From newshub.ccs.yorku.ca!ists!torn.onet.on.ca!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!att!princeton!phoenix.Princeton.EDU!harnad Mon Jun 15 16:04:41 EDT 1992
Article 6208 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!torn.onet.on.ca!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!pacific.mps.ohio-state.edu!linac!att!princeton!phoenix.Princeton.EDU!harnad
>From: harnad@phoenix.Princeton.EDU (Stevan Harnad)
Subject: The Turing Test is not a Trick
Message-ID: <1992Jun11.154029.29686@Princeton.EDU>
Summary: Aim up, then scale down
Originator: news@ernie.Princeton.EDU
Keywords: degrees of freedom; underdetermination of theory by data
Sender: news@Princeton.EDU (USENET News System)
Nntp-Posting-Host: phoenix.princeton.edu
Organization: Princeton University
Date: Thu, 11 Jun 1992 15:40:29 GMT
Lines: 34

The Turing Test is not meant to be a trick, but (1) an empirical task
(generating human performance capacity) and (2) an ADDED intuitive
constraint (indistinguishability of the performance from that of a real
human, as judged by real humans). Replies like "If I told you what is
in my pocket I would help you guess whether I'm a human or a machine"
are not only irrelevant and invalid, but they miss the point that the
purpose of the test is a perfectly serious one -- to build systems with
human capabilities, in the hope that if they match them exactly enough,
they may even have a mind -- rather than Turing's original party game
(a good example, but people seem to be putting too much weight on the
game metaphor).

Think of Turing Testing as calling for a system that is able to
correspond as a pen-pal for a lifetime, as you can; you must simply
never be given any reason to suppose your pen-pal is not a real person,
like you. Under these conditions, the problem of how the pen-pal would
be able to provide even the kind of WRITTEN performance (TT) that in
the rest of us draws on our sensorimotor capacities and experience (as
in saying what is really in our pockets at any time) shows a real and
serious (I think fatal) limitation of the TT, relative to the TTT; this
problem has nothing at all to do with tricks.

(Nor does it have anything to do with handicaps:  Think of how far we'd
get in designing a system that had the TTT-capacity of a car if we
allowed ourselves to aim just for TTT-indistinguishability from a car
that's out of gas, or has a seized crankshaft, or no wheels. First make
something that can do everything a car can do and then worry about
scaling it down to something that can can't drive in reverse or exceed
0.5 mph. Imagine how far Newton have gotten if his laws captured only the
special case of everything standing still...)
-- 
Stevan Harnad  Department of Psychology  Princeton University
harnad@clarity.princeton.edu / harnad@pucc.bitnet / srh@flash.bellcore.com 
harnad@learning.siemens.com / harnad@elbereth.rutgers.edu / (609)-921-7771


