ARTIFICIAL INTELLIGENCE

1
ARTIFICIAL INTELLIGENCE

• Computer systems which exhibit intelligent
  behavior.
• Turing Test
• Symbolic vs. Numeric Processing (rules vs. data)
• Heuristics vs. Algorithms
• Inferencing as a Reasoning Engine
• Pattern Matching

2
AI and KNOWLEDGE SYSTEMS

• WISDOM
• KNOWLEDGE
• INFORMATION
• DATA

3
AI APPLICATIONS

• Expert systems
• Natural language processing (NLP)
• Speech recognition
• Robotics
• Computer vision
• Intelligent tutoring systems
• Neural computing
• Intelligent agents

4
EXPERT SYSTEMS What They Are

• Class of computer programs emerging from AI
  research
• Purpose Attempt to share scarce expertise
• Make expertise available when experts arent
• Train others in experts thought processes
• Attempt to make inferences, or reason, from
• A set of facts that may be INCOMPLETE or
  UNCERTAIN,
• A limited domain of (expert) knowledge
• Two distinct components
• A knowledge base
• An inference engine

5
1st EXPERT SYSTEM MYCIN

• Background Research project at Stanford
• Decision Making Environment Medical Diagnosis,
  specifically infections and appropriate
  medications for treatments
• Data Patient data
• Models Probabilistic rule base
• UI (Dialog) Question/Answer
• Knowledge Rules about medical diagnosis culled
  from doctors at Stanford Medical Center

6
MYCIN ARCHITECTURE
Consultation Program
Explanation Program
PATIENT Database
Knowledge Base
Question/Answer Program
Knowledge Acquisition Prog.
7
KNOWLEDGE REPRESENTATION SCHEMES

• IF THEN RULES
• 1ST ORDER PREDICATE ANALYSIS
• SEMANTIC NETWORK
• FRAMES

8
IF THEN RULES

• IF AND/OR
• AND/OR
• THEN
• with certainty C1 AND/OR
• with certainty C2

9
MYCIN RULE

• IF
• GramStain of organism GramNegative AND
• Morphology of organism Rod AND
• Aerobicity of organism Anaerobic
• THEN
• Organism Bacteroides with Probability .7

10
CONTROL (INFERENCE) STRATEGIES

• Rule Base
• IF X has hair, THEN X is a mammal.
• IF X gives milk, THEN X is a mammal.
• IF X eats meat, THEN X is a carnivore.
• IF X has pointed teeth, AND has claws, AND has
  forward eyes, THEN X is a carnivore.
• IF X is a mammal, AND is a carnivore AND has
  black stripes, THEN X is a tiger.
• IF X is a mammal, AND is a carnivore, AND Has
  dark spots, THEN X is a cheetah.

11
FORWARD CHAINING

• Marlin Perkins sights an animal that has hair,
  eats meat, and has dark spots. What the heck is
  it?!?
• Inference engine works as follows
• X has hair triggers Rule 1. X is a mammal is
  thus added to the working memory rule base.
• X eats meat triggers Rule 5. X is a
  carnivore is added to the working memory rule
  base.
• X is a mammal, X is a carnivore and X has
  dark spots triggers Rule 9, thus X is a
  cheetah. Run for it, Marlin!

12
EXPLANATION (AUDIT TRAIL)

• How did I arrive at this conclusion?
• Rule 1 Rule 5 Rule 9
• I first deduced from the hair that the animal is
  a mammal, and then from eating meat that the
  animal is a carnivore and finally, from the dark
  spots in conjunction with mammal and carnivore
  that the animal is a cheetah.

13
EXAMPLES OF EXPERT SYSTEMS

• TED M1 Abrams Tank Engine Repair
• FRESH CINCPACFLT scheduling and deployment of
  ships in the Pacific
• STEAMER CAI for operation and maintenance of
  1078-class frigate
• Advisor Expert System for MK92 Fire Control
  System (Prof. Kamel)

14
EXPERT SYSTEMSConditions for Application

• No algorithmic solutions are known
• Problem can be solved satisfactorily by an expert
• Significant likelihood of poor decision by
  non-expert
• Significant impact of poor decision
• Lives at stake
• Financial cost
• Time delay
• Problem domain is relatively static
• Knowledge domain is relatively static
• Availability and willingness of expert

15
SOME BENEFITS OF EXPERT SYSTEMS

• Efficiency increased output decreased decision
  time
• Capture of scarce expertise
• Training support
• Operation in hazardous environments
• Accessibility of knowledge knowledge transfer to
  remote locations

16
EXPERT SYSTEMSWhat They Are NOT

• Do NOT behave like humans
• Cant GENERATE expertise or knowledge
• Cant learn from experience
• Operate on a limited domain of knowledge
• Work best in static, vice dynamic, environments
• Have NO understanding of the problem
• User interface can become tedious
• Expert Systems ARE mainstream applications more
  than they are artificially intelligent.

17
AI and the INTERNET

• Intelligent Agents
• Assist Web browsing
• Assist in finding information and matching items
• Filter e-mail
• Access databases and e-catalogs
• Improve Internet security
• Improve network routing
• Expert Systems
• Match queries to users with FAQ
• Intelligent browsing of qualitative databases
• Browse large documents