An alternative hierarchic connection circuit (was Re: Perceptual Cartoon)

[bnhpct]

[Bruce Nevin (2017.07.12.08:53 ET)]

Martin Taylor 2017.07.11.10.35–

Very interesting, Martin.

In the Powers circuit, the reference signal at level n branches into the perceptual input for level n+1 (when the ‘imagination switch’ is closed). Ditto in my “always on” proposal.

In your proposal, the reference signals at level n-1 (= error output from level n) branch into the perceptual input for level n+1, bypassing the perceptual input function at level n where perceptual signals from level n-1 are combined to construct level n perceptual signals.

Error output from level n branches downward to the reference input of a plurality of controllers below. The reference input there receives input from a plurality of error outputs of controllers above.

I have a hunch this may have unforeseen consequences. It may take some time to puzzle out.

···

On Tue, Jul 11, 2017 at 12:06 PM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2017.07.11.10.35]

  On 2017/07/9 6:16 PM, Bruce Nevin

wrote:

…

      The B:CP view of imagination is that a copy of the

reference signal (here, the reference for a missing bit of
input) branches across to create a perceptual input signal
returning to the originator(s) of the reference signal. B:CP
depicts this as a switch actively making and breaking a neural
connection. This has seemed implausible to me, so I proposed
that the imagination signal is always present. When there is
actual perceptual input, the copy of the reference signal
augments it; when there is not, then the copy of the reference
signal provides some (weak) input of that perception.

Here's another circuit that produces exactly the same results as

does the Powers HPCT hierarchy when it is not in imagination mode,
but also does a couple of other things that the Powers circuit does
not do. It has not been tested, but then neither has the Powers
circuit been tested in imagination mode, which is where the two
circuits differ most.

The circuit was inspired by one offered in the "Predictive Coding"

school of thought (Seth A. K., Friston K. J. (2016), * Active
interoceptive inference and the emotional brain* . Phil. Trans.
R. Soc. B 371: 20160007. http://dx.doi.org/10.1098/rstb.2016.0007 ).
Because I came across the paper in an e-mail exchange with Friston,
I call it the “Friston” connection, though it is not exactly the
same as the one in their paper.

Comments:

The diagram does not show the distribution of the reference and the

error to other higher-level units (it’s that same as to the one
shown), and the “input = …” notes in the diagram do not take into
account similarly paired inputs from other lower-level units.

When the circuit is not in "imagination mode" and the perceptual

value is outside the tolerance zone, the perceptual values at every
level are exactly as they are in the Powers circuit, with the caveat
that the “error function” is necessarily at least slightly nonlinear
because of the existence of a possible tolerance zone.

Some things this circuit does differently from the Powers hierarchy:

1. "the imagination signal is always present", but is, as Powers

suggested, the reference signal for the perception the lower-level
control unit is asked to produce. As in the Powers circuit, the
reference is determined contextually, depending not only on the one
higher-level unit depicted, but on other higher-level units as well.

2. In the Powers circuit, it is impossible to perceive a reference

value, other than by turning the switches and going into
“imagination mode”, and it is impossible to perceive an error value
by any internal means. The alternative circuit allows both to be
used on a par with the incoming sensory-based data as distinct
inputs to higher-level perceptual functions.

3. If a control unit has a non-zero tolerance zone (as Bill P. said

that any well-designed control system should do), and if the
perceptual value is within the tolerance zone, the value that is
reported to the next higher level is the reference for that
perception rather than the actual perception. I don’t know if you,
dear reader, is an amateur musician, but I am, and I find that if I
am playing for myself, I am likely to hear myself playing what I
want to hear, but if I am playing with the knowledge there is a
listener, I hear many more mistakes as my tolerance for them
shrinks. So this effect of the alternative connection seems
plausible to me. It’s not so easy to account for with the Powers
circuit that reports the actual perceptual value to the next higher
Perceptual Function regardless of the width of the tolerance zone.

Further comment:

This circuit includes only one switch, the one that breaks the

lower-level loop and prevents it from controlling its perception,
rather than the two switches in the Powers circuit. We need such a
shut-off in any case, to prevent us from acting out our dreams (or
falling out of a tree) when sleeping.

We draw such circuits with single lines that represent paths that

convey “neural currents”. But a “neural current” is a fiction, a way
of combining the firings on a bundle of neurons into a number that
can be managed in simple calculations. When you think of the lines
as bundles in which the firing rates of the individual fibres
matter, some will be responsive to slightly different inputs from
their incoming synapses than are others in the same bundle. The
effect is to spread the firings over the bundle fibres differently
for clearly perceived and for hazily perceived perceptions that have
similar neural currents. The Alternative Circuit deals with this by
always producing a blend of imagination and sensory data across an
“imagination bundle” (reference value fibres) and a “sensory bundle”
that is represented by the error signal. The clearer the lower-level
perception, the less the higher level unit’s perception uses the
imagination (a.k.a.) reference value by itself because the
concentrated perceptual bundle mates directly with the concentrated
reference bundle rather than meeting it in a defocused way.

----------

I'm not proposing this circuit as THE alternative to the Powers HPCT

connection. There are undoubtedly many other possibilities. I just
offer it as a demonstration that we should not get too hooked on the
idea that there’s only one way the hierarchy must work.

The Powers connection does lead to questions, as the quote from

Bruce illustrates. “* How can we perceive what we want, as we
consciously are able to do?”* is another, as is “* How can we
perceive the difference between where we are and where we want to
be* ?” as we also are consciously able to do. Such questions
should lead us to find circuits that don’t lead to such questions,
or that provide their answers in a simple fashion.

If (as Powers claimed) any conscious perception must be represented

in the hierarchy as a perceptual value, and if reference and error
values do not exist in the perceptual side of the hierarchy, then if
the Powers hierarchy were literally correct, both of those questions
would have to be answered with the counter-factual: “We can’t do
that.” The alternative I suggest here answers both questions by
placing reference and error separately into the perceptual upgoing
data flow, so that both are potentially available to perceptual
functions at higher levels as separate signals, perceptible and
available to consciousness.

Martin

[bnhpct]

[Bruce Nevin (2017.07.12.10:23 ET)]

Martin Taylor 2017.07.11.10.35–

"How can we perceive what we want, as we consciously are able to do?" … “How can we perceive the difference between where we are and where we want to be?” as we also are consciously able to do. [BN: You also said you included a question of mine by reference to your quotation from my post of 2017/07/9 6:16 PM, but I don’t see a question in that paragraph.]

