Artificial Intelligence

HAL's Capabilities

• Natural language understanding
• Speech recognition
• Speech production
• Common sense knowledge
• Planning
• Vision
• Face recognition
• Emotion
• Game playing
• Learning
• Lip reading (!)
• Singing (!)

Origins

• Alan Turing

  • Published Computing Machinery and Intelligence in 1950

  • The Turing Test

  • Language allows for in-depth probing of intelligence and consciousness:
    
    Q: In the first line of your sonnet which reads "Shall I compare thee to a summer's day," would not "a spring day" do as well or better?
    A: It wouldn't scan.
    Q: How about "a winter's day?" That would scan all right.
    A: Yes, but nobody wants to be compared to a winter's day.
    Q: Would you say Mr. Pickwick reminded you of Christmas?
    A: In a way.
    Q: Yet Christmas is a winter's day, and I do not think Mr. Pickwick would mind the comparison.
    A: I don't think you're serious. By a winter's day one means a typical winter's day, rather than a special one like Christmas.

• 1956 Dartmouth Conference

  • John McCarthy, Marvin Minsky, Claude Shannon, Nathaniel Rochester, Allen Newell, Herbert Simon, Oliver Selfridge, Trenchant More, Arthur Samuel, Ray Solomonoff

  • "Artificial Intelligence"

Early Successes (late 50s to mid 60s)

• Allen Newell and Herbert Simon (CMU): Logic Theorist, General Problem-Solver

• John McCarthy (MIT): LISP

• Arthur Samuel (IBM): Checkers

• Thomas Evans (MIT): ANALOGY

• Marvin Minsky (MIT): Neural networks

• Frank Rosenblatt (Cornell): Perceptrons

• Joseph Weizenbaum (MIT): ELIZA

The Golden Years of Symbolic AI (late 60s to late 70s)

• Marvin Minsky and Seymour Papert publish Perceptrons in 1969

• Research focuses on knowledge-representation, logic, and planning

• Many rule-based expert systems developed

  • DENDRAL: inferred molecular structure from mass spectroscopy data
  • MYCIN: diagnosed blood infections
  • XCON: configured computer systems
  • PROSPECTOR: analyzed geological data to identify oil and mineral deposits

• Terry Winograd develops SHRDLU at MIT

The Return of Neural Networks (early 80s)

• What is a neural network?

• (Re)discovery of "backpropagation learning algorithm"

• Backpropagation solved some of the problems pointed out by Minsky and Papert

• Other researchers developed new learning algorithms

• NetTalk (Sejnowski and Rosenberg): learned to pronounce English text by example

• ALVINN (CMU): learned to autonomously drive a car on the highway

Other Achievements

• IBM's Deep Blue became the world chess champion on May 11, 1997 by defeating Garry Kasparov (3.5 to 2.5)

• Deep Junior held its own against Kasparov in a February 2003 rematch (3 to 3)

• Gerry Tesauro's TD-Gammon program learned to play world-class backgammon by playing millions of games against itself

• David Cope's EMI program composes music in the style of Bach and Mozart

• Genetic Algorithms use ideas from evolutionary biology to evolve programs

• Rodney Brooks at MIT developed a new approach to robotics ("behavior-based control")

• Robotics is currently an extremely active research area

Future Prospects

• Successes so far have been limited to narrow domains

• AI systems still lack human-level flexibility and adaptiveness

• Computer speed and memory capacity continue to increase exponentially

• Ray Kurzweil predicts that a $1000 PC will match the computing speed and capacity of the human brain by the year 2020

• Hans Moravec predicts that a humanoid robot will be viable by the year 2030

• AI researchers often suffer from over-confidence

• It is not my aim to surprise or shock you—but the simplest way I can summarize is to say that there are now in the world machines that think, that learn and that create. . . . Within 10 years a computer will be world chess champion.
  —Herbert Simon, 1957

• Moral: We should be skeptical of the hype

• Fast, powerful hardware is not enough

• Software is at least as important (probably much more)

• AI has come far since Turing's 1950 article

• AI still has far to go