Information Technology Based on AI

1
Information Technology Based on AI

• What is Artificial Intelligence?
• Artificial Intelligence vs. Natural Intelligence
• Corporate Applications of Artificial Intelligence
• Concept and Applications of an Expert System
• Characteristics and Examples of an Expert System
• Structure of an Expert System
• Development of an Expert System
• Fuzzy Logic
• Neural Networks

2
What is Artificial Intelligence?

• Intelligence simulated after mans intelligence
  by the use of a computer
• Natural Intelligence vs. Artificial Intelligence
• A science that uses a computer technoloogy to
  mimic humans logical behavior
• Rational/logical behavior
• e.g., evaluation, inferencing, judgment,
  problem-solving, etc.

3
Knowledge Proc. vs. Information Proc.
Knowledge Processing Data, Concepts,
Judgments Knowledge Base (Data Base access
possible) Symbol-oriented (doesnt need
algorithms) Solutions/decisions, explanations
Information Processing Data Data
Base Numerically-oriented (uses
algorithms) Numeric information/data
Input
Built-in Base
Processing
Output
4
Artificial Intelligence Vs. Natural Intelligence
5
Corporate Applications of AI

• Industrial Robots
• Natural Language Processing
• Expert Systems
• Visual Recognition
• Machine Learning
• Fuzzy Logic
• Neural Networks

6
What is an Expert System

• Definition A computer program that represents in
  a computer knowledge related to a specific
  problem domain and uses it to solve problems like
  a human expert.
• Applied to diagnosis, prediction, design,
  interpretation, planning, etc.
• Can be helpful for experts as well as novices
• Key capabilities
• inferencing
• drawing conclusions from the results of
  inferencing
• providing explanations for inferencing
• Business applications tax advising, personnel
  recruiting, investment, etc.

7
Categories of Expert Systems
8
ES(Expert System) Vs. DSS
9
Characteristics of Expert Systems

• Based on knowledge
• Uses qualitative, rather than quantitative,
  information
• Makes use of inferencing to draw conclusions
• Uses experts heuristics
• Provides explanations for reasoning
• Can function even when some data are missing
• Capable of handling uncertain situations
• Support for a narrow problem domain

10
Examples of ES Uses

• Credit Card Approval at American Express
• The Authorizers Assistant
• determines whether to approve purchases made with
  an American Express credit card while the
  customer is awaiting to make payment
• The system handles the majority of credit card
  approval work, whereas the experts evaluate the
  remaining 5
• The productivity of the approval experts has
  increased by 20

11
Examples of ES Uses - Contd

• Ticket Auditing at Northwest Airlines
• When Northwest Airlines acquired Republic
  Airlines, its volume of operations increased to
  70,000 tickets per day, that had to be audited
  manually.
• Auditing involved a very time-consuming task of
  checking fare information on a copy of each
  ticket
• A ticket auditing expert system was developed in
  1990
• The system now audits 100 of the tickets. The
  errors made by travel agencies have decreased
  significantly, and as much as 10 million dollars
  is saved each year as a result of using the
  system.

12
Examples of ES Uses - Contd

• Commercial Loan Analysis at a Bank
• Developed by a bank that specializes in loans in
  excess of 30 million dollars (typically for
  construction projects)
• Usually a 6-month study costing 1 quarter million
  dollars is required to determine the
  approvability of the loan
• An expert system is developed to replace the
  costly study.
• The system classifies a loan into three
  categories Approve, Reject, Gray-area.
• Loan officers should take care of only the
  Gray-area loans, thereby significantly reducing
  the loan evaluation costs.

13
The Structure of an Expert System
Consulting Environment (System Use)
Development Environment (Knowledge Acquisition)
14
Development of an Expert System
15
Methods for Developing Expert Systems

• Custom Development
• Can create a system suited to the specific needs
  of an organization
• Is time consuming and requires lots of resources
• Expert System Shell
• Uses a shell to construct a system with more ease
  over a short time period
• The emphasis is placed upon the transfer of
  acquired knowledge by the user to the system
• Off-the-shelf Package
• Can simply purchase a ready-made package and use
  it with minor modification
• Can cost less and avoid the complex development
  project

16
Exsys Professional An ES Shell Illustration
17
Fuzzy Logic

• A technique that deals with uncertainties by
  simulating the process of human reasoning,
  allowing the computer to behave less precisely
  and logically than conventional computers do.
• e.g., tall, somewhat tall, really tall, very
  tall, etc.
• Sample fuzzy logic applications
• vacuum cleaners
• air conditioners
• An example of a fuzzy logic rule
• IF the temperature is very hot AND the water
  level is somewhat low
• THEN add cold water to the container.

18
Neural Networks

• A computer-based technology that is based on the
  research related to the human brain and neural
  system
• Can process large amounts of information
  concurrently
• Can use the learning function of the human brain
  to classify information, based on data of past
  experience
• Neural networks
• A model created after the biological neural
  networks
• The human brain is composed of hundreds of
  billions of neurons, which are interconnected
  with one another in a sophisticated manner.
• Primary advantages of the neural network
  technique
• Dont have to rely on the predefined
  problem-solving knowledge, but seeks a solution
  based on a vast amount of data

19
Components of a Neural Network
PE Processing Element (????)