Artificial Intelligence

Lect. (2)
Dr. Nader A. Mansour
[email protected]
Mechanical Engineering Department
Outlines
Previous Lecture
Artificial Intelligence
applications of AI
Soft Computing Vs. Hard Computing
Expert Systems (ES)
All Components of ES
Rule Based Systems
Individuals involved with ES
Benefits and Limitations of ES
What is Artificial Intelligence?
 System that think like human … ( General Problem Solver)
 [The automation of] activities that we associate with human
thinking, activities such as decision-making, problem solving,
learning... (Bellman, 1978)
 System that think rationally … ( The Laws of Thought Approach)
 The art of creating machines that perform functions that require
intelligence when performed by people (Kurzweil, 1990)
 System that act like human … (Turing Test Approach)
 The study of the computations that make it possible to perceive,
reason, and act (Winston, 1992)
 System that act rationally … (Rational Agent Approach)
 The branch of computer science that is concerned with the
automation of intelligent behavior (Luger and Stubblefield 1993)
Industrial applications of AI
Industrial applications of AI
Soft Computing
 SC is the fusion of methodologies designed to model and
enable solutions to real world problems which are not modeled
or too difficult to be model mathematically.
 The aim of SC is to exploit the tolerance for imprecision,
uncertainty, approximate reasoning and partial truth in order to
achieve resemblance to human like decision making.
 SC consist of several computing paradigms mainly;
Fuzzy Logic (FL) … (Reasoning and Knowledge Representation)
Neural Network (NN) … (Learning)
Genetic Algorithms (GA) … (Optimization)
Soft Computing
 SC can be seen as
Hard Computing
 conventional computing, requires a precisely stated analytical
model and often a lot of computation time
 based on binary logic, crisp systems, numerical analysis and
crisp software
 has the characteristics of precision
Soft Computing Vs. Hard Computing
 requires programs to be written; soft computing can evolve its own
programs
 Hard computing uses two-valued logic; soft computing can use
multivalued or fuzzy logic
 Hard computing is deterministic; soft computing can be stochastic
 Hard computing requires exact input data; soft computing can deal
with ambiguous and noisy data
 Hard computing is strictly sequential; soft computing allows
parallel computations
 Hard computing produces precise answers; soft computing can
yield approximate answers
Expert Systems
Expert Systems
 An automated reasoning system that attempts to mimic the
performance of the human expert
(Rosenman)
 An intelligent computer program that uses knowledge and
inference procedures to solve problems that are difficult enough
to require significant human expertise for their solution.
(Edward Feigenbaum)

 Expert System = Knowledge Base + Inference Engine

Expert Systems
An Expert System is expected to …
 Be specialist
Know facts and procedural rules
 Justify its conclusions
To establish credibility and confidence.
 Estimate the reliability of its answer
 Use heuristics
Interpolate from known facts
Expert Systems
An Expert System is expected to …
 Use heuristics
Interpolate from known facts …
Example:
Find the Catahoula dog in the following pictures:
All Components of ES
 Ideal architecture of Expert System …

