AI/ES (Artificial Intelligence / Expert System) Overview of AI

1. AI/ES(Artificial Intelligence / Expert System)Overview of AI 2012. Fall. SME., Pukyong Nat’l Univ. Kim, Minsoo

2. Contents • What is AI? • History of AI • Research Area • AI Systems

3. What is AI? • In the movies and novels, • Faithful Servants & Friends • Intelligent Machine  Autonomous Robot • Metaphor for Humanism • Man v.s. Machine • Identity Problem • Destruction v.s. New Generation A Space Odyssey 2001 Blade Runner Terminator I. Robot A.I.

4. Problem Solving What is AI? • Human? • Homo sapiens • ‘Man of wise’  Human intelligence

5. What is AI? • AI Research Agenda • Problem Solving with Intelligence • Motor Function  Walking, Driving, … • Sensation & Perception  OCR/OMR, … • Human-Like Problem Solving • Decision Making • Humanism or Humanoid(??) emotion intelligence Problem Solving Human Behavior

6. What is AI? • Four Different Definitions AI • Behavior & Thinking Process • External vs. Internal Characteristics • Reasoning (Ideal Logic vs. Rational Logic) • AreHumans rational or irrational? 1. Systems that think like humans (Cognitive Science) 2. Systems that think rationally (Production Logics) Thinking How to check the intelligence or humanness? 3. Systems that act like humans (Turing Machine) 4. Systems that act rationally (Intelligent Agents) Behavior Ideal Rational

7. What is AI? • Systems that think like humans • Cognitive Science Approach • Mimic human thinking process • Build/Simulate computer model • { all inputs }  AI system  { Human-like outputs } • 1985,John Haugeland • The exciting new effort to make computers think … machines with minds, in the full and literal sense. • “Artificial Intelligence: The Very Idea”, MIT Press • 1978,Richard E. Bellman • The automation of activities that we associate with human thinking, activities such as decision-making, problem solving, learning … • “An introduction to artificial intelligence: Can computers think?”, Boyd & Fraser Publishing

8. What is AI? • Systems that think rationally • Inference Rule Approach • Rational Thinking with Logical Inferencing • Greek Philosopher, Aristotle (Syllogistic logic) • Socrates is a man; All men are mortal, therefore Socrates is mortal • 1985,E. Charniak & D. McDermott • The study of mental faculties through the use of computational models • “Introduction to Artificial Intelligence”, Addison-Wesley • 1992,P.H. Winston • The study of the computations that make it possible to perceive, reason, and act • “Artificial Intelligence”, Addison-Wesley

9. What is AI? • Systems that act like humans • Turing Test based Approach • Can machine think?  Can machines do what we (as thinking entities) can do? • Natural Language Processing, Knowledge Representation and Store, Automatic Inferencing, Pattern Recognition, Machine Learning, … • 1990,Kurzweil • The art of creating machines that perform functions that require intelligence when performed by people • 1991,Rich & Knight • The study of how to make computers do things at which, at the moment, people are better

10. What is AI? • Systems that act rationally • Rational (Intelligent)Agent Approach • Rational behavior: individuals maximize some objective function under the constraints (or under the uncertainty) they face. • Exact inference is required but it cannot be always rational. • There are cases when a simple reflex behavior is rational. • More general and scientific approach • 1990,Schalkoff • A field of study that seeks to explain and emulate intelligent behavior in terms of computational processes • 1993,Lugar & Stubblefield • The branch of computer science that is concerned with automation of intelligent behavior

11. History of AI • The Origins of AI • Alan Turing • ’30: A computer could exhibit intelligence • brilliant mathematician • Worked to crack German codes during WW2 • Worked on the development of the 1st computer that could store a program at Manchester University • The Turing Test (1950) • ability to achieve human-level performance, sufficient to fool an interrogator

12. History of AI • 1st Period, the dawn(1943~1951) • 1943,McCulloch & Pitts • Design of Neural Network • Brain Neuron Study, Propositional Logic, Turing Test • Learning is required in the neuron’s network • 1949,Hebb’s learning rule • Early 1950s,Channon & Turing • Von Neumann computer  chess program • 1951,Minsky & Edmond • Designed SNARC(Stochastic Neural Analog Reinforcement Calculator) • Randomly connected network of Hebb synapses (about 3000 of vacuum tubes and 40 neurons)

13. History of AI • 2nd Period, Early Study(1952~1965) • Nowell & Simon,General Problem Solver • Model human problem solving process Solve restricted puzzle (Tower of Hanoi) • 1958,MaCarthy (Dartmouth  MIT) • Develop LISP • Introduced Time Sharing System • Paper: Programs with Commonsense • Advice Taker: The first proposal to use logic to represent information in a computer. • 1958,Minsky • Microworld’s problem solving (blocks world) • Wide use of Neural Networks • 1962, Widrow’s Adaline (enhanced Hebb’s learning rule) • Rosenblatt, Perceptron’s learning algorithm

14. History of AI • 3rd Period, Dark Era(1966~1974) • 1966,Negative report on machine translation • Devastated natural language research for years • 1968,Marvin Minsky & Seymour Papert • Pinpoint the limitation of Perceptron  NN research’s stagnation • Cause of depression • Early AI programs somewhat lack domain knowledge and deliver information with just simple synaptic links • Tackled somewhat complex problems from the beginning • Limitations in their basic structure/frame for intelligent behavior

15. History of AI • 4th Period, Renaissance(1975~1990) • General search problem  domain specific search problem (with specialized knowledge) • 1975,Success on the Meta-Dendral project • 1980, Spotlight on the Expert Systems • Mid-1980s, Return of NN w/ Backpropagation • Prosperous Era(1991 ~ ) • Wide variety of NN applications • 1990, Agent theory • After 2003, Information search  Mobile Multi-Agent System

16. AI Knowledge Repr. Problem Solving Knowledge System Natural Lang. Processing Learing Robotics Cognitive Model Rule Frame Sensor Controller Envorinmental Problem Solv. RBS Semantic Net. Predicate Logic Machine Translation Document Generation Interface Research Area

17. Research Area • Basic Technology in AI Learning Inference Inference Engine Expert System Knowledge Base Database Learning Model Theorem Proving Game Problem Solving Intelligent System Natural Lang. Processing Pattern Recognition System Recognition Char/Doc/Voice/Image Recognition

18. AI Systems • What is AI Systems? • Implement human mental model • System identification + System automation • 4 components of AI System • User • HCI system • Inference Engine • Knowledge Base(RB + DB) • Considerations • Kn’ Definition: acquisition & understanding • Kn’ Representation: Semantics & Classification • Kn’ Manipulation: Reasoning, Control Strategy, Ambiguity Handling, Learning, Inferencing? • Model Verification: Optimal? Available?

19. AI Systems • In the end, AI system … • acquire knowledge, represent it internally, show the processed result to user via some interface • Proper application areas • No procedural algorithm exists, only heuristics exist • Where human sensation and intuition works good • Limited knowledge workers, non-popular domain • Medical, Law, … • Including uncertain information or data • Reasonable level of data loss or existence of ambiguity • Diagnosis, Inference, Prediction System • Formal knowledge with few flexibility

20. AI Systems • Considerations for applying AI system • Domain adequacy • In this domain proper to apply AI technique? • Blind introduction can be more inefficient • Does it model the real system well? • Is it truly a AI system? • Is it efficient?