If (as Powers claimed) any conscious perception must be represented in the hierarchy as a perceptual value, and if reference and error values do not exist in the perceptual side of the hierarchy, then if the Powers hierarchy were literally correct, both of those questions would have to be answered with the counter-factual: “We can’t do that.” The alternative I suggest here answers both questions by placing reference and error separately into the perceptual upgoing data flow, so that both are potentially available to perceptual functions at higher levels as separate signals, perceptible and available to consciousness.

But this is accommodated by any scheme that accounts for imagination. The three ideas under discussion do this by branching the reference input (or the error inputs that are combined into a reference input) over to the perceptual input side. By those means, a copy of reference (or error) values does exist on the perceptual side of the hierarchy. According to the Powers hierarchy model, then, that’s just what imagination is. So if the Powers hierarchy is literally correct, both of those questions can be answered “This is how we do that. We control copies of our references. We control or observe an imagined perception of what we want (a copy of the reference), and we control or observe a perception of the difference between that imagined perceptual state and our current perceptual state.” The latter is also an exercise of imagination because one term of the difference (a non-imagined Relationship perception) is an exercise of imagination.

···

On Tue, Jul 11, 2017 at 12:06 PM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2017.07.11.10.35]

  On 2017/07/9 6:16 PM, Bruce Nevin

wrote:

…

      The B:CP view of imagination is that a copy of the

reference signal (here, the reference for a missing bit of
input) branches across to create a perceptual input signal
returning to the originator(s) of the reference signal. B:CP
depicts this as a switch actively making and breaking a neural
connection. This has seemed implausible to me, so I proposed
that the imagination signal is always present. When there is
actual perceptual input, the copy of the reference signal
augments it; when there is not, then the copy of the reference
signal provides some (weak) input of that perception.

Here's another circuit that produces exactly the same results as

does the Powers HPCT hierarchy when it is not in imagination mode,
but also does a couple of other things that the Powers circuit does
not do. It has not been tested, but then neither has the Powers
circuit been tested in imagination mode, which is where the two
circuits differ most.

The circuit was inspired by one offered in the "Predictive Coding"

school of thought (Seth A. K., Friston K. J. (2016), * Active
interoceptive inference and the emotional brain* . Phil. Trans.
R. Soc. B 371: 20160007. http://dx.doi.org/10.1098/rstb.2016.0007 ).
Because I came across the paper in an e-mail exchange with Friston,
I call it the “Friston” connection, though it is not exactly the
same as the one in their paper.

Comments:

The diagram does not show the distribution of the reference and the

error to other higher-level units (it’s that same as to the one
shown), and the “input = …” notes in the diagram do not take into
account similarly paired inputs from other lower-level units.

When the circuit is not in "imagination mode" and the perceptual

value is outside the tolerance zone, the perceptual values at every
level are exactly as they are in the Powers circuit, with the caveat
that the “error function” is necessarily at least slightly nonlinear
because of the existence of a possible tolerance zone.

Some things this circuit does differently from the Powers hierarchy:

1. "the imagination signal is always present", but is, as Powers

suggested, the reference signal for the perception the lower-level
control unit is asked to produce. As in the Powers circuit, the
reference is determined contextually, depending not only on the one
higher-level unit depicted, but on other higher-level units as well.

2. In the Powers circuit, it is impossible to perceive a reference

value, other than by turning the switches and going into
“imagination mode”, and it is impossible to perceive an error value
by any internal means. The alternative circuit allows both to be
used on a par with the incoming sensory-based data as distinct
inputs to higher-level perceptual functions.

3. If a control unit has a non-zero tolerance zone (as Bill P. said

that any well-designed control system should do), and if the
perceptual value is within the tolerance zone, the value that is
reported to the next higher level is the reference for that
perception rather than the actual perception. I don’t know if you,
dear reader, is an amateur musician, but I am, and I find that if I
am playing for myself, I am likely to hear myself playing what I
want to hear, but if I am playing with the knowledge there is a
listener, I hear many more mistakes as my tolerance for them
shrinks. So this effect of the alternative connection seems
plausible to me. It’s not so easy to account for with the Powers
circuit that reports the actual perceptual value to the next higher
Perceptual Function regardless of the width of the tolerance zone.

Further comment:

This circuit includes only one switch, the one that breaks the

lower-level loop and prevents it from controlling its perception,
rather than the two switches in the Powers circuit. We need such a
shut-off in any case, to prevent us from acting out our dreams (or
falling out of a tree) when sleeping.

We draw such circuits with single lines that represent paths that

convey “neural currents”. But a “neural current” is a fiction, a way
of combining the firings on a bundle of neurons into a number that
can be managed in simple calculations. When you think of the lines
as bundles in which the firing rates of the individual fibres
matter, some will be responsive to slightly different inputs from
their incoming synapses than are others in the same bundle. The
effect is to spread the firings over the bundle fibres differently
for clearly perceived and for hazily perceived perceptions that have
similar neural currents. The Alternative Circuit deals with this by
always producing a blend of imagination and sensory data across an
“imagination bundle” (reference value fibres) and a “sensory bundle”
that is represented by the error signal. The clearer the lower-level
perception, the less the higher level unit’s perception uses the
imagination (a.k.a.) reference value by itself because the
concentrated perceptual bundle mates directly with the concentrated
reference bundle rather than meeting it in a defocused way.

----------

I'm not proposing this circuit as THE alternative to the Powers HPCT

connection. There are undoubtedly many other possibilities. I just
offer it as a demonstration that we should not get too hooked on the
idea that there’s only one way the hierarchy must work.

The Powers connection does lead to questions, as the quote from

Bruce illustrates. “* How can we perceive what we want, as we
consciously are able to do?”* is another, as is “* How can we
perceive the difference between where we are and where we want to
be* ?” as we also are consciously able to do. Such questions
should lead us to find circuits that don’t lead to such questions,
or that provide their answers in a simple fashion.

If (as Powers claimed) any conscious perception must be represented

in the hierarchy as a perceptual value, and if reference and error
values do not exist in the perceptual side of the hierarchy, then if
the Powers hierarchy were literally correct, both of those questions
would have to be answered with the counter-factual: “We can’t do
that.” The alternative I suggest here answers both questions by
placing reference and error separately into the perceptual upgoing
data flow, so that both are potentially available to perceptual
functions at higher levels as separate signals, perceptible and
available to consciousness.

Martin

[bnhpct]

I see only ASCII trash like this:

···

On Wed, Jul 12, 2017 at 7:45 PM, Erling Jorgensen EJorgensen@riverbendcmhc.org wrote:

[From Erling Jorgensen (2017.07.12 1900 EDT)]

Martin Taylor 2017.07.11.10.35

Hi Martin,

I have some clarification questions about some of your alternative hierarchic circuitry.

