Newsgroups: comp.ai.philosophy
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!hookup!usc!howland.reston.ans.net!pipex!uknet!festival!edcogsci!jeff
From: jeff@aiai.ed.ac.uk (Jeff Dalton)
Subject: Re: Bag the Turing test (was: Penrose and Searle)
Message-ID: <D0Jyv9.8ot@cogsci.ed.ac.uk>
Sender: usenet@cogsci.ed.ac.uk (C News Software)
Nntp-Posting-Host: bute.aiai.ed.ac.uk
Organization: AIAI, University of Edinburgh, Scotland
References: <1994Dec5.152724.10065@oracorp.com> <D0ELL3.9xt@spss.com>
Date: Fri, 9 Dec 1994 16:52:21 GMT
Lines: 42

In article <D0ELL3.9xt@spss.com> markrose@spss.com (Mark Rosenfelder) writes:
>In article <1994Dec5.152724.10065@oracorp.com>,
>Daryl McCullough <daryl@oracorp.com> wrote:
>>markrose@spss.com (Mark Rosenfelder) writes:
>>>-- Its definition is hopelessly vague; c.a.p posters have used the term
>>>for anything from teletype exchanges limited to 5 minutes, to any kind
>>>of external behavior, to any kind of physical observable whatsoever,
>>>to unobservable phenomena as well (e.g. "thinking").  Such a wide range 
>>>of denotations does not amount to a "test" in any useful sense.
>>
>>It isn't hopelessly vague. It is completely clear that Turing meant
>>for the test to be conducted via teletype exchanges. 
>
>True; but not everybody writing in this group follows this restriction.
>I didn't make up any of the extensions to the meaning of the TT referred
>to above; they all come from exchanges I've had on comp.ai.philo.

There is some, I think minor, difficulty about exactly what Turing's
test was.  He starts with an imitation game in which both a man and
a woman are at the far ends of teletypes and the man tries to make
it hard to distinguish him from the woman.  The man and woman are
called A and B respectively something like that) and then Turing
asks us to suppose A (the man) is replaced by a machine.  

So it sounds like the machine is supposed to pretend to be a woman.
But later in the paper the comparison seems to be between a machine
and a man.  I don't think anything interesting turns on this.

But note that in Turing's test we know one of the subjects is a
human and one is not.  When people talk about a Turing Test today,
they usually don't have that kind of 1-on-1 comparison in mind.
The idea is still whether machines can be distinguished from humans,
but it doesn't involve testing by pairs where it's known that one
in each pair is a machine.

It may be that a by-pairs test is more difficult for a machine to
pass.  I think it would have eliminated at least some of the false
positives that have supposedly occurred in the past (e.g. w/ Eliza).
However, the by-pairs test is arguably less fair since it's less
like what we normally do w.r.t. other humans.

-- jd
