ch7- Evolution of the Central Nervous System
1- Understanding Biotic Neural Networks
1.1- Instinctive functionality of neural networks - stimulus/response
1.2- Learned functionality of neural networks - heuristics
2- Comparative Anatomy of Biological Central Nervous Systems
The central nervous system consists of the spinal
chord
(notochord in spineless animals) and the brain. It ties into the
sensory nerves (touch, taste, smell, sight, sound.), the motor nerves (muscles),
and the glandular nerves (hormone producers), among others.
Effectively, the entire system is an integrated neural network. We are
interested in the evolution of the central nervous system and specifically the
functional (evolutionary) anatomy of the brain, which will segue into the next
section: evolution of knowledge.
1-Understanding Biotic Neural Networks
First
let's start with some neural network basics and build
some
logic circuits to see how the networks might work and specifically how
they might lend themselves to evolutionary growth. This exercise is not meant to be a study of neuro-physiology.1.1- Instinctive functionality of neural networks - stimulus/response
Single cells demonstrate
stimulus/response behavior. We can view cell functions as
internally directed, e.g. replication, metabolism, structural,
transport, intra-cell signaling, defense and immunity. Or, cell
functions can be externally directed, e.g. motor/sensor, cell-cell
signaling. So the stimulus would be that of external functions which
will then garner a response triggered via internal functions. Say the
stimulus would be an external input (food chemical)
triggering the sensory functions, and next internal functions
(intra-cell signaling) which would in turn trigger motor
functions (response), propelling the movement of the cell towards the
food source.
Now
take the same stimulus-response behavior
of a single cellular cluster and spread it to two differentiated cell
clusters, half handle
stimulus (sensory neurons), the rest handle the response (motor cells).
In
this drawing the
green cells are sensory neurons and the
red ones
are motor neurons. For simplicity I have 4 of each, representing the
degrees of freedom for a hypothetical animal that moves in two
dimensions. Here, stimulus cells, e.g. detecting food chemicals in
a given direction, fire the motor cells covering that direction. So
you'd end up with movement in that
direction, if a stimulus (e.g. food) is detected. On the other hand, if
the chemical stimulus is of something that is going to eat this
hypothetical animal, then the sensors fire the "get away response" and
you'll have the motor cells firing in the opopsite direction. So even
at this most basic state you have a logic to be executed, if the
stimulus if food, motor cells fire towards, if we are food, fire away
from it. |
|
In
figure 2, I'm trying to demonstrate associative logic. Here you
have
two potential stimuli, A and B; say A represents food (effecting the
green cells) and B represents smell (effecting the blue cells). I've
added another layer of control neurons (black) to complete the logic
circuit. The control neuron, now can make logical decisions, given the
nature of its connection with the stimulus neurons A and B, conditionally firing motor cells, depicted in red. |
|
Here
we have added another layer of complexity to the executable logic for
the control cell cluster. If
A cells detect food and B cells smell food then fire the
motor cell towards ("AND" Logic: fire towards if A+ AND B+). If B is
smells good stuff, and A is dormant, then the decision is to fire
towards and take a closer look at A ( vice versa, "OR" Logic: fire
towards if A+ OR B+). If A cell detects food and B
cells smells chemical danger (enemy!), then the control logic would
fire the motor neuron away from the target ( A+ AND B-). The converse
will hold as well (A- AND B+). Or the control logic could do nothing
with respect to the target (NOT logic, NOT A and NOT B). So we can
assemble a full logical decision profile (AND, OR, NOT) for firing
towards, or away, or do nothing, depending on the makeup of the
stimulus
neurons A, B.
In
this third diagram, I've added a yellow internal censor, say
sensing hunger. Given all of the above logic with A and B, only fire
towards the target if the hunger
cell is turned on. So you have an additional logical control element
that brings in the internal state of the organism in the logical
profile, effecting the "fire towards" decision. |
|
So,
even in this most primitive stage, with only two sensory organs
detecting taste and smell, and one detecting the internal hunger state,
you have a good amount of logical execution. Please note that all of
the behavioral manifestations of these examples are instinctive,
pre-wired, depending on the genetic impetus of agreeable or
disagreeable.
Thus
far we have talked about internal/external functional integration. As
pertinent as that might be, it is noteworthy that a great deal the work
of neural nets involves hierarchical control of internal organs, e.g.
heart, lung, glands, etc. (especially as the morphological complexity
grows). All of that is pre-wired and for the most part not subject to
conscious control.