MT:

EJ: I’m looking at the middle diagram in the lower row, your Possible Alternative for what you call a “Tolerance” arrangement. I’m calling these levels N and N-1, with an upper Error arrow heading off toward level N+1.

EJ: I do not see a mechanism to make e = 0. You claim, with your perceptual input description to level N, that r - e = r =/ p. And, yes, if e = 0, then r - e = r. However, the only way to get e = 0 (in the existing diagram) is if p = r at level N-1, which contradicts the inequality of your claim. Perhaps you are meaning to say that r is routed into the level N perceptual input rather than p (or the p equivalent) from level N-1.

EJ: Based on your heading of calling this a “Tolerance” arrangement, I am imagining that there might be threshold levels introduced to the Comparator at level N-1. And I believe that could be accomplished by graded potentials in neurons. However, I don’t see how thresholds would differ from additional de facto reference inputs. {This leaves aside the neurophysiological problem of needing a double set of thresholds for bi-directional neural control when rate-of-firing signals can only pull, not push – i.e., what is sometimes discussed as “half-wave rectification.”)

EJ: It is possible that thresholds could be introduced by something similar to “gain control,” but again, there is no hint of that in your diagram.

EJ: A more qualitative question: Is your “Tolerance” arrangement supposed to produce what Bill P. called the “Passive Observation” mode? It doesn’t look like it from how you display Powers’ version in the top row. And your diagram of the top middle arrangement suggests that when there is zero error there is zero output. I believe zero error only means that there is zero change in output. For instance, when we are sitting comfortably, with no apparent error, our muscle tone and opponent-process maintenance of posture does not go to zero and become all flabby. It just doesn’t need to change, until there is detectable error (outside of any tolerance zones) again.

EJ: I do think we should be conservative about claiming what exact results such arrangements might produce, until they are tested via simulation. That after all is the reality test built in by the environment, as it “disposes” of some possibilities despite what our perception “proposes.”

All the best,

Erling

---

Disclaimer: This message is intended only for the use of the individual or entity to which it is addressed, and may contain information that is privileged, confidential and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, or the employer or agent responsible for delivering the message to the intended recipient, you are hereby notified that any dissemination, distribution or copying of this communication is strictly prohibited. If you have received this communication in error, please notify the sender immediately by telephone and delete the material from your computer. Thank you for your cooperation.

[_Erling_Jorgensen]

[From Erling Jorgensen (2017.07.12 1900 EDT)]

Martin Taylor 2017.07.11.10.35

Hi Martin,

I have some clarification questions about some of your alternative hierarchic circuitry.

MT:

EJ: I’m looking at the middle diagram in the lower row, your Possible Alternative for what you call a “Tolerance” arrangement. I’m calling these levels N and N-1, with an upper Error arrow heading off toward level N+1.

EJ: I do not see a mechanism to make e = 0. You claim, with your perceptual input description to level N, that r - e = r =/ p. And, yes, if e = 0, then r - e = r. However, the only way to get e = 0 (in the existing diagram) is if p = r at level N-1, which contradicts the inequality of your claim. Perhaps you are meaning to say that r is routed into the level N perceptual input rather than p (or the p equivalent) from level N-1.

EJ: Based on your heading of calling this a “Tolerance” arrangement, I am imagining that there might be threshold levels introduced to the Comparator at level N-1. And I believe that could be accomplished by graded potentials in neurons. However, I don’t see how thresholds would differ from additional de facto reference inputs. {This leaves aside the neurophysiological problem of needing a double set of thresholds for bi-directional neural control when rate-of-firing signals can only pull, not push – i.e., what is sometimes discussed as “half-wave rectification.”)

EJ: It is possible that thresholds could be introduced by something similar to “gain control,” but again, there is no hint of that in your diagram.

EJ: A more qualitative question: Is your “Tolerance” arrangement supposed to produce what Bill P. called the “Passive Observation” mode? It doesn’t look like it from how you display Powers’ version in the top row. And your diagram of the top middle arrangement suggests that when there is zero error there is zero output. I believe zero error only means that there is zero change in output. For instance, when we are sitting comfortably, with no apparent error, our muscle tone and opponent-process maintenance of posture does not go to zero and become all flabby. It just doesn’t need to change, until there is detectable error (outside of any tolerance zones) again.

EJ: I do think we should be conservative about claiming what exact results such arrangements might produce, until they are tested via simulation. That after all is the reality test built in by the environment, as it “disposes” of some possibilities despite what our perception “proposes.”

All the best,

Erling

···

---

Disclaimer: This message is intended only for the use of the individual or entity to which it is addressed, and may contain information that is privileged, confidential and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, or the employer or agent responsible for delivering the message to the intended recipient, you are hereby notified that any dissemination, distribution or copying of this communication is strictly prohibited. If you have received this communication in error, please notify the sender immediately by telephone and delete the material from your computer. Thank you for your cooperation.

[MartinT]

[Martin Taylor 2017.07.11.10.35]

Here's another circuit that produces exactly the same results as

does the Powers HPCT hierarchy when it is not in imagination mode,
but also does a couple of other things that the Powers circuit does
not do. It has not been tested, but then neither has the Powers
circuit been tested in imagination mode, which is where the two
circuits differ most.
The circuit was inspired by one offered in the “Predictive Coding”
school of thought (Seth A. K., Friston K. J. (2016), . Phil. Trans.
R. Soc. B 371: 20160007. ).
Because I came across the paper in an e-mail exchange with Friston,
I call it the “Friston” connection, though it is not exactly the
same as the one in their paper.
Comments: The diagram does not show the distribution of the reference and the
error to other higher-level units (it’s that same as to the one
shown), and the “input = …” notes in the diagram do not take into
account similarly paired inputs from other lower-level units.
When the circuit is not in “imagination mode” and the perceptual
value is outside the tolerance zone, the perceptual values at every
level are exactly as they are in the Powers circuit, with the caveat
that the “error function” is necessarily at least slightly nonlinear
because of the existence of a possible tolerance zone.
Some things this circuit does differently from the Powers hierarchy:

1. “the imagination signal is always present”, but is, as Powers
   suggested, the reference signal for the perception the lower-level
   control unit is asked to produce. As in the Powers circuit, the
   reference is determined contextually, depending not only on the one
   higher-level unit depicted, but on other higher-level units as well.
2. In the Powers circuit, it is impossible to perceive a reference
   value, other than by turning the switches and going into
   “imagination mode”, and it is impossible to perceive an error value
   by any internal means. The alternative circuit allows both to be
   used on a par with the incoming sensory-based data as distinct
   inputs to higher-level perceptual functions.
3. If a control unit has a non-zero tolerance zone (as Bill P. said
   that any well-designed control system should do), and if the
   perceptual value is within the tolerance zone, the value that is
   reported to the next higher level is the reference for that
   perception rather than the actual perception. I don’t know if you,
   dear reader, is an amateur musician, but I am, and I find that if I
   am playing for myself, I am likely to hear myself playing what I
   want to hear, but if I am playing with the knowledge there is a
   listener, I hear many more mistakes as my tolerance for them
   shrinks. So this effect of the alternative connection seems
   plausible to me. It’s not so easy to account for with the Powers
   circuit that reports the actual perceptual value to the next higher
   Perceptual Function regardless of the width of the tolerance zone.
   Further comment:
   This circuit includes only one switch, the one that breaks the
   lower-level loop and prevents it from controlling its perception,
   rather than the two switches in the Powers circuit. We need such a
   shut-off in any case, to prevent us from acting out our dreams (or
   falling out of a tree) when sleeping.
   We draw such circuits with single lines that represent paths that
   convey “neural currents”. But a “neural current” is a fiction, a way
   of combining the firings on a bundle of neurons into a number that
   can be managed in simple calculations. When you think of the lines
   as bundles in which the firing rates of the individual fibres
   matter, some will be responsive to slightly different inputs from
   their incoming synapses than are others in the same bundle. The
   effect is to spread the firings over the bundle fibres differently
   for clearly perceived and for hazily perceived perceptions that have
   similar neural currents. The Alternative Circuit deals with this by
   always producing a blend of imagination and sensory data across an
   “imagination bundle” (reference value fibres) and a “sensory bundle”
   that is represented by the error signal. The clearer the lower-level
   perception, the less the higher level unit’s perception uses the
   imagination (a.k.a.) reference value by itself because the
   concentrated perceptual bundle mates directly with the concentrated
   reference bundle rather than meeting it in a defocused way.
   ---------- I’m not proposing this circuit as THE alternative to the Powers HPCT
   connection. There are undoubtedly many other possibilities. I just
   offer it as a demonstration that we should not get too hooked on the
   idea that there’s only one way the hierarchy must work. The Powers connection does lead to questions, as the quote from
   Bruce illustrates. " is another, as is “?” as we also are consciously able to do. Such questions
   should lead us to find circuits that don’t lead to such questions,
   or that provide their answers in a simple fashion.
   If (as Powers claimed) any conscious perception must be represented
   in the hierarchy as a perceptual value, and if reference and error
   values do not exist in the perceptual side of the hierarchy, then if
   the Powers hierarchy were literally correct, both of those questions
   would have to be answered with the counter-factual: “We can’t do
   that.” The alternative I suggest here answers both questions by
   placing reference and error separately into the perceptual upgoing
   data flow, so that both are potentially available to perceptual
   functions at higher levels as separate signals, perceptible and
   available to consciousness.
   Martin

···

On 2017/07/9 6:16 PM, Bruce Nevin
wrote:

…

      The B:CP view of imagination is that a copy of the

reference signal (here, the reference for a missing bit of
input) branches across to create a perceptual input signal
returning to the originator(s) of the reference signal. B:CP
depicts this as a switch actively making and breaking a neural
connection. This has seemed implausible to me, so I proposed
that the imagination signal is always present. When there is
actual perceptual input, the copy of the reference signal
augments it; when there is not, then the copy of the reference
signal provides some (weak) input of that perception.

•  Active
  

interoceptive inference and the emotional brain*http://dx.doi.org/10.1098/rstb.2016.0007

•  How can we perceive what we want, as we
  

consciously are able to do?"** How can we
perceive the difference between where we are and where we want to
be*

[MartinT]

[Martin Taylor 2017.07.17.23.19]

[From Erling Jorgensen (2017.07.12 1900 EDT)]

Martin Taylor 2017.07.11.10.35

Hi Martin,

      I have some clarification questions about some of your

alternative hierarchic circuitry.

Thanks for these. Answers below.

MT:

      EJ:  I'm looking at the middle diagram in the lower row,

your Possible Alternative for what you call a “Tolerance”
arrangement. I’m calling these levels N and N-1, with an
upper Error arrow heading off toward level N+1.

EJ: I do not see a mechanism to make e = 0.

No, there's no mechanism, nor does there need to be. In engineering

terms “Tolerance” means that if x is supposed to have a value r, and
there is a tolerance zone of half-width t, then when r-t < x <
r+t, x is taken to be equal to r. In the days when I used to build
circuits from components like resistors and capacitors, resistors
came in at least three grades – 1%, 5%, and 10%. If you wanted a
5kΩ resistor that was very precise, you chose a very expensive 1%
resistor. If you just want something around that much resistance,
you chose a cheap 10% one. Either way, your circuit diagram was
marked 5kΩ, with a possible note as to the tolerance. My
understanding was that they were all manufactured together, and then
were sorted into bins according to their actual resistances, so
there were o 10% resistors that were within 5% of the nominal value.
How true this was, I don’t know, but it illustrates the principle.

 Control systems often include a small tolerance zone, to avoid

unnecessary noise-related jiggling around the reference value. So if
in the middle column r-t < p < r+t, then e=0. That’s the same
for both circuits.

      You claim, with your perceptual input description to level

N, that r - e = r =/ p. And, yes, if e = 0, then r - e = r.
However, the only way to get e = 0 (in the existing diagram)
is if p = r at level N-1, which contradicts the inequality of
your claim. Perhaps you are meaning to say that r is routed
into the level N perceptual input rather than p (or the p
equivalent) from level N-1.

Does the above solve this? In the alternative circuit, p at level

N-1 is never routed into the level N perceptual input. The
equivalent is level N-1 r and level N-1 e, which the level N
perceptual function can use, if necessary, as N-1 p by subtracting
r-e.

      EJ:  Based on your heading of calling this a "Tolerance"

arrangement, I am imagining that there might be threshold
levels introduced to the Comparator at level N-1. And I
believe that could be accomplished by graded potentials in
neurons. However, I don’t see how thresholds would differ
from additional de facto reference inputs.

I don't understand what "additional reference inputs" would be.

      {This leaves aside the neurophysiological problem of

needing a double set of thresholds for bi-directional
neural control when rate-of-firing signals can only pull, not
push – i.e., what is sometimes discussed as “half-wave
rectification.”)

