brain for AI
rs...@nycap.rr.com
machine that we know of that works. Some say that Artificial
Intelligence should base its investigations on the brain, should
attempt to emulate the brain. It seems reasonable that we should at
least look at the brain before we try to do so.
The brain is wholly a creature of the genome. The genome establishes a
structure that can accommodate itself to the impinging universe. The
genome does this because it survived. There is no other reason. The
genome has lived since the first cell.
The essential brain lies in the motor program generators. These are
the neural circuits established by the genome to feed and water the
organism in an expected, but not particularly friendly, universe. All
motor acts above the simplest reflex actions originate in the motor
program generators. When we say all, we mean ALL. Without a generator
in action, the organism is nothing but a lump of clay. It is common to
speak of learning an activity, but this is a mistaken view. Learning
is actually modifying. The genome constructs the circuitry that
produces walking; the circuitry may be modified to dance. Without the
walking circuitry as a base, there is nothing.
The generators are centered in the spinal cord and the hindbrain.
There are centers for breathing, posture, locomotion, orienting,
reaching, grasping, manipulating, licking, chewing, swallowing, and
vocalization. This circuitry is highly conserved. It existed before
the forebrain. The notion of a center is a simplification. The
circuitry that leads to the generator, and the circuitry that leads
from the generator to the motor neurons is also set up by the genome.
In humans, completion of the neural circuitry is delayed in some cases
until it is no longer obviously of genomic origin. This is
particularly the case with vocalization. The vocalization circuitry is
first activated at about one year when we hear the infant sounding
individual phonemes. The pre-generator circuitry is modified to
produce words and sentences. The post-generator circuitry is modified
to produce an accent.
The penultimate step of a motor program is to pass through the
thalamus on its way to the motor cortex. Here it comes under the
scrutiny of the thalamic reticular nucleus. The TRN is the repository
of bad outcomes. Every time a motor program is activated under
particular conditions and there is a bad outcome, that result is
recorded in the synapses of the TRN. If the present motor program and
the present conditions match that record, the synapses are activated,
the TRN is turned on, and the motor program is halted. A halted motor
program simply disappears.
Sensory input can also be halted by the thalamic reticular nucleus.
When the neocortex is released from the tyranny of the rush of sensory
input, it is enabled to freely associate. This is dreaming,
associating, reverie.
A succession of halted motor programs, followed by a motor program
that is not halted, is called thinking and deciding.
Ray
casey
>
> Learning is actually modifying. The genome constructs the
> circuitry that produces walking; the circuitry may be
> modified to dance. Without the walking circuitry as a base,
> there is nothing.
This seems to be in accord with the book I recently read
by Eric Kandel "In Search of Memory". Essentially there is
a default set of wiring that enables the animal to act in
a meaningful way in a general environment and yet be able
to adapt its wiring for a particular environment.
Learning selects among a large repertoire of *preexisting
connections* and alters the strength of those connections.
A rat brain starts with more default circuits to be modified
than an Aplysia brain. Rats for example can form maps and
humans can form episodic memories. Of course the actual maps
or episodic memories depends on experience (learning).
John
Don Stockbauer
Has anyone here heard of this "Global Brain" thingy?
rs...@nycap.rr.com
the thalamus like a blanket. All the axons that pass from the thalamus
to the cortex, and from the cortex to the thalamus must pass through
this nucleus. On the way, they give off branches that innervate the
TRN. The output of the thalamic reticular nucleus is wholly
inhibitory. It causes nothing to happen, it can only stop things from
happening.
Every time a bad outcome occurs, the present motor program, together
with the present sensory condition is recorded in the synapses of the
thalamic reticular nucleus. The TRN is the great repository of the
failures in the life history of the organism. When a motor program and
the current condition of the world match the recorded bad outcome, the
motor program is halted and dissipates.
The life history of the organism is thus seen as a history of avoided
bad outcomes.
