Re: consciousness

JGCASEY <jgkjcasey@xxxxxxxxxxxx> wrote:
On May 21, 6:21 am, c...@xxxxxxxx (Curt Welch) wrote:

JC:
Clearly our subjective experiences are just the firing of
neurons. But in what way isn't clear and saying everything
is conscious is as pointless as saying everything is just
made up of atoms.

What's pointless is using the word "consciousness" and "subjective
experience" when you have no clue what it means and you can't define it.
All anyone here does is say that humans have and machines don't. Since
you can't define it, it's pointless to talk as if you were talking about
something real. It's as stupid as me making a property of humans called
selfsense and defining it to be something we obviously have but that
machines don't have. I don't have to define it any more than that since we
are going to just make stuff up and pretend like we are saying something
intelligent.

If a subjective experience is the way it
is for a machine behaving is a particular way then it may
not be a subjective experience for the machine when it is
behaving in another way. We just don't know and that is why
it keeps coming up in philosophy groups.

If selfsense is the way it is for a machine behaving in a particular way
then it may not be selfsense for the machine when it is behaving in another
way.

See how intelligent I'm being? (not)

CW:
In other words, all I'm saying, is that a car is nothing
but the movement of all it's parts. if you keep saying
cars are more than the movement of their parts, I would
like to know what else besides a lot of parts we have
to use to build a car.

JC:
And I am saying that an explanation or *useful* description
of a car is more than saying it is nothing but the movement
of it's parts.

Of course. I agree with you 100% on that point. We have to create a
useful description of what the parts are and what they are doing. We have
to create a description that would allow anyone to build one of these
things after they understood the description.

The words "consciousness", and "self aware" are no more useful than saying
"firring neurons", or saying, selfsense. They mean almost nothing. And
the little bit they seem to imply, our machines are already doing, it they
tell us nothing which isn't already obvious. (that is the machines need
sensors, processing, and effectors).

That statement is true for everything. It
does not explain how a car is different from a horse only
in what way they are the same. It is the difference between
a rock and animal we need to know in order to build AI not
that they are both nothing but interacting atoms.

Exactly. So how does the word "consciousness" or "self aware" help us in
that regard? How has anything you have ever written using those words help
us figure out how to arrange the atoms to make a machine equal in behavior
power to a human? If nothing you can write using those words helps us,
whey do you keep defending the words as if they were good and useful words
every time I point out how pointless and meaningless they are?

CW:
I believe very strongly that a signal processing machine
that processes pulse signals (aka temporal event markers)
is all that is needed to create a machine with human behavior
and human like subjective experience and feelings.

JC:
Having a working belief is fine. But it borders on religion
when you believe without working evidence that other people
can also agree with.

Except I do have tons of evidence. Pick up any book on the brain and read
it. What's in there? Billions of information processing nodes that use
action potentials (pulses) to communicate with each other. So I suggest we
need to build a machine with billions of information processing nodes that
use pulse signals. How can that suggestion be "without working evidence"?
You would have to be an idiot to think that the brain was doing something
else.

And what do I see as people suggesting that this is not enough? They argue
that we have this magic property called "consciousness" even though they
can't define what the property is, they can't produce a machine to measure
it, they can't produce a test to test for it. And they tell me that I am
the one without any evidence? Come on.

I believe in reason (even if I am not all that good at it)
and I don't see any value in beliefs without evidence except
maybe tentative beliefs until the evidence is clear.

How much evidence do you need before you will realize that the brain is an
information processing machine that translates sensory signals to effector
signals and nothing else? Every single experiment done in the past 200
years proves this to be true. Not a single experiment suggests there is
something else there. How can you call it "rational" to not believe that
the brain is nothing more than an information processing device which acts
as the controller for our muscles?

I watched a program last night where Richard Dawkins who
wrote the Selfish Gene spoke with religious believers,
Christians, Muslims, Jews, and it was scary. The evangelist
who apparently has the ear of the president didn't even
understand what evolution meant. And when challenged by
Dawkins he simply arrogantly told Dawkins not to be arrogant.

I would have liked to seen that. I guess it was banned from TV here in the
US. :)

But wasn't Dawkins being just as arrogant as the religious believers?
Testing for arrogance doesn't prove anything about the truth of the belief.