Don't leave it aside. It's probably the way a tolerance zone would

be implemented. All comparators have to be doubled if they are to
accomplish symmetrical outputs of positive and negative values for
the error, because they can’t output less than zero. The tolerance
zone is created by including a slight inhibition that reduces the
output that would be slightly positive so that it becomes zero, and
the same inhibition applies to both positive and negative sides of
the circuit. It’s not a very difficult circuit to produce, and
variation of the inhibitory strength would influence the width of
the tolerance zone. We usually ignore this nicety, but if the
non-negativity of he neural firing rate must be considered, you have
to deal with it.

      EJ:  It is possible that thresholds could be introduced by

something similar to “gain control,” but again, there is no
hint of that in your diagram.

No, but if adjustment of tolerance width is wanted (and I think it

is), it is identical in both circuits, so there’s no point in adding
to the necessary complication involved in comparing the two
circuits.

      EJ:  A more qualitative question:  Is your "Tolerance"

arrangement supposed to produce what Bill P. called the
“Passive Observation” mode?

No. That is not included in the diagrams for either circuit. In the

alternative circuit it means a separate place where the n-1 loop is
broken, whereas in the Powers circuit it means switching only one of
the two switches.

      It doesn't look like it from how you display Powers'

version in the top row. And your diagram of the top middle
arrangement suggests that when there is zero error there is
zero output. I believe zero error only means that there is
zero change in output.

Neither diagram makes any claim as to what zero error means with

respect to output. That is determined by the properties of the
output function. Indeed, if the output function is an integrator (at
least a non-leaky one), zero error means zero change in the output.
But this is a property of the control loop, and the two circuits do
not differ in that respect.

…

      EJ:  I do think we should be conservative about claiming

what exact results such arrangements might produce, until they
are tested via simulation. That after all is the reality test
built in by the environment, as it “disposes” of some
possibilities despite what our perception “proposes.”

True, but when two circuits have the same mathematical analyses, so

that equivalent variables have the same dependence on each other,
then they will behave the same.

Where the alternative circuit needs simulation testing is in the

things that it seems to do that the Powers circuit does not,. In the
case of the circuit I presented, I think that is primarily in the
graded relative importance of imagination in the inputs to a
perceptual function as the perceptual input become more or less
uncertain without changing its value. That’s an effect of thinking
not of neural currents but of the individual fibres whose firings
are added together to produce the neural current.

But there are lots of things that happen when you do the more

realistic thing, recognizing that each of the fibres in the “bundle”
has its individual synaptic connections, which makes them all
slightly different in their sensitivities to “off-optimum” patterns
of input for that bundle. This is a whole new ball of wax, which
should be discussed separately from alternative connection
structures for the same hierarchy.

Martin

[MartinT]

[Martin Taylor 2017.07.13.13.25]

[Bruce Nevin (2017.07.13.13:03 ET)]

Martin Taylor 2017.07.12.16.34 –

Two parts, answered separately.

        > Either the diagram is wrong or you are misreading

it

Here is my reading.

        In the Powers diagram, the reference signal for the

reference input of a specific controller below branches to
join the perceptual signal sent from that controller to the
perceptual input of the specific controller whose error
signal is the basis of the reference signal, making a closed
loop. A caveat is that this branching must comprise any
amplification of the reference signal, which is generally
accomplished by branching to create multiple copies going to
multiple synapses at the reference input. In the Powers
diagram, this branch from reference output to perceptual
input is connected or disconnected to produce several modes
of operation, the control, observation, and imagination
modes.

        In your diagram (as in mine), the branch from reference

output to perceptual input is always connected. In addition,
a copy of the error output of a lower system is sent back up
to the perceptual input of the higher system. The perceptual
signal presumably goes from the lower system to the higher
as well, but that is not shown.

No it's not shown because in the alternative circuit it does not

exist. Its place is taken by the fact that r-e = p if the circuit is
working in pure sensory data mode and p is not in the tolerance
zone. (The equation is approximate if the perception is close to the
tolerance zone, but that depends on separate assumptions about the
shape of the curve that relates e to r-p; those assumptions apply to
both circuits).

        I have assumed that this is an oversight, maybe a

consequence of placing the “input=” label in the way. So as
I understand it, the error signal from every comparator is
sent to two places, the reference output to lower systems
and the perceptual input that the comparator is passing up
to higher systems.

Maybe you are saying as I intend it, but your wording is ambiguous.

Here in words are the connections as I intended. Wherever in the
Powers diagram “p” goes up from level N-1 to the perceptual input at
level N, that “p” is replaced by r and e from level N-1 going up to
the perceptual input at level N. Going down, the two circuits are
identical. The switches that shift modes are different. Within any
level the connections are unaltered, except that in the alternative
circuit there is a switch to break the control loop by preventing
the error signal from connecting to the output function.

The alternative circuit does have a problem that needs to be fixed.

With just one switch, only two modes are available, which we can
label Imagination and Normal Control. To get a third mode
“Observation” the lower loop needs to be broken (which the Powers
circuit does not do), but the error signal still needs to be sent up
to the next level perceptual function. Rather than produce a
plethora of switches for the multiple connections up to the next
level, I suspect a rethink in terms of local and global inhibition
might be a way to address this problem.

        So the perceptual input at level n+1 is influenced not

only by a copy of the reference signal for systems atlevel
n, it is also influence by a copy of the error signal from
each system below,

From level n.

        which determines the reference signals for systems at

level n-1. Same caveat as for Bill’s diagram.

That "which" is ambiguous. My understanding is that the reference

value at level n-1 is always a function of one or (usually) more
output signals from different level n output functions (represented
by the rectangle across the reference connection in both diagrams).
That’s the same in both circuits. In simulations of multilevel
control structures, this function is usually taken to be a simple
addition, but I doubt that could be the case as we go higher in the
hierarchy. [Aside: I expect that this function would be different in
nature level by level, just as is the perceptual function. Bill
guessed that it would be an associative memory, which for me rings
true at higher levels, but it’s just a guess, whoever proposes it.]

No, but you raise an important issue about imagination that is

addressed directly by neither circuit – the counterfactual
imagination. However that issue may be resolved, the resolution
would be the same for either circuit.

