[From Bill Powers (2003.12.21.1641 MST)]
Marc Abrams (2003.12.21.1252)-
[From Bill Powers
(2003.12.21.0-754 MST)]
And reference signals for the endocrine system at least, and
possibly the
immune system, are set by neural signals from the brain via the
pituitary
and (I suppose) other paths.
Not necessarily.
In the pituitary, hormones are generated at variable rates, altering the
concentrations of circulating hormones in blood. These hormones enter all
the organ systems, where they appear to act as chemical reference signals
in the target organs. They cause the organs to release their products
such as glucose, adrenalin, and so forth into the bloodstream. Uniformly,
these substances or their breakdown products feed back to the organs to
inhibit their production. So the hormone from the pituitary stimulates
the output of the organ, and that output, fed back through the
bloodstream to the same organ, directly or indirectly inhibits the output
of the organ. Clearly there is an effective comparator in each organ,
with the difference between the reference input and the feedback from the
output determining the amount of output.
The pituitary itself is a collection of feedback loops, because each
emitted hormone also feeds back into the pituitary to inhibit the rate of
production of that hormone. What excites production of each hormome is a
signal from the neural half of the pituitary gland,where signals from the
hypothalamus are received. Those signals clearly act as reference
signals. Thus we have at least a two-level hierarchy of biochemical
control systems, with the reference signal entering the highest level
being adjusted by the lower levels of the midbrain. Within the organs, of
course, there are many subsidiary loops, which can be broken down into
cellular control systems and even DNA or RNA control systems.
I agree that this interpretation is “not necessarily” so. But
there seems to be a good bit of evidence for it.
The
significance of this depends entirely on what you mean by
“activity.”
I mean ‘command-and-control’.
But that is not supportable by the kinds of measurements available for
determining brain activity.
It seems that a large majority of
the
‘communications’ that ultimately takes place betwen the brain and
spinal
column are chemical and not electrical in nature.
I doubt that very seriously, unless you literally mean
“electrical”, in which case I agree. Neural signals are not
electrical or chemical, but electrochemical. The neural impulses that
carry almost all information between the brain and spinal cord result
from a well-understood process of pumping ions in and out of nerve cells
to generate a voltage potential, which breaks down to initiate a wave of
similar breakdowns along the output axon, after which the ionic pumps
start to restore the ionic potentials again in the cell body and (through
the nodes of Renvier) all along the axon. One place where there is a
different mode of communication through axons is in the connection from
the hypothalamus to the pituitary. Apparently some of the connecting
lines are normal axons, while other “axons” are really pipes
through which neurotransmitters flow. I know of no details beyond
that.
Since neurons are living cells, of course they need nutrients and they
carry out metabolic processes that have nothing to do with signal
transmission. Perhaps these are the sorts of chemical flows you are
referring to.
The communication
through
these pathways are bi-directional as well. Again, I don’t know what you
do
or don’t know and I’m addressing this to a larger audience than
yourself
Which pathways are you referring to? All neural fibers can carry impulses
either way – “normal” or “antidromic.” But normally
the signals go only one way, because receiving dendrites cannot initiate
impulses in an incoming axon from another nerve cell. Antidromic
propagation is often used to see where signals are coming from, but this
doesn’t mean that signals normally travel backward.
The brain uses a lot of
energy and requires metabolic support.
yes, and for the longest time they thought glial cells were the
main
supporting cast. They outnumber neurons 10 to 1. But again, recent
research
has shown that glial cell are the equal of neurons in importance,
in
communication.
However, what their importance is is still not known.
Even the transmission of neural
impulses uses up energy, as does the
generation of
impulses and the pumping up of ionic potentials in the cell body
after an
impulse. So it is not at all surprising that PET scans and the like
should
pick up biochemical activity in areas where there is unusual
neural
activity.
No Bill, this stuff is not related to PET or fMRI’s scans. The
peptide
research I speak of is all about chemical assay’s.
Same differrence. The chemical assay, like PET scans and the others,
tells you only that there was activity in a certain region. It
doesn’t tell you what kind of activity it was.
I think you are confusing what
an input function does with what memory
does. Check this:
The reason I perceive glasses
or watermelons is that I have acquired
perceptual input functions which receive visual sensations and
respond to
a certain combination of them by producing signals indicating that that
a
specific configuration is present. That has nothing to do with
memory.
Sorry, It has everything to do with
memory. The fact that you can
distinguish one configuation from another is due to what you have
stored
your memory. The visual sensations have no meaning in and of
themselves.
