Transcript Faculty of Arts Atkinson College

Welcome
Sixteenth Lecture for ITEC 1010 3.0 A
Professor G.E. Denzel
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Agenda
 Brief discussion of assignment q on changing
background colour inline.
 Finish Chapter 10 in text, dealing On-Line
Analytical Processing (OLAP) and datamining
 Discussion of Artificial Intelligence
approaches
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Using Styles
 Different browsers work differently! View
the following with IE 5, IE 6, NS 4.79, NS
6.2
 http://www.math.yorku.ca/Who/Faculty/Denzel/t
estbody.html
 http://www.math.yorku.ca/Who/Faculty/Denzel/t
estbody2.html
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
What can we do with the stored
data?
 Analytical Processing - the activity of
analyzing accumulated data
 Online analytical processing (OLAP)
 An end-user activity
 Involves large data sets with complex
relationships
 Uses Decision Support Systems models
 Is retrospective
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Online Analytical Processing
(OLAP)
Analysis by end users from their desktop, online,
using tools like spreadsheets
Analyze the relationships between many types of
business elements
Involve aggregated data
Compare aggregated data over hierarchical time
periods (monthly, quarterly, annually)
Present data in different perspectives
Involve complex calculations between data
elements
Respond quickly to users requests
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
What can we do with the stored
data?
 Data mining – intelligent search of data
stored in data marts or warehouses
 Find predictive information
 Discover unknown patterns
 End users perform mining tasks with very
powerful tools
 Mining tools apply advanced computing
techniques (learning, intelligence)
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Data Mining and Analysis
Concerns
Ethical Issues
 Valuable data-mined information may violate individual

privacy
Who is accountable for incorrect decisions that are based
on DSS?
Human judgment is fallible
Job loss due to automated decision making?


Legal Issues
 Discrimination based on data mining results
 Data security from external snooping or sabotage
 Data ownership of personal data
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Chapter Preview
In this chapter, we will study:
 What is meant by artificial intelligence
 How expert systems are developed and how they
perform
 How AI has been applied to other arenas, such as
natural language processing and neural computing
 The concept and usefulness of intelligent agents
 Ethical and legal issues posed by AI
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
‘Intelligent’ Systems?
 Conventional computer systems do not
possess ‘intelligence.’ They simply follow
step-by-step instructions to complete a task
 If a computer system had ‘intelligence,’ it
would…
 Deal successfully with complex situations
 Learn from experience
 Adapt to new situations quickly
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Why do we want ‘Intelligent’
Systems?
 To capture and represent human knowledge
permanently
 To perform tasks requiring intelligence
repetitively, consistently, and capably
 To document the performance of a task
 To conveniently disseminate knowledge
and expertise to others
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Artificial Intelligence
 Branch of computer science that
 Studies human intelligent behavior
 Attempts to replicate that human intelligent
behavior in a computer system
 Employs symbolic processing of knowledge
and heuristics
 Does not really enable computers to ‘think’
 Does enable creation of systems with some
human-like behaviors
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Applications of Artificial
Intelligence
 Expert Systems
 Natural language
technology
 Speech
understanding
 Robotics
 Computer vision
Faculty of Arts
Atkinson College
 Intelligent computer-
assisted instruction
 Machine learning
 Handwriting
recognition
 Intelligent agents
ITEC 1010 A F 2002
What is an Expert System?
 Computer system that solves a problem as
successfully as a human expert
 Incorporates human expertise
 Acquires facts about the problem
 Applies its stored knowledge and expertise
to the problem facts to derive a solution
 Makes recommendations
 Can explain its reasoning and logic
 Successful commercial application of AI
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Key Expert System Terms
 Knowledge acquisition – the process of
obtaining knowledge and expertise from human
experts
 Knowledge representation – the method used
to represent human knowledge and expertise in
the computer system
 Knowledge inferencing – the process of
applying stored expertise to the facts about the
problem to draw conclusions
 Knowledge transfer and use – the
communication of the problem solution and its
justification to the system user
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
More Expert System Terms
 Knowledge base – stored facts and methods of how to




