From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!pindor Thu Apr 30 15:22:55 EDT 1992
Article 5281 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!pindor
>From: pindor@gpu.utcs.utoronto.ca (Andrzej Pindor)
Subject: Re: Intelligence, awareness, and esthetics
Message-ID: <1992Apr27.171625.13126@gpu.utcs.utoronto.ca>
Organization: UTCS Public Access
References: <1992Apr20.191345.27706@javelin.sim.es.com> <1992Apr21.221135.20165@ccu.umanitoba.ca> <1992Apr23.152759.2272@javelin.sim.es.com> <1992Apr24.154950.25222@gpu.utcs.utoronto.ca> <1992Apr24.182714.17683@javelin.sim.es.com>
Date: Mon, 27 Apr 1992 17:16:25 GMT

In article <1992Apr24.182714.17683@javelin.sim.es.com> biesel@javelin.sim.es.com (Heiner Biesel) writes:
>pindor@gpu.utcs.utoronto.ca (Andrzej Pindor) writes:
>
>>You seem to suggest that Turing was so stupid as not to see a difference 
>>between real intelligence and good mimicry of one.
>
>I am not commenting on Turing's intelligence, but rather the efficacy
>of the test proposed by him. I suspect that the difference between
>"real" intelligence and a good mimicry of one is sufficiently subtle
>to require some thought, rather than a simple dismissal. Since the
>debate in this forum has in fact revolved around just this distinction,
>I see no reason to accept the Turing test as inherently infallible, or
>even particularly well defined.
>
I have not seen anyone (at least on this forum) suggesting that the Turing 
test is *inherently infallible*. What concerns of it being *well defined*,
it depends what you mean. If you mean a specific series of questions with
strict criteria which answers count as 'plus' and which 'minus', then you are
right. However, your complaints clearly show that you miss the point.

>>>To reiterate: the Turing test does *not* depend on being able to fool
>>>all the people all the time; hell, even I can't do that.
>>>
>>No! Turing test *depends* on not being able to fool all the people all 
>>the time.                    ^^^
>                        (I assume you meant to elide this)
>
>Really? Then I submit that no machine can ever be said to have successfully
>passed the Turing test, since it requires - by your definition - an
>infinite amount of time to demonstrate.
>
Your submission is granted. However, it seems that you want TT to be something
like , say, a test whether a number is a prime or not. This is where you miss
the point. It can not be such a test because intelligence (or self-awareness)
is also not well defined (unlike e.g. primality of a number). That is where
Turing's genius comes in - the test definitness (if I may say so) matches 
the defnitness of the property it is testing for! (you wouldn't use a pair of
pliers to catch a cloud, would you?).

>Any practical test must obviously have bounded criteria. The number and 
>qualifications of the judges, for example, as well as the period of
>testing, must be defined. "Fooling all of the people all of the time"
>is hardly rigorous, and not realizable in any case.
>
See above.

>Your conception of the Turing test seems to differ substantially from mine.
>Would you please state it simply, so that we can see where the difference
>lies?
>
You are right. Actually Antun Zirdum in the post 6198 have said all about it
and there is no need for me to repeat. Let me stress one thing - the only way 
we know to judge an entity's intelligence (or selfawareness) is its behaviour
and TT is an attempt at controled experiment: we probe with questions and
analyze the resulting behaviour.

>I am suggesting that the Turing test would be sufficient for some but not
>all people, and that that is probably true of any other such test as well.
>As for me, an example of creative art would suffice. This suggests a series
>of alternative tests, each designed to demonstrate some aspect of what
>we call intelligence and awareness. 

You seem to be overlooking the fact that intelligence and awareness
are themselves very vaguely defined concepts.

>....................................The Turing test simply exploits
>verbal capability as a possible indicator of awareness, to the exclusion
>of all other aspects. I fail to understand the almost religious zeal
>with which some people defend the Turing test as *THE* means of establishing
>intelligence - and by inference, awareness - in machines. This zeal seems to
>be at odds with the spirit of the test, which is, after all, a loosely defined
>and easily staged informal affair, surely intended as such by Saint Alan
>himself.
>
I have a very different opinion of what the spirit of the test is and if you
read Antun Zirdum's post (#6198) you will notice that I am not the only one
who thinks so.

>>Obviously you do not fully appreciate what 'Turing test' means.
>
>Perhaps you could explain to what the quoted sentence means,
>and how I can come to fully appreciate what 'Turing test' means? I am
>thus far unaware of any "Turing testing" of any machine intelligence,
>the occasional playing with Eliza clones excepted. The programs I know
>about are without exception abysmally stupid and totally devoid of anything
>approaching awareness. 
>
This has been (to my satisfaction) dealt with by another poster (sorry,
the contributor's name escaped my mind).

>Regards,
>       Heiner biesel@thrall.sim.es.com


-- 
Andrzej Pindor
University of Toronto
Computing Services
pindor@gpu.utcs.utoronto.ca