Explanation
Subsystem

(Inference Engine)
All Components of ES
 Knowledge Acquisition Subsystem (Induction)
 Knowledge Base
 Inference Engine (Deduction)
 User
 User Interface
 Blackboard (Workspace)
 Explanation Subsystem (Justifier)
 Knowledge Refining System
Most ES do not have a Knowledge Refinement Component
All Components of ES
 Knowledge Acquisition Subsystem
is the accumulation, transfer and transformation of problem-
solving expertise from experts and/or documented knowledge
sources to a computer program for constructing or expanding the
knowledge base.
refers to the task of endowing expert systems with knowledge, a
task currently performed by knowledge engineers.
All Components of ES
 Difficulties with Knowledge Acquisition
Knowledge is not easy to acquire or maintain
More efficient and faster ways needed to acquire knowledge.
System's performance dependent on level and quality of
knowledge "in knowledge lies power.”
Transferring knowledge from one person to another is difficult
for these reasons:
expressing knowledge
The problems associated with transferring the knowledge to the
form required by the knowledge base.
All Components of ES
 Difficulties with Knowledge Acquisition
experts busy or unwilling to cooperate.
methods for acquiring knowledge not refined.
acquisition should involve several sources not just one.
it is often difficult to recognize the relevant parts of the expert's
knowledge.
All Components of ES
 Knowledge Base
Knowledge is the primary raw material of ES
Constructing the knowledge base of the Expert System is very
much taxing task.
The knowledge base of expert systems contains
Factual knowledge
Heuristic knowledge
All Components of ES
 Knowledge Base
Factual knowledge
is that knowledge of the task domain that is widely shared, typically
found in textbooks or journals, and commonly agreed upon by those
knowledgeable in the particular field.
Heuristic knowledge
is the less rigorous, more experiential, more judgmental knowledge
of performance.
is rarely discussed, and is largely individualistic.
is the knowledge that underlies the art of good guessing.
All Components of ES
 Inference Engine
 The brain of the ES
 The control structure or the rule interpreter
 Provides a methodology for reasoning
 Inference Engine consists of
 Interpreter
 Scheduler
 Consistency enforcer
All Components of ES
 Inference Engine
 Interpreter
 is the part of ES which uses machine reasoning strategies.
 Scheduler (planner)
 Attempts to organize the overall approach to solve the problem.
 Determines the order of inference
 Consistency Enforcer
 Judges the Confidence Factor of the ES
 Rejects tasks of the ES if confidence factor falls below a desired
level.
All Components of ES
 User Interface
Natural language would be an ideal goal for human consultations
Language processor for friendly, problem-oriented
communication
All Components of ES
 Blackboard (Workspace)
Area of working memory to
Describe the current problem
Record Intermediate results
Records Intermediate Hypotheses and Decisions
All Components of ES
 Explanation Subsystem (Justifier)
Important component of the ES is their ability to explain
themselves.
 Given that the system knows which rules were used during the
inference process, the system can provide these rules to the user as
means for explaining the results and justification of their actions
 By looking at the explanation, knowledge engineer can see how the
system is behaving and how the rules and data are interacting
Rule-Based Systems
 A very common inference engine is based on representing
knowledge in a rule base.
if X then Y
 X is the condition while Y is the consequent.
 A rule is triggered if all the conditions are satisfied and then the
consequents are fired.
 Rule based systems are used as a deduction systems
 Deductions can be either data-driven (Forward chaining - FC)
or goal-driven (Backward chaining - BC)
Rule-Based Systems
 Forward Chaining
an inference technique which uses IF THEN rules to deduce a
problem solution from initial data
Rule-Based Systems
 Forward Chaining (Example)
Rule-Based Systems
 Forward Chaining (Example)
Rule-Based Systems
 Forward Chaining (Example)
Rule-Based Systems
 Backward Chaining
an inference technique which uses IF THEN rules to repetitively
break a goal into smaller sub-goals which are easier to prove
Rule-Based Systems
 Backward Chaining (Example)
Rule-Based Systems
 Backward Chaining (Example)
Rule-Based Systems
 Backward Chaining (Example)
Individuals involved with ES
 Domain Expert
The individuals who currently are experts in solving the problem.
Has the special knowledge, judgment, experience and methods to
give advice and solve problems
 Knowledge Engineer
encodes the expert’s knowledge in a declarative form that can be
used by the ES.
 The User
The individual who will be consulting with the ES to get an
advice which would have been provided by the expert.
Individuals involved with ES
 System Builder (Programmer)
 Tool Builder
 Vendors
Benefits of Expert Systems
 Increased Output and Productivity
 Decreased Decision Making Time
 Increased Process (ES) and Product Quality
 Reduced downtime
 Capture of Scarce Expertise
 Flexibility
 Easier Equipment Operation
 Elimination of the Need for Expensive human experts
Operation in Hazardous Environments
Benefits of Expert Systems
Increased Capabilities of Other Computerized Systems
Integration of Several Experts' Opinions
Ability to Work with Incomplete or Uncertain
Information
Provide Training
Enhancement of Problem Solving and Decision Making
Improved Decision Making Processes
Improved Decision Quality
Ability to Solve Complex Problems
Knowledge Transfer to Remote Locations
Benefits of Expert Systems
Reduce risk of doing business
Doesn’t betray, cheat
Improved products and customer service
A sustainable strategic advantage
Limitations of Expert Systems
Knowledge is not always readily available
Expertise can be hard to extract from humans
Each expert’s approach may be different, yet correct
Hard, even for a highly skilled expert, to work under
time pressure
Users of expert systems have natural cognitive limits
ES work well only in a narrow domain of knowledge
Limitations of Expert Systems
Most experts have no independent means to validate
their conclusions
The vocabulary of experts is often limited and highly
technical
Knowledge engineers are rare and expensive
Lack of trust by end-users
Knowledge transfer is subject to a host of perceptual
and judgmental biases
ES may not be able to arrive at conclusions
ES sometimes produce incorrect recommendations
Famous Expert Systems
 Dendral (Stanford, Ed Feigenbaum, 1965) - organic chemical
analysis
 Macsyma (MIT, 1965) - symbolic math problems
 Mycin (Stanford, 1972) - diagnose blood diseases
 Prospector (SRI, 1972) - mineral exploration
 Caduceus (1975) - internal medicine
 Xcon and R1 (1980, 1982) - computer system configuration
(for DEC)
 Harpy - document retrieval
 Hearsay - Speech understanding
 Max (Nynex, 1992) - telephone network troubleshooting
bacterial infection diagnosis
Useful Links
Expert Systems Demos
(Exsys Corvid Expert System Demos )
References
 Expert Systems in Education and
Training, Tomas D. McFarland &
Reese Parker.