Motor programs from the basal ganglia and the cerebellum enter the
thalamus through the VA-VL complex (the ventral anterior and ventral
lateral nuclei). Sensory energy enters through the lateral geniculate
body (visual-eye), the medial geniculate body (auditory-ear), and the
ventral posterolateral-ventral posteromedial nuclei (somatic-body).
Even as motor programs can be intercepted by the thalamic reticular
nucleus, so can sensory energy be intercepted by the TRN. Sensory
energy is periodically interrupted for milliseconds to allow for extra
neural activity in the neocortex as the organism proceeds. In times of
great danger, the TRN is inhibited and sensory energy flows without
pause. We experience this as, “time slows down”.
Sensory energy can be halted for extended periods. If accompanied by
sleep, this is called dreaming. If the organism is awake, it is called
day-dreaming. reverie, even thinking.
Ray
rs...@nycap.rr.com
>
> Has anyone here heard of this "Global Brain" thingy?- Hide quoted text -
>
A most interesting concept, but not particularly amenable to scientic
investigation. My own interest lies in the organism, abroad in the
world, driven by a neural system.
Ray
Don Stockbauer
You're very correct; it's not particularly amenable to scientific
investigation. However, a few people do deliriously believe in it and
promote it, for they feel that the payoff is beyond all expectations,
since self-awareness creates sentience and intelligence.
N
that'll be the system
N
identity and reasons for existance perhaps? etc lol
good easy listening module
http://britac.studyserve.com/home/Lecture.asp?ContentContainerID=132
more selections here
http://britac.studyserve.com/home/default.asp
Don Stockbauer
> well the AI brain stuff will need filling with artificial facts like,
> identity and reasons for existance perhaps? etc lol
Those will be emergent phenomena.
casey
> ...
> Every time a bad outcome occurs, the present motor program, together
> with the present sensory condition is recorded in the synapses of the
> thalamic reticular nucleus. The TRN is the great repository of the
> failures in the life history of the organism. When a motor program and
> the current condition of the world match the recorded bad outcome, the
> motor program is halted and dissipates.
It seems to me none of what you have written is useful in terms
of building any kind brain for AI?
JC
casey
The answer to everything, it will all emerge. Do you have anything
new to say about how social systems self organize?
JC
N
social systems that self organise? how to 'stack' robots...that kind
of thing? or more like how they function together? or self repairing,
self sustaining systems like programmes perhaps? as far as I
understand, anything that prevents repair from damage incidental or
deliberate harm is inadequate, thats if the object of the machinery is
to fulfill some kind of special purposes.
A car is a car...a box on wheels, its purpose is to transport. The
damage to its surface texture isnt necessarily harmful to its purpose,
although damage to the surface of 'some' machines such as to shields
and to armour plate is. This would be like a imagine...some highly
specilized design? so self repair of some surfaces is really
important.
And then say that the purpose of the machine is to run a programme,
and the transportation is essential, but if it doesnt get to its
destination, the programme inside can still function and keeps
running. Then the programme will have to 'self repair' or consider
alternative ways to complete its routine.
An artificial social systems could function like...either high
individuation where each robot is self sustaining or specialized, only
occassionally going in for a recharge for example 90% independent, or
has a high dependency or totally shared and can never function without
continued checks, 10% independency.
Adaptability is the other thing that comes into my mind, large mutual
dependent systems run more slowly, although I imagine have greater
force.
Don Stockbauer
Yes, I do, but the self-organization of social systems is at such a
high level it's unverifiable by myself, only verifiable by the social
systems themselves, and any such speculation only serves to further
playing little games here in Usenet.
J.A. Legris
You're getting warmer.
Self-organization, -assembly, -maintenance and -regeneration are so
basic to living organisms that they can be taken as defining
attributes - an artificial system exhibiting analogous competence
would surely qualify as "intelligent". Deep and thorough dependence on
chemistry and physics, i.e. properties of materials, is implicit,
which suggests that the notion of a purely computational
"intelligence" is a contradiction in terms.
