Joe Lau

Alan Turing - Can machines think?

Who was Alan Turing?

The paper has four parts:

  1. The imitation game
  2. How a computer works
  3. Replying to 9 objections against thinking computers
  4. Learning machines

The imitation game

Picture of setup

  • Turing proposed to consider the question "Can machines think?", and suggested that it should be replaced by the question of how machines will perform in an "imitation game". Why?
  • Original version of the imitation game : judge talks to a man and woman through teletype and has to decide which is which. Turing asks : what if a machine takes the man’s place? Turing proposes to replace the original question with this one.
  • Now we say: Passing the Turing test = the judge cannot do better than guessing. Sometimes the Turing test is conceived as a simple test where a computer tries to deceive a judge into thinking that it is human.
  • Other versions - A jury asks questions of a computer, and the computer passes the test when the majority believes that it is a human. Similar to the Loebner competition (but the most human-like machine wins).
  • Turing: two advantages of the proposal: (a) physical appearance and abilities should not matter (b) suitable for testing any area of knowledge.
  • Chatbots: Eliza
  • Loebner Prize: http://www.loebner.net/Prizef/loebner-prize.html

How a computer works

@I believe that in about fifty years’ time it will be possible to programme computers, with a storage capacity of about 10^9, to make them play the imitation game so well that an average interrogator will not have more than 70 percent chance of making the right identification after five minutes of questioning. ... I believe that at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted. ^^^Turing's prediction@

Note: 50 years = 2000, 10^9 (decimal units) = 1.25 gigabyte? (a high-end harddisk was around 20G in 2000)

9 objections against AI

1. The Theological Objection

  • Objection: Only those with souls can think. Only humans have souls.
  • Turing
    • The almighty is omnipotent.
    • Theological arguments "have often been found unsatisfactory in the past."

2. The ' Heads in the Sand ' Objection

  • Objection: We should not believe in thinking machines because if they were to exist, then there will be terrible consequences.
  • Turing: Many people are influenced by this line of thought because they want to think that humans are superior to everything else. But very bad argument and "consolation would be more appropriate".

3. The Mathematical Objection

  • Objection: Gödel's incompleteness theorem shows that for every machine there are questions that it cannot answer correctly.
  • Turing: No reason has been given to show that it does not apply to human beings also. We often give wrong answers.
  • Comment: Modern version of this argument from Roger Penrose's The Emperor's New Mind.

4. The Argument from Consciousness

  • Objection: A machine cannot have feelings or consciousness.
  • Turing: We rely on external evidence for the existence of consciousness.

5. Arguments from Various Disabilities

  • Objection: Machines cannot do X ( X = be kind, resourceful, beautiful, friendly, make mistakes, fall in love, learn from experience, do something new ... )
  • Turing
    • Existing machines with limited memory storage cannot. But future machines might.
    • Cannot make mistake: mistake of functioning vs mistake of conclusion

6. Lady Lovelace's Objection

@The Analytical Engine has no pretensions to originate anything. It can do whatever we know how to order it to perform.^^^Ada, Countess of Lovelace (1815-52)@

  • Objection: Machines have no originality.
  • Note: "Ada" is a programming language named after Lady Lovelace. The first female programmer?
  • Turing: Machines can take people by surprising. Original work by human beings can also be a matter of following general principles.
  • Comment: Computer music

7. The Argument from Continuity in the Nervous System

  • Objection: The nervous system (brain) is not a discrete-state machine.
  • Turing: Makes no difference to the test.

8. The Argument from Informality of Behaviour

  • Objection: Human behavior is not determined by a set of rules. Machines are. So machines cannot act like humans.
  • Note: Descartes on the creative use of language.
  • Turing: We should not confuse "rules of conduct" with "laws of behavior".

9. The Argument from ESP

  • ESP includes: Telepathy, clairvoyance, precognition, psychokinesis. Turing seems to think there is evidence for "telepathy".
  • Turing: With ESP anything might happen. Not clear how it affects the test. Put everything in a "telepathy-proof" room?
  • Note: Belief in ESP was quite common at that time. People reported experiments confirming telepathy, e.g. Rhine (1934) Extra-Sensory Perception. But the results could not be duplicated. See here.

Learning machines

  • Turing discussed building learning machines to pass the test.
  • Some ideas mentioned
    • Comparison of the learning process with evolution
    • Using a random element in searching for a solution
    • Chess as an area where machines will compete with humans

Some issues to consider - "Computation cannot explain X"

  • Current computers cannot do X, technical problem or in-principle obstacle
  • Bridging argument - X requires Y. Computation cannot produce Y.
  • Is X essential for consciousness / cognition? e.g. emotions, creativity
  • How about computation + special ingredient

An argument to support the computer model

See http://consc.net/papers/qualia.html

  • Imagine replacing a brain cell by a tiny computer that performs the same role, but carries out a digital computation inside.
  • Imagine replacing some of the cells connected to the computer by similar computers, now all connected digitally.
  • Expand the replacement to other brain cells until all brain cells have been replaced.
  • At which point would consciousness and cognition cease to exist?

Questions to think about

  1. The Dutch computer scientist has said, "the question of whether Machines Can Think ... is about as relevant as the question of whether Submarines Can Swim". (In "The threats to computing science") Do you agree? Why or why not?
  2. How is thinking different from consciousness? Is it possible for an AI system to be able to think but it still lacks consciousness?