Reading: Kershner, Ch. 16

1. What is artificial intelligence?

-- the attempt to make machines "intelligent"

-- two goals:

1. to model human intelligence
2. to make computers more useful

2. What are the marks of intelligence?

3. How would we decide whether a computer is intelligent?

-- The Turing test (Alan Turing, 1950), modified version:

A human "judge" sits at a terminal that has a cable going under a

closed door into a room. In the room, at the other end of the cable, is

either a computer or a terminal with a human being typing at the

keyboard. The judge doesn't know which it is. The judge can type

questions etc., and reads the response that comes back from the

room. The judge must then decide whether she is communicating

with a computer or a human being at a keyboard.

IF it is a computer at the other end, and the judge guesses that

it is a human, then the computer is said to PASS the Turing test.

1. Some areas of Artificial Intelligence research:

Natural language understanding, natural language generation

Speech recognition

Vision

Robotics

Theorem-proving

Game-playing, puzzle-solving

Machine learning

Expert systems

-- and many more

1. Back to the Turing test:

-- the Loebner competitions

-- ELIZA (Joseph Weizenbaum, 1966)

-- these are NOT intelligent programs!

1. Ingredients of natural language processing (understanding & recognition):

-- syntax: grammar & parsing

-- semantics: word meaning, sentence meaning

-- pragmatics: beyond word meaning to commonsense & world knowledge