One thing that is often forgotten, related to the fibre-bundle

↔ wire simplification, is that when we draw one control unit
for a particular perception (such as the location of a chair), that
single diagrammed unit probably represents a host of units operating
in parallel. So your “it has seemed necessary to be able to control
the same perception redundantly by two controllers, one in
imagination mode and one in control or observation mode” is by no
means an extension of the theory. All it suggests is that there need
not be a conflict between perceiving in imagination and from sensory
data at the same time with different values. Whether they are
compared is a question of whether there exists a relationship
perceiving unit above them both that takes input from each.
Yes, I think that sort of dual perception is quite regular. It
doesn’t happen in a simplified hierarchy of a diagram on a page with
just one control unit per type, but I think it would be natural in
the kind of proliferating hierarchy Bill envisaged.

Martin

···

[Bruce Nevin (2017.07.13.13:13 ET)]

          Martin Taylor

2017.07.12.16.34–

              BN> we control or observe a perception of the

difference between that imagined perceptual state and
our current perceptual state."

MMT> No.

              BN> The latter is also an exercise of

imagination because one term of the difference (a
non-imagined Relationship perception) is an exercise
of imagination.

            MMT> The alternative circuit can be used that way.

The Powers circuit cannot.

            MMT> I expect there are many other possible

circuits. It has just seemed to me that the original
Powers HPCT is taken too easily to be the received
truth, rather than being a largely intuitively derived
concept. The advantage of this particular suggestion is
that it is based rather closely on the Seth-Friston
circuit, for which they identify brain locations for the
different components and functions. If their suggestions
are correct, and if this proposal makes any sense, the
same mapping should apply.

        BN> Ah, yes, I jumped a cog. In my thinking about the

experience of comparing real with imagined perceptions it
has seemed necessary to be able to control the same
perception redundantly by two controllers, one in
imagination mode and one in control or observation mode,
with a third making the comparison. Are you saying that this
can be done within the same controller using your
MMT/Seth-Friston circuit?

        BN> I refer to experiences like getting a vivid

imagined flash image of an object toppling to the floor and
moving it or otherwise acting to ensure that this doesn’t
happen in fact. Visiting a place and comparing its
remembered character with its present state. Various
planning and contingency-response situations. And so on.

[bnhpct]

[Bruce Nevin (2017.07.13.13:03 ET)]

Martin Taylor 2017.07.12.16.34 –

Two parts, answered separately.

Either the diagram is wrong or you are misreading it

Here is my reading.

In the Powers diagram, the reference signal for the reference input of a specific controller below branches to join the perceptual signal sent from that controller to the perceptual input of the specific controller whose error signal is the basis of the reference signal, making a closed loop. A caveat is that this branching must comprise any amplification of the reference signal, which is generally accomplished by branching to create multiple copies going to multiple synapses at the reference input. In the Powers diagram, this branch from reference output to perceptual input is connected or disconnected to produce several modes of operation, the control, observation, and imagination modes.

In your diagram (as in mine), the branch from reference output to perceptual input is always connected. In addition, a copy of the error output of a lower system is sent back up to the perceptual input of the higher system. The perceptual signal presumably goes from the lower system to the higher as well, but that is not shown. I have assumed that this is an oversight, maybe a consequence of placing the “input=” label in the way. So as I understand it, the error signal from every comparator is sent to two places, the reference output to lower systems and the perceptual input that the comparator is passing up to higher systems.

So the perceptual input at level n+1 is influenced not only by a copy of the reference signal for systems atlevel n, it is also influence by a copy of the error signal from each system below, which determines the reference signals for systems at level n-1. Same caveat as for Bill’s diagram.

(Attachment 5.2 PowersFristonCircuit_v34.jpg is missing)

···

On Wed, Jul 12, 2017 at 5:10 PM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2017.07.12.16.34]

[Bruce Nevin (2017.07.12.08:53 ET)]

        Martin Taylor

2017.07.11.10.35–

Very interesting, Martin.

      In the Powers circuit, the reference signal at level n

branches into the perceptual input for level n+1 (when the
‘imagination switch’ is closed). Ditto in my “always on”
proposal.

      In your proposal, the reference signals at level n-1 (=

error output from level n) branch into the perceptual input
for level n+1, bypassing the perceptual input function at
level n where perceptual signals from level n-1 are combined
to construct level n perceptual signals.

Either the diagram is wrong or you are misreading it. The circuit is

intended to produce exactly the same effects as the Powers circuit
except that I have explicitly included a pair of diagrams for the
case in which there is a finite tolerance zone and the perceptual
value is within that zone. Here’s the diagram again. I do not see
the level-jumping you mention. Level n perceptual input comes from
many level n-1 “reference-error=p” units, and the output of level n
units goes similarly many-to-many to produce the level n-1 reference
values. I think (and hope) you are misreading it.

      Error output from level n branches downward to the

reference input of a plurality of controllers below. The
reference input there receives input from a plurality of error
outputs of controllers above.

That is (if by "error output" you mean the output of the control

unit’s output function") exactly as in the Powers circuit.

      I have a hunch this may have unforeseen consequences. It

may take some time to puzzle out.

/Bruce

and in a later message

        [Bruce Nevin (2017.07.12.10:23

ET)]

          Martin

Taylor 2017.07.11.10.35–

MMT> “* How can we perceive what we
want, as we consciously are able to do?”* … “* How can we perceive the
difference between where we are and where we want to be* ?” as we also are consciously
able to do. [BN: You also said you included a question
of mine by reference to your quotation from my post of 2017/07/9 6:16 PM, but I don’t
see a question in that paragraph.]

          MMT> If (as Powers claimed) any conscious perception

must be represented in the hierarchy as a perceptual
value, and if reference and error values do not exist in
the perceptual side of the hierarchy, then if the Powers
hierarchy were literally correct, both of those questions
would have to be answered with the counter-factual: “We
can’t do that.” The alternative I suggest here answers
both questions by placing reference and error separately
into the perceptual upgoing data flow, so that both are
potentially available to perceptual functions at higher
levels as separate signals, perceptible and available to
consciousness.

        But this is accommodated by any scheme that accounts for

imagination.

Not so. In the Powers circuit, if you are in imagination mode, the

upgoing perception is switched out. You get either the sense based
data provided by the Real World (and lower-level imagining) or you
get the reference value provided by imagination. This what you said
in the message of 2017/07/9 6:16 PM:

  [BN earlier] The B:CP view of imagination is

that a copy of the reference signal (here, the reference for a
missing bit of input) branches across to create a perceptual input
signal returning to the originator(s) of the reference signal.
B:CP depicts this as a switch actively making and breaking a
neural connection. This has seemed implausible to me, so I
proposed that the imagination signal is always present. When there
is actual perceptual input, the copy of the reference signal
augments it; when there is not, then the copy of the reference
signal provides some (weak) input of that perception.

