Turings Computing Machinery and Intelligence

1
Turings Computing Machinery and Intelligence

• Part II

2
Replacing the question

• Turing wants to replace the question? Why?
• What's the replacement?
• The imitation game. What is the game?
• What advantage does the new question have over
  the old?

3
Problems with machine performance

• Machine doesn't have a body - so can't demand a
  performance.
• We don't ask a human to race an airplane (Turing
  was a runner), so we shouldn't ask the computer
  to enter beauty contests.

4
Positive Arguments

• Does Turing give any?
• He does talk about storage capacity. (A capacity
  of 109 would be enough.)
• What possible arguments could one adduce?

5
Postive Arguments

• Here's one the constraints imposed by these
  machines are really minimal. The question is how
  such machines are programmed. And Turing isn't
  saying that this has been done. Just saying the
  question of whether machines can think boils down
  to the question of whether a machine could be
  programmed to play the imitation game.

6
Objections

• Most of the paper is Turings response to
  objections.

7
Theological objection

• Whats the objection?
• Turings response
• To say machines couldn't think is to place a
  limitation on God.
• Other problems with this objection?

8
Heads in the sand Could happen but it would be
bad.

• Turing says consolation should be available in
  the possibility of the transmigration of souls.
• Turing actually held out some hope that a
  non-materialist notion of the mind supported the
  possibility of the survival of death.

9
Mathematical objection

• Well skip this one for now.

10
Argument from Consciousness

• Objection Computers are not conscious.
• Responses
• You'd have to be the machine to be sure.
• Solipsism - Turing thinks that is the consequence
  of accepting the argument from consciousness.
• (Solipsism is the view that youre the only
  person who exists. Think back to Descartes.)

11
Other minds the general issue

• Skepticism about other minds I know my mind
  through introspection. How can I know that others
  have minds?
• Solipsism perhaps Im the only mind.

12
Resolving the Problem of Other Minds

• Argument from analogy
• Behaviorism
• The explanatory power of folk psychology We
  attribute beliefs and desires and other mental
  states all the time its extremely helpful as a
  theory. We probably couldnt get by without it.
• Now think of Turings claim in this light its
  really the same, but w.r.t. machines.

13
Arguments from various disabilities

• Computers can't enjoy strawberries and cream
• Machines can't make mistakes
• Argument from continuity of the nervous system
• Argument from informality of behavior
• Argument from ESP

14
Lady Lovelace's objection

• Computer cant "originate" intelligence.
• Machines can't do anything new
• Response Consider Deep Blue

15
Learning

• Can computers learn?

16
Computer Chess

• Human chess play takes intelligence.
• What does computer chess play tell us about
  intelligence?
• Is there an analog of the Turing Test for chess?

17
How do computers play chess?

• The set of possible moves is called the problem
  space of chess.
• Computers cannot exhaustively search the problem
  space.
• Why?
• There are more possible moves than atoms in the
  universe.

18
How can there be so many moves?

• Twenty possible opening moves
• Twenty possible responses
• So 400 possible positions by the second move.
• 20,000 possible positions after the second move
• Trillions of possible positions after five moves

19
Key functions in chess machines

• Legal move generator
• Generates relevant allowable moves
• Legal move evaluator
• Evaluates the moves generated by the legal move
  generator. evaluation function

20
Chess machines use heuristics

• Heuristics are rules of thumb that help limit
  the number of moves generated and evaluated.
  minimax rule
• How many moves do humans consider and evaluate?
• Depth and breadth searching

21
HALs Human-style play

• HAL uses non-optimal strategy
• Goes for trappy strategy over optimal move.
• Uses satisficing criteria
• A strategy that generates a good enough move,
  not the best possible move.

22
Do actual computers play like HAL?

• Deep Blue and other modern chess computers
  evaluate millions of moves per second.
• This is mitigated brute force.
• Does this show that computers dont play the way
  humans do?
• How do humans play chess?

23
Deep Blues limitations

• Can humans distinguish computer chess play from
  human play?
• Its now very hard to do so.
• But there are still some differences.
• The role of emotion
• The use of the concepts never eventually and
  any time