solve a problem
Heuristic – rule of thumb that can be applied in a
problem solution
Inference engine – processing logic stored in the system
that correctly applies the stored knowledge to the problem
to develop a solution
Domain expert – one or more humans who have
achieved a high level of expertise in solving a problem
Knowledge engineer – person who develops expert
systems
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
How is an Expert System
Created?
 Knowledge engineer works with domain expert to
extract domain knowledge
 Knowledge engineer encodes domain knowledge
in knowledge base using appropriate knowledge
representation
 Knowledge engineer tests system on sample
problems and refines system knowledge with help
from domain engineer
 Refinement continues until system is solving
problems with human expert capability
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
How Does an Expert System
Perform?
 System asks user a series of questions to gather
facts about the problem
 System uses inference engine to form conclusions
from the facts, including a measure of certainty
about the conclusions
 System displays its recommendation or solution to
the problem
 If asked, the system can display its reasoning and
logic as to how it arrived at the conclusion
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Explanation
facility
Knowledge
base
Faculty of Arts
Atkinson College
Inference
engine
Knowledge
base
acquisition
facility
User
interface
Experts
User
ITEC 1010 A F 2002
Expert System Structure
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
More on Expert Systems
 Strengths
 Rapid, consistent


problem solutions
Ability to justify and
explain reasoning
Easy to replicate and
distribute to non-expert
users
Faculty of Arts
Atkinson College
Limitations
 Can only solve
problems in a narrow
domain
 Can only be applied to
certain problem types
 Cannot learn from its
experience
 Hard to acquire
knowledge from human
expert
ITEC 1010 A F 2002
Other Intelligent Systems
 Natural Language Processing
 The ability to communicate with a computer in
your natural language
• Voice (speech) recognition and speech
•
Faculty of Arts
Atkinson College
understanding – system recognizes spoken words
and understands their meaning
Voice synthesis – computer produces natural
language voice output that sounds ‘human’
ITEC 1010 A F 2002
Other Intelligent Systems
 Neural Computing
 A computer model that uses architecture that




mimics certain brain functions
Performs pattern recognition well
Can analyse large data sets and discover
patterns where rules were previously unknown
Can ‘learn’ by analysing new cases and
updating itself
Many potential business applications
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Figure 11.2 Neural Internet-based optical character recognizer.
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
More Neural Nets
Discussion of using Neural networks to predict the stockmarket --- why not?
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Other Intelligent Systems
 Case-Based Reasoning
 Uses solutions from similar problems and
adapts them to new problems
Useful in solving very complex cases

 Fuzzy Logic
 Enables systems to effectively deal with

uncertainty
Often use in combination with other
technologies to improve productivity
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Rules for a Credit Application
(Could be from neural net or expert system)
Mortgage application for a loan for $100,000 to $200,000
If there are no previous credits problems, and
If month net income is greater than 4x monthly loan payment, and
If down payment is 15% of total value of property, and
If net income of borrower is > $25,000, and
If employment is > 3 years at same company
Then accept the applications
Else check other credit rules
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Intelligent Agents
 Software agent that autonomously performs
tasks on behalf of a user with certain goals
or objectives
 Can tirelessly perform repetitive tasks over a



network
Includes knowledge base and ability to learn
Can be static (on the client only) or mobile
(move throughout a network)
Often used to facilitate search and retrieval on
the Internet and to assist in e-commerce tasks
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Examples of Agents in use today
 Search engines (yahoo, alta vista, ask Jeeves,
etc.)
 Stock trackers
 http://www.botspot.com
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Virtual Reality
 Simulation of a physical environment in a
highly realistic way
 Useful for communication and learning
 Many potential business applications,
especially marketing
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002
Intelligent Systems Concerns
 Potential to use the power of intelligent
systems in unethical ways
 Who will be accountable for decisions
made by intelligent systems?
 Who ‘owns’ knowledge and expertise? Can
an expert be ‘forced’ to reveal his/her
expertise?
Faculty of Arts
Atkinson College
ITEC 1010 A F 2002