The alternative circuit gives you exactly what you suggested ought

to be the case. The reference signal is always available, and is
never switched out. If you treat the “wires” in the circuit as fibre
bundles rather than wires carrying a neural current, you also can
smoothly augment the perceptual input from below from total
perception to total imagination, without affecting the controlling
done by lower-level units (other than if you control using much
imagination, your control will not be very good!

The alternative circuit has only one switch per control unit, one

that breaks the control loop, so that the unit in question does not
try to control its perception – something you don’t want to do when
you are controlling in imagination, as, for example when you are
planning some future activity.

        [BN now] The three ideas under discussion do this by

branching the reference input (or the error inputs that are
combined into a reference input) over to the perceptual
input side. By those means, a copy of reference (or error)
values does exist on the perceptual side of the hierarchy.

Yes, but in the Powers hierarchy not in a way that can be compared

with the perceptual value for which it is a reference, and in the
Powers hierarchy the error is not available either directly or by
computation.

        According to the Powers hierarchy model, then, that's

just what imagination is.

Yes, I kept that exactly the same. As I said, I tried to make the

circuit function as like the Powers circuit as I could. The point
was to show that it isn’t the only circuit that can produce the same
results. The other nice properties of the alternate circuit are just
gravy. I haven’t seen any downside yet, though I’m willing to bet
there must be some.

        So if the Powers hierarchy is  literally correct, both

of those questions can be answered "This is how we do that.
We control copies of our references. We control or observe
an imagined perception of what we want (a copy of the
reference),

Yes.

        and we control or observe a perception of the difference

between that imagined perceptual state and our current
perceptual state."

No.

        The latter is also an exercise of imagination because one

term of the difference (a non-imagined Relationship
perception) is an exercise of imagination.

The alternative circuit can be used that way. The Powers circuit

cannot.

I expect there are many other possible circuits. It has just seemed

to me that the original Powers HPCT is taken too easily to be the
received truth, rather than being a largely intuitively derived
concept. The advantage of this particular suggestion is that it is
based rather closely on the Seth-Friston circuit, for which they
identify brain locations for the different components and functions.
If their suggestions are correct, and if this proposal makes any
sense, the same mapping should apply.

Martin

[bnhpct]

[Bruce Nevin (2017.07.13.13:13 ET)]

Martin Taylor 2017.07.12.16.34–

(Attachment 5.2 PowersFristonCircuit_v35.jpg is missing)

···

On Wed, Jul 12, 2017 at 5:10 PM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2017.07.12.16.34]

[Bruce Nevin (2017.07.12.08:53 ET)]

        Martin Taylor

2017.07.11.10.35–

Very interesting, Martin.

      In the Powers circuit, the reference signal at level n

branches into the perceptual input for level n+1 (when the
‘imagination switch’ is closed). Ditto in my “always on”
proposal.

      In your proposal, the reference signals at level n-1 (=

error output from level n) branch into the perceptual input
for level n+1, bypassing the perceptual input function at
level n where perceptual signals from level n-1 are combined
to construct level n perceptual signals.

Either the diagram is wrong or you are misreading it. The circuit is

intended to produce exactly the same effects as the Powers circuit
except that I have explicitly included a pair of diagrams for the
case in which there is a finite tolerance zone and the perceptual
value is within that zone. Here’s the diagram again. I do not see
the level-jumping you mention. Level n perceptual input comes from
many level n-1 “reference-error=p” units, and the output of level n
units goes similarly many-to-many to produce the level n-1 reference
values. I think (and hope) you are misreading it.

      Error output from level n branches downward to the

reference input of a plurality of controllers below. The
reference input there receives input from a plurality of error
outputs of controllers above.

That is (if by "error output" you mean the output of the control

unit’s output function") exactly as in the Powers circuit.

      I have a hunch this may have unforeseen consequences. It

may take some time to puzzle out.

/Bruce

and in a later message

        [Bruce Nevin (2017.07.12.10:23

ET)]

          Martin

Taylor 2017.07.11.10.35–

MMT> “* How can we perceive what we
want, as we consciously are able to do?”* … “* How can we perceive the
difference between where we are and where we want to be* ?” as we also are consciously
able to do. [BN: You also said you included a question
of mine by reference to your quotation from my post of 2017/07/9 6:16 PM, but I don’t
see a question in that paragraph.]

          MMT> If (as Powers claimed) any conscious perception

must be represented in the hierarchy as a perceptual
value, and if reference and error values do not exist in
the perceptual side of the hierarchy, then if the Powers
hierarchy were literally correct, both of those questions
would have to be answered with the counter-factual: “We
can’t do that.” The alternative I suggest here answers
both questions by placing reference and error separately
into the perceptual upgoing data flow, so that both are
potentially available to perceptual functions at higher
levels as separate signals, perceptible and available to
consciousness.

        But this is accommodated by any scheme that accounts for

imagination.

Not so. In the Powers circuit, if you are in imagination mode, the

upgoing perception is switched out. You get either the sense based
data provided by the Real World (and lower-level imagining) or you
get the reference value provided by imagination. This what you said
in the message of 2017/07/9 6:16 PM:

  [BN earlier] The B:CP view of imagination is

that a copy of the reference signal (here, the reference for a
missing bit of input) branches across to create a perceptual input
signal returning to the originator(s) of the reference signal.
B:CP depicts this as a switch actively making and breaking a
neural connection. This has seemed implausible to me, so I
proposed that the imagination signal is always present. When there
is actual perceptual input, the copy of the reference signal
augments it; when there is not, then the copy of the reference
signal provides some (weak) input of that perception.

The alternative circuit gives you exactly what you suggested ought

to be the case. The reference signal is always available, and is
never switched out. If you treat the “wires” in the circuit as fibre
bundles rather than wires carrying a neural current, you also can
smoothly augment the perceptual input from below from total
perception to total imagination, without affecting the controlling
done by lower-level units (other than if you control using much
imagination, your control will not be very good!

The alternative circuit has only one switch per control unit, one

that breaks the control loop, so that the unit in question does not
try to control its perception – something you don’t want to do when
you are controlling in imagination, as, for example when you are
planning some future activity.

        [BN now] The three ideas under discussion do this by

branching the reference input (or the error inputs that are
combined into a reference input) over to the perceptual
input side. By those means, a copy of reference (or error)
values does exist on the perceptual side of the hierarchy.

Yes, but in the Powers hierarchy not in a way that can be compared

with the perceptual value for which it is a reference, and in the
Powers hierarchy the error is not available either directly or by
computation.

        According to the Powers hierarchy model, then, that's

just what imagination is.

