[From Bjorn Simonsen (2005.03.18, 10:55 EST)]
Martin Taylor 2005.03.17.10.27
In that "quasi-real" world, the dynamics may be faster (I can imagine
a multi-hour flight in a few seconds), but the consequences of my
imagined actions on my imagined perceptions have to be reasonably >
like what would happen in the real world. There must be gain around
imagined perceptual loops, whether the match to the real world is
close or completely wrong. But the effects are caused differently,
not by time delays but by forced (imagined) mapping of the temporal
effects of real-world delays onto the rapid imagined feedback.
I don't know if I understand what you have written in the same way you
understand it yourself. It looks to me as if you are both talking about gain
and time delay. What I write below is the way I think. And I appreciate
comments if I think wrong relative to you and other who know PCT.
When you talk about gain and time delay you talk about concepts in a
simulation. In the brain there is no gains and time delays. The brain
function as it does. And we know that a tendon reflex happens quicker than
if you are asked to point out a green card with two circle figures and a
cube inside a double lined frame. We know we control lower levels quicker
than we control higher levels.
The way I understand the gain of a system is as follows.
The effectiveness of a control system depends on its loop gain. If we make a
simulation of behavior we have to use a gain that makes the simulation
effective. If the gain is too little the simulation will be unstable. If we
increase the gain, we achieve a tighter control.
The way I understand the time delay of a system is as follows.
Higher levels are informed with perceptual signals from lower levels.
Therefore we have to put in a time delay at higher levels (all levels). The
time delay may have a value 0.001 or 0.0001 or ..... . When we put in a time
delay like 0.001 it has an effect on the qo. When we put in a time delay
like 0.0001 it has a less effect on the qo. When it has a less effect on qo,
the system needs longer time before it is effective.
But we don't talk about gains and time delays when we talk about the brain.
The brain is normally effective.
When we make an effective simulating model of how we control our
perceptions, we can say we have a knowledge about how the brain may
function.
We have a wonderful model telling us how HPCT (the brain) functions at the
lowest levels. And the way it tells us how it functions at the higher
levels, I think "is the best there is".
When we look at the imagination mode, Rick tells us there is no g. I
understand him as though there is no function describing the physical
connection from qo to qi. qo = qi = r. But then there is a function, qo = qi
(=r). As Rick says: "we perceive what we want".
But I guess we can imagine on more levels at the same time. I guess we can
have a simulation loop on three levels where the highest level send output
signals to the lowest level and where the lowest level send perceptual
signals to the highest level. Here I think we have to put in a gain and a
time delay to get an effective system. Maybe the most effective variable is
the time delay. And we have to make it huge relative to 0.0001.
If we make it huge, the system is effective and we can imagine a flight from
A to B in a second.
I appreciated your musings.
bjorn