Introduction To AI

PSU CS441/541 Lecture 1
September 27, 2001

• Conceptually, What Is AI?
  • Make machines (artificial)
    • think like people (artificial brains)?
    • act like people (robots)?
    • interact like people (Turing Test)?
    • act like animals/bugs/...(bugbots)?
    • act rationally (stock pickers)?
    • act logically (theorem provers)?
    • solve problems (chess)?
  • But wait, what is AI? No right answer
  • What about
    • the soul (Descartes)?
    • introspection (Einstein)?
    • intuition (Chinese Room)?
• Operationally, What Is AI?
  • What AI researchers do!
  • Easy to understand, hard to master: see Ginsberg 1.3
  • Topics at a recent AAAI Conference (2000, Austin TX)

    Agents
    Cognitive Modeling
    Constraint Satisfaction
    Game Playing
    Human-Computer Interaction
    Knowledge Representation and Reasoning
      -Boolean Satisfiability
      -Case-Based Reasoning
      -Computational Complexity of Reasoning
      -Decision Theory
      -Logic
      -Nonmonotonic Reasoning
      -Ontology
      -Reasoning About Actions And Time
      -Spatial Reasoning
      -Uncertainty
    Machine Learning and Data Mining
    Natural Language Processing and Information Retrieval
    Planning and Scheduling
    Robotics
    Search
    
    Applications (IAAI-2000)
      

  • The [hard..soft] continuum
  • Search, KR, and the problem-solving approach
  • What is happening in my ``old country''?

    

• What ``Works'' In AI?
  • History
    • 60s-70s, grandiose predictions
    • 80s, scale-down and die-off
    • 90s-present, success and disillusionment
  • ``If computers can do it, it is not AI'': chess machines
  • Specifically
    • hacking does not work (c.f. Minsky, Semantic Information Processing, 1968
    • learning sometimes works
    • brute-force often works
    • ``pre-solving'' always works...
  • What we need is faster computers?
    • faster computers cannot win...
      • complexity theory
      • algorithmic experience
    • ...but faster computers are mandatory
      • qualitative changes with instance size
      • ability to run enough experiments
      • capacity for real problems
• The Pragmatist's Approach To AI
  • computing machines:thinking :: mechanization:labor
  • Remove the part people hate and are bad at, retain the rest
  • ``Continuum of intelligence''
• Specific Outstanding Hard Problems
  • Build semantic model corresponding to arbitrary natural language utterance
  • Automate highest-level mathematics
  • Provide accurate predictions of behavior of individual human
• Agents
  • Capture notions of ``smart'', ``entity'' (without HW)
  • Captures some of the thermostat-bug-chimp-human continuum
  • Good SW Enginering idea
  • Overgeneralizes