--
Joe
N
>
> - Show quoted text -
any problems with your old mysterious 'stalking donkey' Joe? and does
it retrieve what it wants?
casey
>
>
> ... the self-organization of social systems is at such a
> high level it's unverifiable by myself, only verifiable
> by the social systems themselves, and any such speculation
> only serves to further playing little games here in Usenet.
I would dispute the idea that current social systems are at
such a high level we cannot understand them. Sure the details
at any point may be beyond us but the details may not matter
any more than the details of the actions of any molecule
matter in understanding the behavior of a gas.
It is an assumption that you cannot understand the machinery
behind the self organization in a social system. A simple
social system can be found in crowd behavior and that can be
understood and controlled even if the crowd itself cannot
represent such understanding of itself. In other words the
crowd isn't as smart as the parts that make it up and it is
certainly not more "complex".
I wouldn't jump to the conclusion that a social system is
"brain like" either. It may be very complex and yet not have
the ability of self verification. Only in humans for example
have neural systems reached the point where they are capable
of understanding that they are neural systems.
Your views about not being able to understand social systems
seems to be like the views held by religious people who
believed they could not understand the works of God. Science
has thrown doubt on such beliefs.
JC
Don Stockbauer
Well, whatever. Nice thing to debate, anyway. And debate is what
forms the global brain.
J.A. Legris
Huh?
Don Stockbauer
The whole idea of putting forth things to debate is to help form the
Global Brain, for it is this telecommunication which forms the GB,
messaging between its "neurons" ( = people). Sneaky devil, aren't
I????
Don Stockbauer
I'm so sorry, I thought you were huh-ing me, not N.
rs...@nycap.rr.com
> On Oct 29, 12:50 am, "rs...@nycap.rr.com" <rs...@nycap.rr.com> wrote:
>
> > ...
> > Every time a bad outcome occurs, the present motor program, together
> > with the present sensory condition is recorded in the synapses of the
> > thalamic reticular nucleus. The TRN is the great repository of the
> > failures in the life history of the organism. When a motor program and
> > the current condition of the world match the recorded bad outcome, the
> > motor program is halted and dissipates.
> > ...
> My mumblings are a description of the workings of a mammalian brain. Is this of any use to a person contemplating “building any kind brain for AI”? Some seem to think so. Many don’t. You pays your money, and you takes your choice.
Specifically, it does not rely on any soul (mind). The actions of
doubt, thinking, and deciding are shown as the outcome of a thalamic
reticular nucleus that is the repository of all the bad outcomes of
the organism.
Ray
Aleksandar Kuktin
> The brain is wholly a creature of the genome. The genome establishes a
> structure that can accommodate itself to the impinging universe. The
> genome does this because it survived. There is no other reason. The
> genome has lived since the first cell.
Actually, this is not so.
The embryonic development is a child of *interaction* between the genome,
the out-of-fetus influences (mother's hormonal status, for example) and
wholly "third-party" influences such as viruses.
This interaction is implemented and moderated via the cells and tissues
which undergo development.
So, to assume that understanding the "blank" brain only requires the
understanding of genome of the individual is mistaken.
> The generators are centered in the spinal cord and the hindbrain. There
> are centers for breathing, posture, locomotion, orienting, reaching,
> grasping, manipulating, licking, chewing, swallowing, and vocalization.
> This circuitry is highly conserved. It existed before the forebrain. The
> notion of a center is a simplification. The circuitry that leads to the
> generator, and the circuitry that leads from the generator to the motor
> neurons is also set up by the genome.
This is probably an over-simplification of the matter.
No single part of the brain can function without the other parts. In
fact, no single part of the brain can function without the body that
houses it.
It is true that (for example) the centers for posture are located in the
brain stalk. However, these do not work nor should they work without
other parts of the brain (I'm thinking primarily of the primary motor
cortex). These centers provide a sort of "hardware acceleration" for some
boring tasks. Like standing upright and not-falling over.
> The penultimate step of a motor program is to pass through the thalamus
> on its way to the motor cortex. Here it comes under the scrutiny of the
> thalamic reticular nucleus. The TRN is the repository of bad outcomes.
> Every time a motor program is activated under particular conditions and
> there is a bad outcome, that result is recorded in the synapses of the
> TRN. If the present motor program and the present conditions match that
> record, the synapses are activated, the TRN is turned on, and the motor
> program is halted. A halted motor program simply disappears.
In order to detect the "bad outcome", the TRN would need to (1) have an
idea of what is going on outside and (2) have a faculty to analise the
data and REALIZE the bad outcome came to pass.
I am not aware of such faculties _within_ the TRN. However, they do exist
in the forebrain. And the thalamus is, ultimately, under total command of
the cortex. So these tasks get done in the cortex, while the thalamus
(and basal ganglia) only serve as executors.
rs...@nycap.rr.com
> On Mon, 26 Oct 2009 17:24:34 -0700, rs...@nycap.rr.com wrote:
> > The brain is wholly a creature of the genome. The genome establishes a
> > structure that can accommodate itself to the impinging universe. The
> > genome does this because it survived. There is no other reason. The
> > genome has lived since the first cell.
>
> Actually, this is not so.
A flat assertion, followed by a denial, followed by mutual
recrimination. Fiddle-Dee-Dee.
>
> The embryonic development is a child of *interaction* between the genome,
> the out-of-fetus influences (mother's hormonal status, for example) and
> wholly "third-party" influences such as viruses.
> This interaction is implemented and moderated via the cells and tissues
> which undergo development.
I think it better to say that we have had a half-century of
psychology, followed by a half-century of physics, and now we are to
have (at least) a half-century of genomics. Why not?
> So, to assume that understanding the "blank" brain only requires the
> understanding of genome of the individual is mistaken.
The “blank” brain was a notion of 19th century philosophy. I have no
need of the tabula rasa. My brain is fully furnished by the genome. In
2009 to see what the genome can do, we can look at the Brush Turkey of
Australia. Hatched in a pile of rotting vegetable matter, it never
knows its parents. It slithers to the ground, and not stopping to
learn to walk, it runs madly for cover. Immediately or within an hour
(some say two), it can fly. It never looks back It does no imprint. It
has nothing to imprint on.
> > The generators are centered in the spinal cord and the hindbrain. There
> > are centers for breathing, posture, locomotion, orienting, reaching,
> > grasping, manipulating, licking, chewing, swallowing, and vocalization.
> > This circuitry is highly conserved. It existed before the forebrain. The
> > notion of a center is a simplification. The circuitry that leads to the
> > generator, and the circuitry that leads from the generator to the motor
> > neurons is also set up by the genome.
>
> This is probably an over-simplification of the matter.
I don’t think so.
> > The penultimate step of a motor program is to pass through the thalamus
> > on its way to the motor cortex. Here it comes under the scrutiny of the
> > thalamic reticular nucleus. The TRN is the repository of bad outcomes.
> > Every time a motor program is activated under particular conditions and
> > there is a bad outcome, that result is recorded in the synapses of the
> > TRN. If the present motor program and the present conditions match that
> > record, the synapses are activated, the TRN is turned on, and the motor
> > program is halted. A halted motor program simply disappears.
>
> In order to detect the "bad outcome", the TRN would need to (1) have an
> idea of what is going on outside and (2) have a faculty to analise the
> data and REALIZE the bad outcome came to pass.
I think you have recourse to a soul (mind). Only the soul (mind) knows
what is going on outside. Only the soul (mind) REALIZE’s. A bad
outcome is pain. Pain is accompanied by the release of a neurohormone
that floods the thalamic reticular nucleus. The neurohormone causes
all recently exercised synapses to be strengthened. This causes the
neurons of the TRN (in effect, in effect only) to avoid pain in the
future by halting a motor program.