I of course understand that in searching for the solution to this problem
we need to pick working hypothesis and then try to advance our designs
based on our current best understandings. But what I don't understand is
why people keep defending the concepts of "consciousness" and "subjective
experie3nce" when both can be ruled out as invalid hypotheses before we
even start. They are invalid because they haven't been defined and you
can't work with an hypotheses that hasn't been defined.

CW:
I believe that all we need to do is build a black box that
is N pulse inputs (for the sensory data) and N pulse outputs
(for the effector outputs). I believe there is nothing magic
or special about what we will find we need to put inside this
black box. Fairly standard processors I believe will do, but
there's a small chance that the computation will be hard
enough that we will have to build special custom hardware
optimized for performing the computations.

JC:
A tentative belief. Now some practical evidence or a mathematical
proof of concept.

Again, pick up any, and every book written about the brain and you will
find all the practical evidence you need. There are probably millions of
them. What on earth are you waiting for before you understand what's right
in front of you?

Pick up any and every book about building or programming robots and you
will find all the evidence you need to understand that we know how to build
machines that produce effector outputs by doing data processing on sensory
data. Everything these machines do are very similar to what humans do. We
see things, and we react to them. The robots see things, and they react to
them. All the hard stuff about what we need to do has already been figured
out. We know how to build ears, and eyes, and arms and legs and power
sources.

All that's left is some slightly improved data processing algorithms. But
yet, a large number of people in this group don't believe that. They
believe that we aren't even close because we haven't yet made a machine
"conscious" or made it have "subjective experience". But that's all pure
bull *** because they have been brainwashed into believing that humans
have magic powers called "consciousness" and that machines don't have, even
though no one has every shown them a single bit of evidence to support that
believe, and they never even bothered to define what "consciousness" was
before they brainwashed you into believing you had it.

If you forget everything you have ever been taught, and just start over
from ground zero, and look at all the objective evidence we have, and stop
listening to the people that have brainwashed you into believing things
that have no proof, you will see where the evidence leads. The objective
rational evidence says exactly what I've been saying. We are just machines
very much like our robots, but with better signal processing algorithms
(and most important, better learning algorithms).

I agree that whatever solution will work will
not involve any magic stuff.

That's good. Now if you could only learn to ignore the crap that the
philosophers have brought up and just look at the objective evidence the
scientists have uncovered, you would be all set.

CW:
To be a bit more precise about the complexity issue, as you
know, I believe we need to build a distributed reinforcement
learning algorithm into the box. That is, we need to build
a ton of individual decision making nodes (100 billion or
maybe a few orders of magnitude more than that) and they all
need to be trained by global reinforcement reward and
punishment signals.

JC:
Where I differ is that I think you can build it up in stages
just as we did with the computer and just as evolution had to
do with animals. If you have the concept right demonstrate how
it works with a simpler machine such as one that learns to walk
then perhaps relearns when you rip one off its legs off.

Sure, that's consistent with what I've been doing. I try to understand
simple algorithms with small networks and try to make them solve simple
problems.

But when we start small, we still should have a direction in mind if we
expect to get to the end of the road. I have some very specific directions
I'm heading in. Many don't seem to have a direction at all. All they do
is pick one problem that they think might be related to the big problem,
and try to make progress on that. When we don't know where we need to go,
that's all we can do. We explore and see where it leads us.

But if you can't understand the big picture of what humans are, it's highly
unlikely that any work you do will be in the right direction. If you can't
cut through the bull *** about misguided and confused concepts like
consciousness and subjective experience by looking only at the objective
evidence we have access to, you will just wander lost forever.

This group is very much about debating what the big picture is for AI. That
is, to debate the big picture about the very basic nature of what humans
are. I think all the objective evidence which can be found in books about
the brain and books about human behavior make it very clear exactly what we
are. We are living machines that have a data processing brain that
controls how our muscles move. There's nothing more obvious that this.
And if you understand this, you should have been able to get past all the
nonsense about "subjective experience" by now and move on to debating the
exact nature of these data processing systems that exist in the brain. But
yet, instead of focusing on data processing algorithms, you keep suggesting
that the idea of "subjective experience" is somehow important and if we
don't get the algorithm correct, it might not have "subjective experience".