You need to give it meaning.
No, I’m sorrier than you are, it has nothing to do with memory. What
you’re proposing is related to the “template” model of
perception, by which incoming patterns are compared with stored templates
in memory. When a match is found, the input is identified. But this model
has the flaw that an undefined mechanism is needed to receive a stream of
neural impulses, find the pattern in it, and perform comparisons with a
huge number of stored patterns in some undefined form. And this must be
done for every different perception that we can have.
The greatest problem with the template model of perception is that it is
useless for control. Merely classifying and naming patterns is not
sufficient to permit controlling them. We need continuously-variable
representations that can be in a range of states, so we can pick up the
watermelon, turn it to various angles to inspect it, and set it down on
one side preparatory to cutting it. What you see at various stages of
this process is not just “a watermelon”. It is an ovoid in
various orientations, looking circular from some angles, oval from
others, and with the long axis pointing in various directions in space.
Unless we want to believe that there is a template for every possible
orientation, size, color, and pattern of markings, we must look for a
completely different explanation of how perception of a continuously
variable configuration works.
That model is in the class that depends on “feature detection,”
though I am not sure that concept is adequate. Oliver Selfridge, I
believe, called it the “pandemonium” model. The idea is that
you have a large number of detectors that sense simple dimensions of the
array of input information, each one independently reporting how much is
present of the particular pattern it is tuned to. The one that yells the
loudest wins, according to the pandemonium model, thus explaining the
name. This is the approach that leads to a collection of perceptual input
functions each one specialized to detect just one dimension of perception
at the level in question. This leads in turn to a relatively simple model
of perception and control, but requires a large number of independent
control systems. That’s the principle behind HPCT.
It’s like a photocell with a
narrow-band color filter producing a signal
indicating the presence of light at a certain wavelength. The
recognition
is not done through comparison with memories, but through a neural
network
operating in real time.
Yes, and a photocell cannot tell you what it is perceiving
either.
It can tell you how much of its perception is present. Neural
signals can’t tell you what they represent, either. The brain learns to
identify them in terms of their relationships with other
signals.
How would you allow the photocell
to ‘tell’ you what color the wavelength is?
You would need an array of photocells with different filters in front of
them. Then you would need a way of combining signals from adjacent
photocells with weights corresponding to their filters. The resulting
signals would tell you that particular colors were present. We are
speaking, of course, of how color vision in the eye works.
Memory would enter if my brain
stored the perceptual signal coming out of
the glasses-recognizing input function, and later played it back
into the
same axons. The result, for the local control system and all
higher
systems, would be similar to the effect of experiencing the
same
sensations
again.
I don’t believe memory works this way.
Then how do you explain hearing a tone on a pitch pipe, and a few seconds
later remembering the tone? I can’t see any other way for that to
work.
Terrific. Thank you. We really
don’t differ all that much but there are
subtle and important differences between our views. It’ll be a few more
days
but I think it’ll be worth the wait. Again, your views are well stated
and
clear.
Thank you.
Bill, I’m not sure I understand
this. Are you saying that emotions only
reflect intrinsic error and not error associated with anything that
gives
rise to them (i.e. error signals)
No, just the opposite. I am saying that the feeling component of emotion
arises from variables we have sensors for in the body. But I am saying
that these sensors probably do not detect the variables I call
“intrinsic,” but only indirect effects of changes in the
intrinsic variables. So an intrinsic variable might have to do with the
level of carbonic acid in the blood, but the sensation we might feel when
this variable gets too high is “shortness of breath,” or
“tightness in the chest.” I’m guessing that the reorganizing
system works with variables much closer to actual biochemistry than the
sensations we feel.
Yes, which brings up another set of
questions. In B:CP you do not address
the hierarchy in this fashion.
If the emotion chapter had been retained, it would have.
What would your ‘biochemical’
hierarchy look
like?
Some guesses are described in the preceding. One would have to sit down
with a pile of books on biochemistry and physiology to work out more
details. Interested in taking on a project?
I love the concept. I was
thinking of something very similiar, with
3 to 5 lower levels of abstraction for the physical and at the 4th or
6th
level having a level of networked nodes representing ‘cognition’ and
the
various interconnections that might be present.
Don’t forget that you’re trying to put together a system can can control
at all levels of abstraction. Also, the biochemical hierarchy I have in
mind has nothing to do with the levels of perception and control in the
brain’s hierarchy. The two systems interact but they are different
systems.
Best,
Bill P.