> I am not aware of such faculties _within_ the TRN. However, they do exist
> in the forebrain. And the thalamus is, ultimately, under total command of
> the cortex. So these tasks get done in the cortex, while the thalamus
> (and basal ganglia) only serve as executors.
This love affair with the cerebral cortex can be overdone. I rather
like “Exploring the Thalamus and Its Role in Cortical Function”,
Sherman and Guillery, 2006, p403. “However, we suggest that cerebral
cortex without thalamus ia rather like a great church organ without an
organist: fascinating but useless.”
Ray
casey
> A bad outcome is pain. Pain is accompanied by the release
> of a neurohormone that floods the thalamic reticular nucleus.
Pain is a high level "opinion" the brain can have about itself
and learning is not just restricted to situations that the
brain finds painful.
> The neurohormone causes all recently exercised synapses to
> be strengthened. This causes the neurons of the TRN (in
> effect, in effect only) to avoid pain in the future by
> halting a motor program.
We are also pleasure seeking as well as pain avoiding.
> This love affair with the cerebral cortex can be overdone.
And a love affair with the thalamus be overdone. I doubt all
our learning is simply a blocking or selection by the thalamus
of innate motor programs. Long term learning can take place
in many parts of the brain.
The bottom line however is you need to model what it is you
are talking about if you want it to be applicable to AI. These
models will test your theories to see if they can do what you
claim and make it clear what it is you are talking about.
John
Don Stockbauer
But most of those models are pretty air-headed, like Cindy Crawford
and the like. Better to get someone who knows some science to prove
your theories.
Aleksandar Kuktin
Well, one has to start from somewhere. :)
The thing is that any love affair in the matters of the brain is likely
to be overdone. The simple reason is the fact that all parts of the brain
are interconnected. Secondly, the brain not only acts as a whole, within
itself, it also acts as a whole in relation with the surrounding (as
showed by experiments with sensual deprivation). Third, the notion of a
"soul" as in a singular conscious entity (presumed to be incorporated in
some part ("centre for soul") in the brain) is, AFAIK, unprovable.
But, if we take the interconnected brain and granulate it, we can model
our complex behavior.
For example, lets say we want to model an emotional response. A bad one.
First, in neurobiology, "emotion" is defined as a state of readiness for
a particular set of actions. So, "sadness" is a state of readiness for
performing "sad" actions (maintaining a sad facial expression, crying,
etc.). Note that nowhere in the definition are absolute terms used. The
spectrum of these actions is, therefore, not chiseled in rock. Actions
can change (so laughing can also theoretically be a sad action).
This has an important repercussion, one which I will detail presently -
namely, it is not the act that defines itself, it is the response which
defines it. Details below.
Pain can occur somewhere in the body (note: this will not be an attempt
to describe the pain pathway but an attempt to clarify what I said
earlier). Via complex bodily functions, action potentials arrive to the
brain, to a "pain induced saddnes centre". Now, fundamentally, there is
no such thing as a "brain centre for-this-or-for-that". These are our
human constructs we use to make sense of the world. This "pain induced
saddnes centre" in the example works not by somehow being conscious about
the pain, but by (i.e.) activating other centers. Namely, those which
actuate the "sad actions". So, the pain centre activates various motor
networks (the upper motoneuron is essentially never activated by non-
motor centers, only by other motoneurons) which create and maintain a sad
facial expression. The pain centre, furthermore, activates memory
systems, both to remember the current event and to recall previous. Et
cetera, et cetera..
Now, while this "pain centre" is active, it is important to note that an
emotion (state of readiness) is, if fact, not implemented by the so-
called "centre". It is implemented by everything else that this centre
influences! I say influences because only a minority of synapses are
excitatory, most are inhibitory.
The centre may be out of cortex (in fact, the "primary brain centre" for
pain IS out of cortex, in midbrain). What is important is that any sort
of brain activity if actually /emergent/.
