Re: cognitive != observable? (project three)

"Bill Modlin" <[EMAIL PROTECTED]> writes:
>"Neil W Rickert" <[EMAIL PROTECTED]> wrote in message
>news:[EMAIL PROTECTED]
>> Patty <[EMAIL PROTECTED]> writes:

>> >Perhaps I should emphasize more that the stratagy is to
>> >create a platform on which we can *start* evolving.  Even a
>> >baby does not know its references when it *starts* to babble.

>> Very much the point.

>> I got into this area at precisely that point -- how does a baby (or
>> any organism) make sense of the bloomin' buzzin' confusion of signals
>> and establish reference?

>> I believe I have solved the problem.  But it turns out that
>> practically nobody is in the least bit interested.  It seems that
>> almost everybody take reference for granted, and sees me as not
>> having solved any problem at all.

>Neil, you and I have spent quite a bit of time talking past each other
>in previous years, so I'm not all that hopeful that anything good will
>come of it, but I'd like to hear your solution once again.

It will require you to be willing to put in the effort to try to
understand.  We don't share a sufficiently large set of assumed
premises.

>I also got into this area by wondering how an organism makes sense of
>the buzzing confusion of signals and establish reference.

>I certainly don't take reference for granted.

We perhaps have different meanings for "reference".  It seems to me
that you do take it for granted.

Specifically, it seems to me that you assume the sensory organs
innately make reference, and that what your neural network must do is
to discover what is already "known" to the sensory organs.

I see that as a wrong starting point.

>I believe I have a good start to a solution, though I can't say that I
>have it fully solved.

>I would be very much interested in understanding your proposed solution,
>if I can.

>>From previous conversations, I think we both feel that the solution
>involves developing appropriate perceptual mechanisms, and that this
>development is driven or guided by observable correlations among signals
>available to the organism.

But we would use correlations in very different ways.  Correlations,
by themselves, are not nearly sufficient.

>Again from previous conversations, I think that our primary difference
>of opinion is that I imagine mechanisms which build on correlations
>among clusters of signals from any source, while you seem to insist that
>the relevant correlations can only be found (at least initially) between
>action signals generated by the organism and the resulting changes in
>sensory input.

That's a misstatement.  I don't doubt that there will be correlations
all over.  The signals at one end of a neuron will be strongly
correlated to those at the other end of the same neuron.  But
building a cognitive system on such purely internal correlations is
madness.  The important correlations are those that reflect
conditions in the external world.  It is important to concentrate on
those, even if they are far weaker that other internal correlations.

>I think we have to build up a lot of perceptual structure before we can
>begin to see the correlations you are talking about, you seem to claim
>that those correlations will be the only ones sufficiently evident at
>first to be useful.

On the contrary, other correlations will be far more evident.  But
they won't be useful in developing perceptual structure.

>Would you care once again to explain your solution?  Perhaps somehow
>this time I will understand more of what you are saying.

I will try.  But I am putting it off until Monday or Tuesday, since I
won't have good access to usenet for a few days.