Yes, I kept that exactly the same. As I said, I tried to make the

circuit function as like the Powers circuit as I could. The point
was to show that it isn’t the only circuit that can produce the same
results. The other nice properties of the alternate circuit are just
gravy. I haven’t seen any downside yet, though I’m willing to bet
there must be some.

        So if the Powers hierarchy is  literally correct, both

of those questions can be answered "This is how we do that.
We control copies of our references. We control or observe
an imagined perception of what we want (a copy of the
reference),

Yes.

        and we control or observe a perception of the difference

between that imagined perceptual state and our current
perceptual state."

No.

        The latter is also an exercise of imagination because one

term of the difference (a non-imagined Relationship
perception) is an exercise of imagination.

The alternative circuit can be used that way. The Powers circuit

cannot.

I expect there are many other possible circuits. It has just seemed

to me that the original Powers HPCT is taken too easily to be the
received truth, rather than being a largely intuitively derived
concept. The advantage of this particular suggestion is that it is
based rather closely on the Seth-Friston circuit, for which they
identify brain locations for the different components and functions.
If their suggestions are correct, and if this proposal makes any
sense, the same mapping should apply.

Martin

[MartinT]

[Martin Taylor 2017.07.12.16.34]

Either the diagram is wrong or you are misreading it. The circuit is

intended to produce exactly the same effects as the Powers circuit
except that I have explicitly included a pair of diagrams for the
case in which there is a finite tolerance zone and the perceptual
value is within that zone. Here’s the diagram again. I do not see
the level-jumping you mention. Level n perceptual input comes from
many level n-1 “reference-error=p” units, and the output of level n
units goes similarly many-to-many to produce the level n-1 reference
values. I think (and hope) you are misreading it.
That is (if by “error output” you mean the output of the control
unit’s output function") exactly as in the Powers circuit.
and in a later message
Not so. In the Powers circuit, if you are in imagination mode, the
upgoing perception is switched out. You get either the sense based
data provided by the Real World (and lower-level imagining) or you
get the reference value provided by imagination. This what you said
in the message of The alternative circuit gives you exactly what you suggested ought
to be the case. The reference signal is always available, and is
never switched out. If you treat the “wires” in the circuit as fibre
bundles rather than wires carrying a neural current, you also can
smoothly augment the perceptual input from below from total
perception to total imagination, without affecting the controlling
done by lower-level units (other than if you control using much
imagination, your control will not be very good! The alternative circuit has only one switch per control unit, one
that breaks the control loop, so that the unit in question does not
try to control its perception – something you don’t want to do when
you are controlling in imagination, as, for example when you are
planning some future activity.
Yes, but in the Powers hierarchy not in a way that can be compared
with the perceptual value for which it is a reference, and in the
Powers hierarchy the error is not available either directly or by
computation.
Yes, I kept that exactly the same. As I said, I tried to make the
circuit function as like the Powers circuit as I could. The point
was to show that it isn’t the only circuit that can produce the same
results. The other nice properties of the alternate circuit are just
gravy. I haven’t seen any downside yet, though I’m willing to bet
there must be some.
Yes.
No.
The alternative circuit can be used that way. The Powers circuit
cannot.
I expect there are many other possible circuits. It has just seemed
to me that the original Powers HPCT is taken too easily to be the
received truth, rather than being a largely intuitively derived
concept. The advantage of this particular suggestion is that it is
based rather closely on the Seth-Friston circuit, for which they
identify brain locations for the different components and functions.
If their suggestions are correct, and if this proposal makes any
sense, the same mapping should apply.
Martin

(Attachment 5.2 PowersFristonCircuit_v33.jpg is missing)

···

[Bruce Nevin (2017.07.12.08:53 ET)]

        Martin Taylor

2017.07.11.10.35–

Very interesting, Martin.

      In the Powers circuit, the reference signal at level n

branches into the perceptual input for level n+1 (when the
‘imagination switch’ is closed). Ditto in my “always on”
proposal.

      In your proposal, the reference signals at level n-1 (=

error output from level n) branch into the perceptual input
for level n+1, bypassing the perceptual input function at
level n where perceptual signals from level n-1 are combined
to construct level n perceptual signals.

      Error output from level n branches downward to the

reference input of a plurality of controllers below. The
reference input there receives input from a plurality of error
outputs of controllers above.

      I have a hunch this may have unforeseen consequences. It

may take some time to puzzle out.

/Bruce

2017/07/9 6:16 PM:

  [BN earlier] The B:CP view of imagination is

that a copy of the reference signal (here, the reference for a
missing bit of input) branches across to create a perceptual input
signal returning to the originator(s) of the reference signal.
B:CP depicts this as a switch actively making and breaking a
neural connection. This has seemed implausible to me, so I
proposed that the imagination signal is always present. When there
is actual perceptual input, the copy of the reference signal
augments it; when there is not, then the copy of the reference
signal provides some (weak) input of that perception.

        [Bruce Nevin (2017.07.12.10:23

ET)]

          Martin

Taylor 2017.07.11.10.35–

MMT> “* How can we perceive what we
want, as we consciously are able to do?”* … “* How can we perceive the
difference between where we are and where we want to be* ?” as we also are consciously
able to do. [BN: You also said you included a question
of mine by reference to your quotation from my post of 2017/07/9 6:16 PM, but I don’t
see a question in that paragraph.]

          MMT> If (as Powers claimed) any conscious perception

must be represented in the hierarchy as a perceptual
value, and if reference and error values do not exist in
the perceptual side of the hierarchy, then if the Powers
hierarchy were literally correct, both of those questions
would have to be answered with the counter-factual: “We
can’t do that.” The alternative I suggest here answers
both questions by placing reference and error separately
into the perceptual upgoing data flow, so that both are
potentially available to perceptual functions at higher
levels as separate signals, perceptible and available to
consciousness.

        But this is accommodated by any scheme that accounts for

imagination.

        [BN now] The three ideas under discussion do this by

branching the reference input (or the error inputs that are
combined into a reference input) over to the perceptual
input side. By those means, a copy of reference (or error)
values does exist on the perceptual side of the hierarchy.

        According to the Powers hierarchy model, then, that's

just what imagination is.

        So if the Powers hierarchy is  literally correct, both

of those questions can be answered "This is how we do that.
We control copies of our references. We control or observe
an imagined perception of what we want (a copy of the
reference),

        and we control or observe a perception of the difference

between that imagined perceptual state and our current
perceptual state."

        The latter is also an exercise of imagination because one

term of the difference (a non-imagined Relationship
perception) is an exercise of imagination.