Bolting down this emergence to a simple automation such as the one
presented in the first article is, IMHO, an oversimplification.
That is the reason I wrote my article as I wrote it. I understand I was a
bit blunt, my bad.
But I am open to objections, ofcourse. That is why I am here.
casey
>> The bottom line however is you need to model what it is you
>> are talking about if you want it to be applicable to AI.
>> These models will test your theories to see if they can do
>> what you claim and make it clear what it is you are talking
>> about.
>
>
> Well, one has to start from somewhere. :)
AI starts with computer models. These can be high level
cognitive models or low level neural network models or
something in between.
> Bolting down this emergence to a simple automation such as
> the one presented in the first article is, IMHO, an over
> simplification.
And in the form it took not very useful for it offered no
way to model it on a computer to show exactly how it might
work. There was no attempt to show for example what form
a "motor program" took or what mechanisms generated them.
And an explanation is NOT "they were generated by module x"
if it doesn't explain how to build a working testable
example of a module x".
At the most primitive level you might build a central
pattern generator using the equivalent of neural oscillators.
Show how these are wired together and triggered and modulated
by higher neural circuit inputs and how they make use of force,
position and touch feedback to bring about the desired motor
actions in variable environments.
Some of these "explanations" are useless, like saying walking
involves the use of your legs. Making reference to brain
parts instead of legs doesn't improve the explanation.
Unfortunately I don't have access to “Exploring the Thalamus
and Its Role in Cortical Function”, Sherman and Guillery, 2006,
to access if anything useful has been discovered or explained.
JC
zzbu...@netscape.net
> On Mon, 02 Nov 2009 13:09:18 -0800, casey wrote:
> > On Nov 3, 5:32 am, "rs...@nycap.rr.com" <rs...@nycap.rr.com> wrote:
>
> >> A bad outcome is pain. Pain is accompanied by the release of a
> >> neurohormone that floods the thalamic reticular nucleus.
>
> > Pain is a high level "opinion" the brain can have about itself and
> > learning is not just restricted to situations that the brain finds
> > painful.
>
> >> The neurohormone causes all recently exercised synapses to be
> >> strengthened. This causes the neurons of the TRN (in effect, in effect
> >> only) to avoid pain in the future by halting a motor program.
>
> > We are also pleasure seeking as well as pain avoiding.
>
> >> This love affair with the cerebral cortex can be overdone.
>
> > And a love affair with the thalamus be overdone. I doubt all our
> > learning is simply a blocking or selection by the thalamus of innate
> > motor programs. Long term learning can take place in many parts of the
> > brain.
>
> > The bottom line however is you need to model what it is you are talking
> > about if you want it to be applicable to AI. These models will test your
> > theories to see if they can do what you claim and make it clear what it
> > is you are talking about.
>
> > John
>
> Well, one has to start from somewhere. :)
Well, that's so true. Since so many AI philosophers are always
Portney Complaining about why some engineers believe that
might makes right. Well they don't. They believe that Satellites
GPS, Digital Terrain Mapping, and Data Fusion make right,
for the simple, simple, simple reason that so many stooge
philosophers of science believe that Gallileo makes right.
And they believe that Blue Ray, HDTV, Home Broadband,
Flat Sceen Debuggers, Holographics Integration, Self-Assembling
Robots, Digital Books,
USB, Rapid Prototyping, and On-Line Publishing makes right, for the
even simpler reason
that so many stooge chemists believe that Nobel Prizes make right.
And they believe that Cyber Batteries, Flash Memory, All-In-One
Printers, mp3, mpeg,
Atomic Clock Wriswatches, Light Sticks, Multiplexed Fiber Optics,
UAVs, Drones
and the 21st Century make right for the trivial reason that idiot
physicists believe that
Ben Franklin makes right.
> But I am open to objections, ofcourse. That is why I am here.- Hide quoted text -
Don Stockbauer
wrote:
You know, you beat me to that post. Word for word.