All we need, is hard objective data on human behavior. We then need to
build machines that produce that behavior. There's nothing more complex
here to be done. There's nothing to look for called "subjective
experience", or "consciousness". Unless you can turn those vague concepts
into hard behavior facts that I can test for in my machines, the concepts
are useless. But no on in this group has ever been able to do that even
though I and others like Don have asked people how think the concept of
consciousness is important to do it.

CW:
The decision nodes I believe will be very simple (something
on the order of complexity of my pulse sorting nodes). But
the task of updating all of them for every reward (or large
subsets of all nodes that were active in the past few minutes),
might be computationally unworkable for doing this with
standard processors (even thousands of them). It might have
to update 100 billion nodes 100 times per second for example.
To solve this, instead of using large arrays of off the shelf
processors, we might have to build custom chips that have
dedicated electronics to do the processing. They might even
be analog in nature not because it's required, but because
it's cheaper and allows us to build a human like machine with
less power and less parts.

I believe the only thing left to figure out here, is the
details of the processing of these signals and the details of
how you distributed reinforcement learning in such a network.

JC:
The devil is in the detail.

For sure.

CW:
So, then you go off and say that "subjective experience is
not just the firing of neurons". Then what is it?

JC:
The evidence is that it does, in some way, involve the
firing of neurons. In what way is not clear.

CW:
Why would a machine of the class I talk about above not have
subjective experience?

JC:
No idea on way or the other. I don't know why firing neurons
can have subjective experiences only that they can.

Don't you see that what you just wrote is total bull ***. You say you
"known" that they can have "subjective experience" but yet you don't know
what subjective experience is, you can't define it, you can't show it to
me. All you can do, is repeat the bull*** that has been brainwashed into
you - "I AM GOOD ROBOT AND I WAS TOLD TO SAY THAT I HAVE SUBJECTUVE
EXPERIENCE AND MACHINES DON'T AND I WILL KEEP SAYING IT NO MATTER HOW DUMB
IT MAKES ME LOOK".

Open you eyes and think for yourself instead of repeating what you have
been told to believe. What is it about yourself that makes you so sure you
have this stuff called SE which machines don't have? If you can't test for
it in yourself, how can you test for it in one of my networks and know it
doesn't already have it? You can't. And if you can't, what you are saying
about SE is total bull ***. You have been brainwashed into believing that
you, as a human, have something special about you, which machines don't
have.

All we know, is that humans produce very special behaviors which our
machines can't produce. Humans can pass a turning test. This is a strict
objective test of behavior. No machine has every equaled the performance
of a human on these tests. There are also millions of other very simple
behavior tests we can give to humans which machines have not yet passed.
This the the only truth about what makes humans special. This is the truth
that you keep getting confused when you talk about the stupid SE crap.
Drop the SE crap, and start talking about what humans can be tested to do,
and lets see what we can do to make our machines act more like humans.

CW:
Why would one of my pulse sorting nets not already have
subjective experience?

JC:
Again I don't know what is required for anything to have a
subjective experience. It apparently is what is like for some
kinds of machines.

Just my point. You can't tell if a machine has it, you can't tell if you
have it, but yet you believe, out of pure irrational faith, that you have
it and that machines don't have. Even though you have a name for it, you
don't even know what "it" is, but yet you keep talking about it as if this
undefined nothing was somehow important to the problem of AI. Can't you
grasp how irrational that type of thinking is and how unproductive it is to
AI work?

CW:
What type of machine do you think we need to build to create
subjective experience?

JC:
No idea.

CW:
I've outlined the type of machine I think we need to create
subjective experience. For me, the subjective experience
question is solved. That part of the machine I already know
how to build.

JC:
If you know how to build it then build it. I will give it
a good kick and see if it feels pain :)

:)

My networks have been feeling pain for 30 years. Making it feel pain is
not the problem. Making it learn to say, "Damn that hurts", and "I wonder
what pain is?", is the problem. :)

CW:
All machines with sensory inputs and effector outputs which
interact with the world already have subjective experience.

