Introduction to Artificial Intelligence

What is AI?

Knowledge Representation

Knowledge Representation Hypothesis (symbols + inference):

Knowledge Representation is the most important notion in AI and has broad implications.  In every decision support systems there is some portion of the program that represents knowledge.

Symbolic approaches to knowledge representation takes many forms. For each, there is a specific notation and type of reasoning done:

1. Formal Logic

A. Propositional Calculus

B. Predicate Calculus

Every deck of cards has at least one ace:

Representations based on formal logic use Resolution, Theorem Provers to perform inference.

2. Production Rules

3. Horn Clauses
  4. Semantic Networks

Semantic networks represent knowledge as consisting of a large network of nodes and links.  Each node represents a concept. Links represent relationships among concepts.  Graph matching techniques (comparing how two graphs are similar) enable inferences to be made on semantic networks.

5. Frames/Scripts

6. Objects (Formalism of frames/semantic nets) 7. Mathematical Equations
  8. Neural Networks
9. Genetic Algorithm/Artificial Life 10. Declarative/Procedural Representations,  Controlling the Execution
  11. Expressivess vrs Tractability
  Questions Primitive Search - Tree Searching Knowledge Engineering The "Knowledge Engineer" is the person trained in the techniques and tools of developing expert systems and other AI applications.  The KE acts as an impartial interviewer (psychologist, friend, interrigator) who interacts with the expert to extract knowledge of the domain.
    Same person may be both expert and knowledge engineer

    Project Life-Cycle
  1. Clearly Identify the Problem to be Solved
    1. Who decides?  (User driven vrs Technology Driven)
    2. Is the problem simple or hard, or just right?
  2. Identify an Expert
    1. Is he/she knowledgable?
    2. Is he/she available?
    3. Can he/she communicate?
    4. Incentives?
    5. Local vrs Distant?
  3. Reading Phase
  4. Domain Model (Conceptual Model of the Domain)
  5. Knowledge Acquisition Phase
  6. Rapid Protyping
  7. Iterative Cycling (Repeat these steps)
  8. Testing Phase
    1. Verification (Does the program do what you want it to do, is it correct)
    2. Validation (Does the program give good advice)
    3. Evaluation (Do they like it)
    4. Bottom Line (Does it make money)
  9. Marketing/Politics
  10. Maintenance

Knowledge Acquisition

Light Requirements
Falling Branches Hazard
Maintenance Requirement
Pros: Always get an answer, easy to modify
Cons: May get more than one answer

AI Languages

Prolog DEMO

Related Links

Introduction to AI