12.9.15

A thing passes the Turing test for consciousness by being able to give any response that consciousness can give.

The analogous test for playing Smoke on the Water could be passed both by a guitarist and the LP.

Advertisements

9 Comments

  1. stevegbrown said,

    December 9, 2015 at 3:51 pm

    But if a thing passes the Turing test does it know that it passes the Turing test?

  2. Peter said,

    December 14, 2015 at 4:28 pm

    To pass the Turing test the LP would have to not only play Smoke on the Water, but also improvise the guitar solo.

    • December 14, 2015 at 6:15 pm

      You’re either changing the topic or saying something obviously false. The analogy is this

      Turing test : (unnamed) Playing SOTW test :: can say whatever a conscious being can say : can play whatever a player of SOTW can play.

  3. Peter said,

    December 15, 2015 at 9:36 pm

    I think you are leaving out a crucial part of the Turing test. To pass the test a computer needs to do more than just “say whatever a conscious being can say”, it needs to be able to respond to input as a conscious being would respond. Interactivity is required, otherwise the computer could just replay a chat transcript – which would be equivalent to your example of playing Smoke on the Water.

    That said, I don’t think the Turing test is much of a test. It doesn’t detect consciousness or any property that a conscious being might have, it just tests whether something is a good enough simulation of consciousness that it can fool the average human. You might as well devise a “duck-ness” test where a robotic duck lure has “duck-ness” if it can fool other ducks (or humans) into thinking it is a real duck.

    Mr. Turing seemed to be a bit impatient with the question of whether a computer can think or what that might mean, so he effectively sidestepped the question.

    • December 15, 2015 at 10:56 pm

      a computer needs to do more than just “say whatever a conscious being can say”, it needs to be able to respond to input as a conscious being would respond.

      You can’t mean that the computer must be conscious in order to “respond to input as a conscious being would respond”, since this would be begging the question.

      Interactivity is required, otherwise the computer could just replay a chat transcript.

      One meaning of “say” just is “say in response”, and that’s clearly the sense used in the test, in the OP, and in everything I’ve ever said on this thread.

      which would be equivalent to your example of playing…

      You talk of “the example of playing” when I say that thing can be played in two different ways, one of which involves a conscious being and another which doesn’t. Replace “played” with “say” or (as you prefer) “interact”.

  4. Peter said,

    December 16, 2015 at 8:42 pm

    You can’t mean that the computer must be conscious in order to “respond to input as a conscious being would respond”, since this would be begging the question.

    Isn’t that the assumption underlying the Turing test? If a computer can respond as a thinking being would, then it (presumably) can think. Or at least Mr. Turing thought that in practise it didn’t matter because people would accept it as able to think in any case. He didn’t actually refer to consciousness, but it seems to me the argument is the same if you substitute “think” with “have consciousness”.

    One meaning of “say” just is “say in response”, and that’s clearly the sense used in the test, in the OP, and in everything I’ve ever said on this thread.

    I would agree with that, but then an LP doesn’t “play in response” – except in the trivial sense that it plays music when you flip the switch.

    You talk of “the example of playing” when I say that thing can be played in two different ways, one of which involves a conscious being and another which doesn’t.

    1st way = playing back a recording
    2nd way = improvising, in response to music being played (i.e. playing a guitar solo extemporaneously)

    The problem with the test is it wouldn’t be clear whether the 2nd actually does involve a conscious being, it might just be a sophisticated example of the 1st with lots of conditional branching and preprogrammed musical theory.

    In theory the human observer is supposed to discriminate between them, but they can only do this subjectively.

    • December 17, 2015 at 3:50 pm

      Isn’t that the assumption underlying the Turing test?

      But you’re not objecting to the Turing test, but to me. This rules out simply assuming it is true.

      an LP doesn’t “play in response” – except in the trivial sense that it plays music when you flip the switch.

      Is this your claim: consciousness is interactive (or able to respond or improvise or whatever), therefore whatever we compare to it must be interactive or able to respond or improvise?

      But forget all that: I say a LP’s and Guitarists both play and Turing machines and persons both speak in response (I could have said that carpenters and saws both saw, or any other analogous name applied to a primary and secondary agent) Agree?

  5. Peter said,

    December 17, 2015 at 5:59 pm

    But you’re not objecting to the Turing test, but to me. This rules out simply assuming it is true.

    Actually I’m objecting to both 🙂
    There are problems with the Turing test – a computer could pass it and it would still be doubtful whether consciousness/thought has been detected.
    But even so, an LP which simply plays or a computer which simply speaks,/i> can’t be considered to have passed. The playing or speaking has be interactive. The test has to reveal signs that the machine has engaged with a mind and responded appropriately before it can grant a pass.

    Is this your claim: consciousness is interactive (or able to respond or improvise or whatever), therefore whatever we compare to it must be interactive or able to respond or improvise?

    I would say yes, but more importantly that is IMO what the Turing test assumes.

    I say a LP’s and Guitarists both play and Turing machines and persons both speak in response … Agree?

    Yes they do, but in the case of LPs we can say they definitely do not do so in the same way. Only the guitarist plays interactively. In the case of Turing machines it’s a little more doubtful: do they respond interactively or only appear to?

    The common sense assumption is that being “truly interactive” (whatever that might mean) is at least a necessary condition of being conscious, so if a machine passes a Turing test it’s at least possible that it might be conscious. I guess that could be wrong: maybe there can be “truly interactive” machines that are still just machines. In any case it’s an unanswered question.

    We’re stuck because we don’t have much of an idea what consciousness actually is, only some rather vague ideas about the sorts of behaviour it ought to cause. As far as we are concerned when we use the word “consciousness” we may as well be talking about ineffable souls in that we can’t detect it the way we might detect dark matter or radiation and it’s not even conceivable that we could. So we try to infer it from behaviour instead.

    The fact that we could even consider the Turing test plausible simply shows that we have a serious limitation in our ability to account for phenomena in the natural world. It’s a problem hiding in plain sight, that there is this thing called consciousness that is as mysterious today as it would have been in the Middle Ages if not more so.


Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: