Turing and The Test

[From Rick Marken (960131.0800)]

Hans Blom (960131) --

All I know about others is through their "wonderfully superficial"
external appearances. Not much different from a Turing Test, I would
say.

Yes. But if you had learned perceptual retrofaction (nee control) theory
(PRT) you would know how to test to determine what variables people are
retrofacting. Then you would be able to see beyond the superficial
appearance of behavior to the perceptual goals of retrofactive systems.

Since you seem to have a deep personal interest in _not_ learning perceptual
retrofaction theory (it's been well over three years and there's been no
change in your understanding that I can detect;-)) it looks like you have,
indeed, committed yourself to looking at behavior from the conventional
(Turing) perspective: so enjoy yourself in the "wonderfully superficial"
world of external appearances. You'll have lots of company;-)

Remi Cote (310196.0716) --

I agree with Hans.

Oops. Bad sign;-)

At first. What I think you think is that you can proove scientificaly that
retrofaction exist in the way M. Powers described it (even with the gain
factor modulated by error?).

Your first inclination was correct. Stick with it and learn the methods of
perceptual retrofaction theory. You _can_ prove scientifically that
retrofaction exists; we do it all the time using The Test for the Controlled
Variable. Read about The Test in B:CP and _do_ the coin game to see how The
Test works. Then read about a particularly amazing application of The Test in
the "Behavior in the first degree" chapter of my "Mind Readings" book.

Can you describe the frame of the protocol that can allow one to affirm
beyound doubt that gain factor ara ALWAYS modulated by factor.

I don't think I understand what you're asking here. Sorry.

Best

Rick

[Martin Taylor 960131 1331]

Rick Marken (960131.0800)

Sarcasm has its place, I guess. But it reflects on both parties.

Hans Blom (960131) --

All I know about others is through their "wonderfully superficial"
external appearances. Not much different from a Turing Test, I would
say.

Yes. But if you had learned perceptual retrofaction (nee control) theory
(PRT) you would know how to test to determine what variables people are
retrofacting. Then you would be able to see beyond the superficial
appearance of behavior to the perceptual goals of retrofactive systems.

Since you seem to have a deep personal interest in _not_ learning perceptual
retrofaction theory (it's been well over three years and there's been no
change in your understanding that I can detect;-)) it looks like you have,
indeed, committed yourself to looking at behavior from the conventional
(Turing) perspective: so enjoy yourself in the "wonderfully superficial"
world of external appearances. You'll have lots of company;-)

If you are conducting a Turing test and want to use the PCT "Test", then
do so. The Turing Test restricts your perceptions and actions in only
one way: you can affect the test object only through language, and you
can see only its language actions. Otherwise, do what you will. "The Test"
is as much a part of the Turing test as is anything else you might choose
to do.

Now, since you insist on it, could you please expand on your method of
perceiving what is inside a "test object" without observing what it does?

ยทยทยท

----------------
I am about to yield to temptation to be sarcastic--internal conflict:-)

Since you seem to have a deep personal interest in _not_ learning that
magic doesn't work (it's been well over three years and there's been no
change in your understanding that I can detect;-)) it looks like you have,
indeed, committed yourself to looking at behavior from the conventional
(magical) perspective: so enjoy yourself in the "wonderfully intuitive"
world of internal appearances. You'll have lots of company;-)

Sorry. But I DO remember a series of interchanges about 3 years ago in
which you did commit yourself to a belief that retrofactive systems worked
by magic;-)

Martin

Remi Cote 310196.2012(EST)

To whom it may retrofit:

In a post(960130.0730(MST)) Bill P. just make it clear, as always:
"So a second functional relationship must exist: X1 = g(y-yo)"

Does this function exist in E.Coli too?

There is a way to imitate retrofaction without using this second
functional relationship with if-then producing standard output,
output that doesen't vary in unit, and that is why I mention the
Turing test... because that point is the same that Turing made
50 years ago (or so). When dealing with black box (complex) there
is no way to be sure about architecture, so we look at output only.

But Turing (having read his "Mind" article) was aware of this limitation.
He was a great scientist (probably killed by some sort of CIA agent
because he was gay and therfore subject to blackmail not blackbox).
But he knows at that time that he was limited to this tool. And we
still are today as loong as we deal with black boxe. The question
is "Is the brain a real black box?" For some yes, for other they have
faith.

So when we look at output of a retrofaction (presumed) we may also
look at a imitation of retrofaction. That is retrofaction without
the second functional relationship X1=g(y-yo).

Why am I insisting about this, is that I want to do a retrofaction
model that will predict behavior at work, when worker are subject
to organisational changes. And it is easier to conceive a
pseudoretrofaction model than a retrofaction model. If I got
more hint from you, reader, I may even be able to frame a turing
test, or at least some sort of validation an compare the prediction
of pseudo vs analog(that is real Powerian)retrofaction model.

And also compare retro and pseudoretrofaction with expectancy model. Or
Atkinson model, or any other model. But right now I am not ready
for all this... That is why I depend on you. All of you reader that
are the real expert on this field (that is evolving fastly (good sign))
that is changing name every month (from my point of view since I am
on the list since approx. 1 month).

At least I am expecting something. That is a good start but wrong
paradigm, or is it not? Am I right in saying retrofaction without
X1=g(y-yo) is pseudoretrofaction or worst the manifestation of
an evil s-r.

REMI

[From Rick Marken (941123.1430)]

Me:

The Turing Test is based on the idea that the best model of a living system
is the one that mimics the most superficial appearances of the system's
behavior and/or does this mimicing in the most convincing way.

Jan Talmon (941122) --

I think Rick is wrong here. What Turing had in mind when he proposed the
imitation game was to have a test to decide whether computers can think. He
never intended to consider the mechnical behaviour.

But the decision regarding whether the behavior you see (verbal behavior in
the case of the original Turing Test) is that of a human or an imitation is
based on judgements about overt behavior; if the verbal replies (which are
overt behaviors) to your queries SEEM to you like they are coming from an
intelligent system, then you would guess that they are coming from an
intelligent system. Turing defined intelligence strictly behavioristically.
So your claim that

the Turing test is a TEST in the PCT sense.

is not true at all. Turing never gave clear criteria for deciding whether a
particular type of response to a question was "intelligent" (at least I
don't think he did; if so, please post these criteria); he (as far as I
know) left that judgement up to the subjective opinion of the tester. Things
are quite different in PCT; the existance of a control system is evidenced by
the fact that a variable is under control by the system. The Test is based on
specific criteria for determining whether or not a variable is under
control.

...Stevan Harnad...has argued that the Turing test is insufficient to decide
whether machines can think. He argues that a system that processes only
symbols can never pass as a model for the human cognitive powers. Such a
system would need "connections" with the physical world to relate the symbols
to what is happening in the physical world: Also the lower levels of HPCT are
needed.

An object that walks, quacks and shits like a duck is only said to to be a
good model when it's behaviour in the real physical world - including
variable disturbances - is indiscernable for the behaviour of a duck.

The Harnadian version of the Turing Test is just as behavioristic as the
original -- despite all Harnad's verbal mumbo jumbo. The concept of a
disturbance implies the existance of a controlled variable; a disturbance is
a variable that WOULD have an effect on a variable if that variable were NOT
under control. Could you show me were in the world Harnad gives clear evidence
that he knows what a controlled variable is? I've read a dismally large
amount of Harnad's writing and have seen evidence that he understood anything
about the nature of control.

It is (at least for the PCT believers) the S-R models that are used in the
existing systems that make them fail the Turing test.

Whether a system fails or passes the Turing Test is a completely subjective
affair. A control model would have the same chance of passing (being taken
for being alive in) the Turing Test as an S-R model. I have an "inverse"
version of the "Mind reading" program that illustrates this fact. The
program consists of 5 numbers that "run around" the screen in two dimensional
patterns; only one of the numbers is a control (living) system -- it is
controlling its two dimensional position on the screen with respect to a
varying reference. The other four numbers are not control (living) systems -
- they simply go to the location specified by a varying internal "command"
signal.

The position of all five numbers is also influenced by the position of the
mouse -- which is the "disturbing" variable. When there is no disturbance,
all five numbers move along different paths around the screen and it is
impossible to tell which number is the "live" one. You can "test" to see
which one is alive by applying disturbances (which are like the questions
asked of the system in the Turing Test). If you apply the disturbance
abruptly (move the mouse a short distance and stop) you can see all the
numbers move abruply; but the live number "springs back" to its original
path. If this kind of "spring back" looks like "intelliegnce" to you, then
you will correctly identify that number as the one with a "mind" -- the one
that is "alive". If, on the other hand, it doesn't look like intelligence,
you might instead be inclined to pick another number as the "intelligent" one
-- perhaps a number that moves in the fanciest two-dimensional path.

It is also possible to make one of the unintelligent numbers an "S-R" device -
- which always produces a certain response (change in position) when it is
"stimulated" by the disturbing variable. This S-R number will appear to
"spring back" when disturbed, just like the controlling number. So either
number might be selected as "intelligent" if "spring back" (irritability, I
think it's called in biology) is your criterion for intelligence as reflected
in behavior. So the S-R number could pass the Turing Test while the number
that has a mind (intentions) fails.

It is easy to "fool" someone doing the "Turing Test" (someone who doesn't
understand control theory); that is, it is easy to get such a person
to identify one of the non- controlling numbers as the one that is "alive";
all you have to do is make a number respond to a disturbance in a way that
seems "lifelike" (or "mindlike") to the person doing the test. It is much
more difficult to fool a person who knows how to do "The Test"; that is, it
is difficult to get such a person to identify one of the non-controlling
numbers as the one that is doing the controlling. This is because a person
who knows how to do The Test knows what to look for when applying
disturbances, viz. lack of effect of the disturbing variable on the
controlled variable.

Best

Rick