1.2- Learned functionality of neural networks - heuristics
Now
lets talk about memory, where training is required. Lets take the above
example and replace the smell stimulus, which is instinctive
(pre-wired), with sound stimulus, which requires the training of memory
circuits for what would become an agreeable or disagreeable sound
stimulus.
Pavlov's dogs put that association logic
on the map. Say B (sound) by itself is a benign stimulus, but it is
consistently present when A (food) is present (sounds like food?), then
the logic cells would get trained over time and fire if B (sound) is
present, even if A (the food) isn't present. In this example sounds form a pattern. And in
normal life Pavlov's dogs would have learned to differentiate arbitrary
background chatter from their owner calling them with something like
"Spot, Food!". Neural networks naturally form memory circuits as the
following drawings demonstrate:
| The dogs'
hearing memory circuits get trained over time to the sound
"Spot, Food!" and a pattern corresponding to the sound is established
in the network (visualize it as dynamic color patterns, with colors
representing strength of connections- above figure). Please note that training of
e.g. sound memory clusters requires affirmation by the instinctive
stimulus neurons, e.g. taste. It is the instinctive neurons that signal
agreeable and/or disagreeable. So in figure 2 and 3 (above), you can plug in the sound memory network for the smell
stimulus (blue)
cell clusters and you'll have the memory of an event executing the
same types of associative logic as in the example. But the feed also
has to go backwards from the control neural cluster to the memory
clusters, imprinting it as an agreeable or disagreeable event (figure, right). In that
sense the sound memory cluster could then behave just like smell
stimulus neural cluster (B) firing a B+ (agreeable) or B-
(disagreeable) stimulus. | 
|
Memory
circuits record events. You have Inherent Reality presenting an event
involving external entities or phenomenon; sensory organs (sound, sight,
smell, touch, taste) pick up the event and record them in associated
memory circuits, affirmed as agreeable/disagreeable or benign, thereby
internalizing the Perceived Reality of the event.
| Now
suppose you have a memory network cluster tied
into the
sound stimulus neurons and another memory cluster tied into the
sight stimulus neurons. You can execute associative control logic by
having an integrator/arbitrator neural cluster between the sight and
sound memory clusters, as the figure (right) shows (memory
clusters in gray, brown, integrator cluster in
magenta, control cluster in black, all two way connections,
affirmative cluster in red, one way connection). So you can integrate
the sight and sound information and
based on learned (affirmed) memories execute logical action. Say Pavlov
trained
his dogs to the sound of "Spot Food" as well as the Red color of their
food bowl. Then the response would come from the integrator/arbitrator
neural cluster, integrating/arbitrating sound and sight memories and
triggering the control neural cluster (or not), depending on the
outcome of the arbitration. |  |
If
you think about it, the complexity of the executable logical profile of
such a network would be rather remarkable. First, you have to consider
that now you have 3 input sources: A= taste (affirming input), B= sound
memory, C= site memory. And these inputs have a strength (intensity)
ranging from A+ to A-, B+ to B- and C+ to C-. And the affirmation of
the training is a continual process, not a one shot deal. So the memory
has to be continually affirmed and the integration/arbitration has to
be continually affirmed as well. If you put all of that together, you
would not expect a deterministic behavior. Rather, you would expect trial and error, give up and try again, hunt and peck,
otherwise rather chaotic behavior to establish agreeable outcome in a
consistent manner. And if we consider the behavior of animals, from
insects, to dogs to people, that is exactly what we see. The bottom
line is that the heuristic process is inherently chaotic, the learning
process may or may not lead to a stable result set. Fortunately we can
take the heuristic network design quite a bit further.
Suppose
action itself could be remembered. That is, not only the animal is to
try an action based on taste, sound and site inputs, but it has a
benchmark of what it has tried before and the resultant affirmation of
the outcome (figure, right). Then it wouldn't have to hunt and peck so
much, it could refine the actions that it has tried before. Then trial
and error isn't continually random. Given enough time, repetition and
luck, agreeable outcomes could be honed on systematically. And once a
set of actions are perfected, then they can be repeated over and over
with relatively minor alterations. And that is exactly what we see in
heuristic networks and resultant behavior in higher order animals.
