text 6. complement ii

Upload: cristina-eileen-hodgson

Post on 02-Jun-2018

224 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/11/2019 Text 6. Complement II

    1/21

    Can (non-human) machines think?

    Alan Turing:

    Computing Machinery andIntelligence

  • 8/11/2019 Text 6. Complement II

    2/21

    What is a computer?

    A person who makes calculations or computations; a

    calculator, a reckoner;spec.a person employed to

    make calculations in an observatory, in surveying,

    etc. Now chiefly hist. [OED]

    ENIAC: Electronic Numerical Integrator and

    Computer (1946) was the worlds first electronic

    Turing complete computer. [Wiki]

  • 8/11/2019 Text 6. Complement II

    3/21

    Logic Gate: A Simple Computer

  • 8/11/2019 Text 6. Complement II

    4/21

    More Gates, Smaller Package

  • 8/11/2019 Text 6. Complement II

    5/21

    30 years: 250,000 times more RAM

    1985:

    Kaypro: 64 KB RAM

    2012:

    Mac: 16 GB RAM

  • 8/11/2019 Text 6. Complement II

    6/21

    Could Computers Think?

    What isthinking?

    Turing means more than mere calculating.

    What kindsof things can think?

    Human beings?

    Non-human animals?

    Computers?

    How can we tellif something can think? How do I know if you are a thinking thing?

    How do I knowthatyouare conscious?

  • 8/11/2019 Text 6. Complement II

    7/21

    Turing:

    The realquestion concerns how it is that wecan tellif something is thinking.

    Turing offers a functional definition ofthinking:

    A thing thinks if it meets the same behavioral

    criteria as do the paradigm cases of thinkingthings,

    i.e., if it acts (in the relevant ways) the same way thathuman beings act.

  • 8/11/2019 Text 6. Complement II

    8/21

    The Imitation Game:

    Played with 3 people a man (A), a woman

    (B), and an interrogator (C).

    The interrogator stays in a room apart from

    the other two.

    The interrogator asks questions and tries to

    determine, from the answers, which answerer

    is the man and which is the woman.

  • 8/11/2019 Text 6. Complement II

    9/21

    Turings Strategy

    Instead of asking whether or not a computer

    can think,

    How could we measurethat?

    Ask instead whether or not a computer could

    successfully play the imitation game, and

    fool a questioner trying to figure out which

    answers come from a human.

    This is at least something we can measure.

  • 8/11/2019 Text 6. Complement II

    10/21

    Turings Imitation Game

  • 8/11/2019 Text 6. Complement II

    11/21

    Can digital computers be made to play

    satisfactorily the part of A in the imitation

    game?

    The new problem has the advantage of

    drawing a fairly sharp line between thephysical and the intellectual capacities of a

    man.

    i.e., by looking at the answers provided (rather

    than what the answerer looks like, etc.), it

    focuses on what is essential to thinking.

  • 8/11/2019 Text 6. Complement II

    12/21

    Turing:

    I believe that it will be possible to program

    computers, with a storage capacity of 109[i.e.,

    10 gigabytes], to make them play the imitation

    game so well that an average interrogator will

    not have more than 70% chance of making the

    right identification after five minutes of

    questioning.

  • 8/11/2019 Text 6. Complement II

    13/21

    Turings Claim

    Any computer (hardware plus software plusdata) that can successfully play the imitationgame

    i.e., one that can provide answers to our questionsto it that we cant distinguish from the answers

    provided by a human being

    thinks!

    I have no more reason to deny that it isconscious or has inner states than I do to denythatyoudo.

  • 8/11/2019 Text 6. Complement II

    14/21

    Turings Reasoning

    The Imitation Game gives us afunctional definition(a behavioral criterion) for whether or not somethingcan think.

    Since we cant look inside other peoples mindsthis is the best criterion we can get.

    If computers can successfully play the imitation

    game, then they meet the behavioral criterion, and so(we would have no reason to deny that) they think. So, if it quacks like a duck, its a duck!

  • 8/11/2019 Text 6. Complement II

    15/21

    Objection: The Argument from

    Consciousness

    1) Only things which are conscious

    (i.e., that have conscious mental states)can think.

    2) Computers are not conscious (i.e.,

    do not have conscious mental states). 3) Therefore, computers cannot think.

  • 8/11/2019 Text 6. Complement II

    16/21

    Turings Response:

    But how do we knowthat computers arentconscious?

    Isnt this essentially the same question we are discussing

    i.e., whether or not computers can think?

    The objection begs the question:

    The real question is whether or not computers can beconscious. The objection just assumesthat they cannot.

    How can we tell if other human beings are consciousthat other human beings can think?

  • 8/11/2019 Text 6. Complement II

    17/21

    Who thinks?

    I am pretty sure thatIthink!

    I am directly aware of my own thoughts.

    How can I tell if someone (something) elsethinks?

    If I need to be directly aware of their thoughts,

    then I cant know anyone thinks except me.

    If I dont need to be directly aware of their

    thoughts, I must rely on observablecriteria.

    The Turing Test provides such criteria.

  • 8/11/2019 Text 6. Complement II

    18/21

    I Think: Do You?

  • 8/11/2019 Text 6. Complement II

    19/21

    The upshot:

    I cant see anyones conscious mental states butmy own.

    All I can see is their behavior (how they answerquestions, etc.).

    If I judge that other people think, simply on thebasis of observing their behavior, without directlyseeing their conscious mental states, then I mustreach the same conclusion about computers.

    Alternately put: If I deny that computers can thinkbecause I cant see their inner states, then I willhave to deny that other humans can think, for thesame reason. This is solipsism.

  • 8/11/2019 Text 6. Complement II

    20/21

    Conclusion:

    The criteria we in factuse to attribute thinking(consciousness) to other human beings are behavioral.

    These criteria concern linguistic behaviorhow a think talks, not whatit looks like.

    If these are the criteria I use with other human beings, it wouldbe inconsistent to demand some higher standard ofcomputers.

    So, if a computer could meet the same standards of linguisticbehavior as do other human beings (i.e., if it could

    successfully play the imitation game), we must, on pain ofinconsistency, claim that it thinks,i.e., that it is conscious.

  • 8/11/2019 Text 6. Complement II

    21/21

    Underlying Assumptions?

    Personal Reflections

    How dowe tell that some other being (human or

    non-human) thinks or is conscious?

    Is it really by observing their behavior?

    Arent infants/pets aware of their parents/owners

    emotions (mental states)?

    Are they inferring this on the basis of the linguistic

    behavior they observe?

    Did you have to learnthat others are conscious?