Transcript Artificial Intelligence

Artificial Intelligence
CIS 479/579
Bruce R. Maxim
UM-Dearborn
Overview
• What is artificial intelligence (AI)?
• Why study AI?
• How can you tell whether a machine or
program is intelligent?
• What kinds of things (if any) can be
learned by a machine?
• When can (should) machines replace
human experts?
What is AI?
• AI is the study of ideas that enable
computers to be intelligent
• Intelligence might be defined as the
capacity to acquire and apply
knowledge
• Ai is the part of computer science
concerned with design of computer
systems that exhibit human intelligence
AI Problem Characteristics
•
•
•
•
•
•
Decomposable to smaller or easier problems
Solution steps can be ignored or undone
Predictable problem universe
Good solutions are obvious
Uses internally consistent knowledge base
Requires lots of knowledge or uses
knowledge to constrain solutions
• Requires periodic interaction between human
and computer
Goals of AI
• To make computers more useful by
letting them take over dangerous or
tedious tasks from human
• Understand principles of human
intelligence
Physical Symbol Hypothesis
(Newell and Simon)
• Given enough symbols related to one another
• Given collection of operators capable of
producing new symbol expressions
• You have the necessary & sufficient
conditions for intelligent behavior
• The problem is can this critical mass be
achieved in real-time within the memory
capacity of some machine
AI Technique Characteristics
• They capture generalizations
• Can be understood by domain expert
• Easily modified to correct errors or
reflect changes in the world view
• Used in lots of different situations
• Reduce its own size/bulk by narrowing
the range of possibilities it must
consider at any given time
AI System Taxonomy
(easiest to hardest)
•
•
•
•
•
•
•
•
Data acquisition
Expert System
Problem Solving
Automatic Programming
Robotic Control
Pattern Recognition
Natural Language Processing
Vision & Scene Analysis
Turing Test
• Used as criteria for evaluating success of AI
system
• Inquisitor, AI System, Human
• Inquisitor connected to AI System and Human
by “wire” and free to ask either any question
• If Inquisitor can not determine which is the
Human without guessing AI System is
successful
AI Success Criteria
• Is the task clearly defined?
• Is there an implemented procedure
performing the task?
• Is there an identifiable set of regularities
or constraints that the procedure uses
to derive its power?