) (4,4)
AI RESEARCH (flight analogy)
•Feathers
AI RESEARCH (flight analogy)
•Feathers
•Flapping
AI RESEARCH (flight analogy)
•Feathers
•Flapping
•Feathers & Flapping
AI RESEARCH (flight analogy)
•Feathers
•Flapping
•Feathers & Flapping
•Beak
AI RESEARCH (flight analogy)
•Feathers
•Flapping
•Feathers & Flapping
•Beak
Facts: lift, air pressure, laws of physics,
etc.
RECENT PROJECTS
Aerial Spray Optimization
Peanut Harvest Optimization
Medication Testing/Analysis
Snake Hunting (special math problem)
Intelligent ISs and DSSs
Weather Prediction
Robotics