But
suppose we take this a step further. What if memorized actions and
their resultant outcomes where to be compared. That is the logic of the
actions could be affirmed across the spectra of actions. Then there
would be an internal memory of logic. I don't know about other animals,
to be honest, but we certainly have that. We can remember why and how
we did what we did with respect to a given circumstance that turned out
well and use that as a benchmark to try it with respect to other
similar circumstances, i.e. preform deductive reasoning. | 
|
So
heuristic networks are inherently chaotic. In simpler networks
(animals) all we see is the continual hunt and peck that is the
manifestation of instability of these chaotic networks. But heuristic
networks are particularly amenable to evolutionary aggregation. That
is, animals that have evolved the additional neural networks to embody
memory of actions, can hone in on continually more agreeable outcomes,
and subsequently stabilize that behavioral processes. We call that
perfecting the learned behavior. Moreover, if additional network layers
can be put to use to remember the logic that gave rise to the actions,
then the logical processes themselves can be honed. And if you have a
set of well honed logical processes committed to memory then
commonalities will emerge. Then these common, well affirmed, logical
decisions could be honed. That in essence is the process of evolution
of knowledge.
We are just about done, but before we move on I need to make a couple of other points.
Memories
that are occasionally affirmed form short term memory. Those are
manifested as internal connection strengths of the neurons in the
memory circuits. However, if an action is continually affirmed, it ends
up as permanently imprinting on the memory circuit, by the way of
switching on a "long term memory gene" in the neurons of the memory
circuit, thereby establishing long term memory. We need this to establish the next point.
These
neural networks form a hierarchy of distinct, yet associated control
networks in the brain. To make this point, I'll give an example of
driving a stick-shift car. Many may not remember, but before automatic
transmissions you had to shift gears with a stick shift. Think about
the logical controls that go into that operation. (1) Your right foot
has to let off the gas petal. (2) Next, your left foot has to press
down the clutch. (3) Then your right hand has to shift the stick from
one correct gear to the next. (4) Then your left foot is to lift off
the clutch, and (5) your right foot has to step back on the gas. While
all of that is going on (6) your left hand has to do the driving,
corresponding to the (7) sight information that is to (8) lead you to
your destination and hopefully avoid crashing the car in the process.
And you have to continually hone into sound information coming from the
car, telling you whether to shift gears or not. So all of these
controlled actions are distinct yet associated, some in succession,
others simultaneous. Further, they are hierarchical in the context of
getting you to your destination.
One
last point on this
subject, heuristic behavior, when repeated and affirmed adinfinitum can
appear as instinctive as pre-wired behavior. Why? Because of the
switching on of the long term memory gene in the memory neurons. In
fact, as daunting as
driving a stick-shift car may be for a while, after years of doing the
controlled actions, it becomes instinctive. One does not think about
it, is not even conscious of it, he just does it. In this way, memory
can turn in to instinct. Memory that is affirmed to the point of
becoming literally instinctive (genetically driven) can itself serve as
the affirming factor in decisions of a control circuit.
Going
back to the start of this exercise, we had the basic sensory neurons
for taste and smell to serve as the affirming neurons for the memory
circuits. By this point in the exercise we see how memory circuits
themselves can be turned into instinctive affirmation factors. Once you
bring in internal action memory and internal logic memory into this
picture, that is when action and logic memories are affirmed to the
point of becoming instinctive, then you have the basis for building a
hierarchy of abstractions and controlled actions that we humans exhibit
with regularity.
Ok, lets sum up.
1-
Functionally, biotic neural networks are the result of the evolution of
stimulus-response functions of cells. In the first order of
morphological hierarchy, cell-cell communication functionality gives
impetus to the evolution of pre-wired, instinctive, control networks.
2- Biotic neural networks naturally form memory clusters, holding memory of
stimuli and when affirmed the memories can form inputs to the control
networks.
3- Memory based control networks lend to heuristic behavior, that is self (autonomous) learning.
4- Neural networks naturally do fuzzy logic. The dog would generate
likely similar response to "Spo, Food" and would probably ignore "Sa,
Fa", i.e. one is close to the trained patterns and the other is far
from it.
5- Neural network clusters can be strung together hierarchically and in parallel to carry out complex associative logic.
6-Actions
and logic itself can be committed to memory. As a result learning can
be progressively perfected. And as a corollary, such networks can do
deductive logic as well, i.e. infer a solution. For
example a monkey sees that hitting something can break it. Then he sees
a broken nutshell and eats the nut. Then he sees an unbroken nutshell
and deduces that hitting it with a rock will break it, making the nut
accessible (all manners of animals do problem solving). And by the
virtue of having internal memory of the behavior itself, if a deductive
reasoning set is effective in one situation, it can evoke deductive
reasoning in other situations, i.e. the reasoning process itself
becomes a memory pattern.
7- Neural Networks in general, and memory and control networks
specifically, easily lend themselves to evolutionary growth in
complexity (associative and deductive). You just need more of them, and a lot of appropriate
training, to do more logic.
8-
Inherent Reality, the domain of abstract entities and phenomenon,
presents inputs to these networks. Higher order networks, by the virtue
of memory of action and logic, can form abstractions of these inputs as well as their own actions and
logic. These abstractions themselves are imprinted on memory, forming
Perceived Reality. These imprints are inherently
incomplete, and subject to growth, and at least in us humans, the brain
by itself attempts
closure. For us these attempts come off as mental simulations, day or
night dreams, learning through trial and error, or learning from
others, or just
plainly making up stuff if all else fails. That is the subject for the
next chapter, evolution of knowledge.
2- Comparative Anatomy of Biological Central Nervous Systems
Now lets look at some animals in order of complexity to see
the
evolution of the nervous system at work. The earliest neural nets we
see are in the earliest morphologies. Here are drawings of a sea
anemone and a sea star. What you see is a simple stimulus-response
network without much control circuitry in between. There is no brain
formation (source:
http://www.emc.maricopa.edu/faculty/farabee/BIOBK/BioBookNERV.html).
The
very early development of a central nervous system can be seen
in arthropods (insects, crustaceans). In this drawing the nervous
system is drawn in blue, the circulatory system in yellow and the
digestive system in green (source:
http://www.cals.ncsu.edu/course/ent425/tutorial/nerves.html#2). |
|
This drawing
focuses on the central nervous system of the
arthropod.
In the upper part of the picture you see the formation of a primitive
brain in 3 round pairs of ganglia (lobes), the forebrain, midbrain and
the hindbrain. The forebrain (Protocerebrum) is largely associated with
vision. The mid-brain (Deutocerebrum) processes sensory information
collected by the antennae. The hind-brain (Tritocerebrum) and the
ganglia below it innervate the mouth parts and integrate sensory inputs
from forebrain and mid-brain. It also links the brain with the rest of
the ventral nerve cord and the visceral and the peripheral systems,
i.e. the wiring of the internal organs and limbs.
Even at this early stage neural nets can generate complex
behavior. Honey bees, for example, can discern and communicate
direction and distance of the food source. |
|
In
the next slide, I'll show you a developing
zebrafish's
nervous system. You essentially see the same fore, mid and hindbrain
structures:
In the next set of
slides we'll look at fish, reptilian,
amphibian, and bird brains. You'll see the evolution of cerebrum and
the optic lobe from the elemental forebrain, and the cerebellum and
medulla oblongata from the hind-brain. Notice the relative increase in
the size of the cerebrum (source:
http://k-2.stanford.edu/InfoPackets/2-BioSys.7.0.html).
Fish
|
 Amphibian
|
Reptile
|
Bird
|
Next lets look at the
brains of a number of mammals to compare
the relative sizes.
And
finally we'll see a human brain and the
functional areas:
By
a comparative view of the nervous systems of animals in a hierarchical
order, we have traversed stimulus-response domain of behavior to
heuristic domain of behavior. We have seen that in fact the evolution
of the central nervous system and the brain, abstraction and problem
solving, is the impetus of survival, a consistent
evolutionary impetus for of heuristic behavior. The animals that evolve
in that direction get
rewarded and can have a competitive edge. So we see abstraction and
problem solving fairly consistently among various animals in very
different branches of the tree of life, honey bee, octopus, parrot,
raccoon, dolphin, chimp, and man.
Chapter Key:
Morphological Flows, entities going through functional constructs
thereby creating more complex entities with more complex
functionalities:
Cell Stimulus-Response == cell differentiation ==> sensory and motor neurons
sensory neurons +motor neurons +controller neurons == life actions ==> instinctive behavior
Sensory neurons + neural memory circuits +affirming neurons == training ==>
learning
Sensory neurons + memory circuits + integrator/arbitrators + controllers + motor cells == life actions ==> heuristic behavior
Inherent
Reality [abstract entities and phenomenon] + heuristic networks ==
imprinting ==> Perceived Reality [abstractions and an internal
notion of what is physical]
Links:
The
nervous system - good introductory site
Arthropod
nervous system
Zebra-fish
development, including the nervous system
Comparative
brains
Functional
brain anatomy
Cranial
nerves