JC:
May I ask the question Richard Dawkins asked the religious
people about their beliefs, what is your evidence?

My evidence is that you can't define "subjective experience" and everything
you make a vague reference to when you try to define it I can easily prove
my networks already have it.

You ask me to show you evidence that my hardware has something you can't
test for and you can't define? Do you know how stupid asking me this makes
you look?

Before we can take anything serious, we need objective hard evidence to
document what it is we are talking about. Show me how to test for
subjective experience, and I'll do the test. If you can't show me, then I
know you have no clue what you are talking about and you should also have
figured that out by now.

CW:
If you think I'm wrong, then lets talk about what type of
machine you think we need to build to create subjective
experience because I think we all agree that humans have it
and if we want to duplicate the powers of a human in a machine
we better make it have subjective experience and awareness of
it's environment and of itself.

JC:
I don't think you can make it have subjective experiences.
This would result as a consequence of how the machine parts
are organized and on their activities. In what way they need
to be organized and what kind of activities would be involved
I have no clear view.

No one has a "clear view" of what subjective experience is because it's a
concept that has never been defined. It's just been taking on faith that
is a real and valid thing - the same stupid error the religious people
make. They trust what feels right to them, instead of seeing things
rationally. They trust their feelings instead of trusting rational logic.

My rational logic tells me quite clearly that our feelings are the behavior
of a operant conditioned learning controller which is evolving it's
behavior for the purpose of maximising future rewards. As such, anything
that it can do, or sense which creates higher future rewards, is reinforced
and repeated. As such, when we say things that indicate we are special -
that we have powers which other things don't have, it makes us feel "good"
- we keep repeating those behaviors. We keep going to see superman moves,
because it makes our brain raise its prediction about our future - that is,
if superman can be a normal person with hidden special powers, maybe we
have hidden powers we don't know about - and that thought alone, makes us
fell better. We do anything that makes us believe we have more power.

People get sucked into this God crap because it makes them feel better. It
makes them believe that something with great power cares about them and
cares about what they do. It's a very seductive belief. But it's
irrational. There is no great power that cares a rats ass about us. NO
matter how better it makes us feel, it's a still a lie. And if your faith
in reason is strong enough, you can see this. But if your faith in reason
is not strong enough, your faith in these things that make you feel better,
end up ruling your behavior. Humans are not born being rational, it's a
skill we learn.

You have an irrational faith in the belief that you have a power called
"subjective experience" which doesn't exist in our current machines (but
might exist in them if we build them right). It's irrational because you
can't even define what IT is. You can't point to anything objective and
say, that's what subjective experience is.

And you can't define it, the concept is irrational.

People, because we are reinforcement learning machines, are drawn to the
things that make life better for us (the things that create higher future
rewards). This is why we work so hard to collect things like money, and
food, and shelter, and friends. They all represent higher future rewards
for us. Without these things, we expect to have less future rewards. But
in addition to the physical objects we collect and hoard, we do the same
with our behaviors. We repeat any behavior which is expected to create
higher future rewards, and we avoid behaviors which are expected to lower
future rewards. We avoid falling from high places. We seek safety. But
do what it takes, to make our predictions about the future better. This
includes producing language. If one type of language makes us feel better,
we keep repeating it. If another type of language makes us feel worse, we
avoid it.

Talking as if "subjective experience" was a hidden super power we had is a
seductive, but irrational, faith that many people have been sucked into.
So many people keep saying it, that everyone believes it is real - so
everyone is feeding each others primitive need to believe we are special -
that we have a this super power called consciousness or subjective
experience.

Step back and try to figure out what the hell you are actually talking
about, and why you keep talking as if it were a super power that only
humans have when it's not. It's the same power camcorders have. It's the
same power that all my current networks have. It's nothing special. I can
look at my desk and identify the objects I see. We have computers that
can do the same thing (though not as well as humans can). How is being
able to do that a super power that deserves a special name called
subjective experience.

I can tell red things from blue things. So can robots. How is this a
super power?

Computer programs can learn to be attracted to things that produce rewards,
and avoid things that produce punishment. You call this "pain avoidance"
and believe it to be another super power. I believe that TDGammon and all
reinforcement trained machines already do this so it's not a human super
power at all. It's just what reinforcement learning machines do.

