Hello World! An Invitation to Computer Science
Lecture Slides
Artificial Intelligence
Monday May 12, 2008
- universal Turing machines
- Church-Turing thesis
- the halting problem is not computable!
From incompleteness to A.I.?
- Gödel's and Turing's results often misinterpreted:
computers and mathematics limited as compared to humans
- not what results say
can be interpreted as quite the opposite
has a lot do with birth of A.I. as field
A brief intro to A.I.
- why do we care so much about AI? specialness
- practicality: we want computers to do tasks that are
too repetitive, too time consuming, too menial, ...
- classically, two tasks:
chess
natural language
- algorithms, Church-Turing thesis, Strong AI
- behavior-based approach:
if it quacks like a duck...
Turing's Test
- based on "imitation" game
- like Turing machines, more philosophical than practical
- Turing anticipated (and refuted) many possible objections:
theological: are we special?
consciousness: may be separate from "thinking"
informality of behavior: are humans machines?
- Lanier: rather than computers passing test
maybe humans fail
- Strong AI v. Searle's Chinese Room
(Turing's original paper
"Computing
machinery and intelligence" from Mind, 1950.)
Hardware v. software and substrate neutrality
- is there something special about carbon-based brains?
- universal TMs are universal regardless of what they are made
- strong AI: intelligence is in software
- how important is parallelism?
- Penrose: could quantum mechanics play a role?
- if to err is to be human, are errors requisite for intelligence?
A.I. on the Web