Turing, Tests

[From Rick Marken (960201.1540)]

In flagrant disagreement with ME, Bill Powers (960201.0500 MST) says to
Martin Taylor (960131 16:00) --

I agree with your assessment [of the Turing Test] .

I'm ready for my close-up, Dr. Kevorkian;-)

All Turing asked was whether it is possible, through verbal/symbolic
communication alone, to determine whether you are interacting with a
machine or a human being.

OK, I see my problem. I was thinking about the Turing Test in terms of how
one might go about testing to determine whether or not a system was human.
The idea of the Turing Test, as I recall, is to ask questions and see if the
answers _look_ like those that would be given by a human. The basis of the
comparison is in terms of overt behavior; not (of course) in terms of
controlled (spontefacted) variables.

But (as Martin pointed out) the way you conduct the test is really not part
of Turing's suggestion for the Test. Turing's idea was simply to see if you
could tell (based _in some way_ on observable behavior) whether you were
dealing with a machine or a human. So the principle of the Turing Test (blind
comparison of observable behavior) doesn't assume a particular view of
behavior. So Martin is (I can't believe I'm going to say this) RIGHT and
I'm WRONG.

The point I was trying to make is made more clearly, I think, by my program
called "FindMind", which is available at this easy to remember Web address:

gopher://gopher.ed.uiuc.edu:70/1D-1%3A24119%3AMarken%20Programs

If you are lucky enough to have a Mac (while Apple still exists), download
FindMind.sea AND the MindRead.sea (FindMind uses a file in the MindRead
folder).

If you can get the program to run, you will see five numbers moving around
the screen; one is alive (its a control -- spontefaction -- system), the
others are not (they are producing programmed outputs). The idea (as in the
Turing Test) is to determine which number is "alive". The first thing to note
is that there is no way to tell which number is alive by just looking at the
visible behavior of the numbers; they move around in the same _kind_ of
pattern. So at this level, the numbers pass the Turing Test.

You can "question" the numbers by moving the mouse. But you have to know what
kinds of "answers" qualify a number as being "alive". All numbers respond in
some way to the "questions" (actually, disturbances) caused by the mouse; if
you didn't suspect that the "living" number might be controlling...er.
spontefacting ... it's momentary position, you probably wouldn't notice
anything particularly unusual about its response to the disturbance
(question); any number could still qualify as being "alive".

In all the discussions of the Turing Test with which I am familiar (in
cognitive psychology and AI), the determination of whether the observed
behavior is that of a human or not is based on superficial characteristics of
behavior (equivalent to looking at the pattern of movement of the numbers in
FindMind). This is why I said that the Turing Test is based on a superficial
view of behavior. But it's not really the Turing Test itself that is based on
a superficial view of behavior (as Martin correctly pointed out); it is just
that the way the Turing Test is typically carried out is (like conventional
behavioral research) superficial.

Bruce Abbott (960201.1650 EST) --

Me:

have you figured out a way to tell whether organisms in an operent
conditioning experiment are behaving as retrofaction systems or as
Killeen Machines?

Bruce:

the basic Killeen Machine is an equilibrium system...However, Killeen's
expanded Killeen Machine does add a spontefaction system to spontefact

So the answer to my question would be...?

Best

Rick