How is this thing you call "subjective experience" anything other than a
story you tell yourself to feed your basic need to increase your
expectation of future rewards.

Someone here has even said outright that if humans didn't have these super
powers, they wouldn't want to know about it. (or something like that).
This is exactly why the belief in consciousness is so seductive. It's a
drug that any good reinforcement learning machine will have a hard time
resisting. But, if something else makes us see the true danger in the
drug, we can resist it - because our desire to use it is overpowered by our
prediction of the dangers. If you are correctly trained and conditioned in
the value of rational thought - you will see the dangers of believing in
these ideas that have no rational foundation. But most in the US haven't
been so conditioned. They choose to believe in God stores because they are
a drug that makes them feel so good, they can't resist it. But what's
more obvious, is that it's not the believing in God that is as important as
the rejection of the alternative - they greatly fear that if there is no
God, that the world as no purpose, that there is no meaning to life, and
that they are lost, and on their own in this universe. And being able to
escape those thoughts, is what makes the belief in God so seductive, and
why it makes people feel so good. If you can "find God", or "find Jesus"
it means you have been able to total escape the belief that we are alone.
And this is something of prime importance to a reinforcement learning
machine.

Your belief that what you sense happening in your is "subjective
experience" and that other machines like my networks can't sense the same
things, is nothing more than a idea that's too seductive to ignore. Unless
you can be rational about it and look at the objective evidence and
understand why you are being seduced. Once you understand, you will fill
greater pain for being tricked by a lie, than you did by believing in the
lie. And once you get to that point, you will see the truth.

You won't feel better. Because the truth isn't always good. You will feel
much better if you can actually believe in the magic power of being human
and the magic power of our creator God. To feel the best, you have to do
whatever you can to trick your brain into making higher predictions about
future rewards and lower predictions about future pain. If you can
actually make your brain believe that when you die, you will experience
endless joy - it will be orgasmic. If you can make your brain believe that
pain is something you can't avoid - God made it happen and God is something
you can't predict - then you won't have any worries about making the wrong
decision. And that will cause your brain to raise it's expectation of
future rewards. Any trick that works, humans have used. And humans have
found a lot of them.

But if you want to understand what humans are so you can build a machine
that duplicates our powers, you have to stop letting yourself fall prey of
believing you have super powers when you don't. You have to stop letting
your gut instincts direct your beliefs, and instead, look only at the
objective facts.

The objective facts have been clear since before most of us were born.
Humans are just operant conditioned learning machines. All this desire to
be more, is just what operant conditioned learning machines do. They
reject any idea that suggests they don't have special powers.

To believe that we are just machines that do what we are programmed to do
is so fucking scary to most people that they can't even consider it a
possibility. Their mind makes them run away from the thought before they
can even given it a serious consideration.

To believe that we are just operant conditioned reaction machines and
nothing more, is a thought so scary to most people, that they run away
from the idea the fist chance they get. They put down the ideas of
behaviorism because to believe what the behaviorists have been saying, is
to accept the fact that we are nothing more than machines that our shaped
by our environment. Most people can't accept this, so they ignore it
without even understanding it. There's been a huge movement away from
behaviorism and what it suggests, because of the simple fact that they
behaviors were right, we are operant conditioned machines and being such,
we lie, and stick our heads in the sand, and do whatever we have to do, to
escape the ideas that scare us. And what the behaviorists found, is scary
to most.

But if you want to solve AI, you have to get past all this confusion
created by the mass hoards of people how are to scared to see the truth.

The entire concept of "consciousness" and "subjective experience" is
nothing more than a a complex an elaborate smoke screen put up by
reinforcement learning machines that are unable to face the truth about
what they are, and what they are not.

CW:
It better be able to see red and green and experience them
internally as different subjective experiences or else it's
not going to do what humans can do.

JC:
Ok.

CW:
I think my machines already do all this, and the only thing
left to do is make it perform better. I don't think the stuff
people spend so much time talking about here such as whether a
machine can be conscious, or whether it's aware or whether it
has qualia is a question that still needs to be answered. I
don't believe there's a hard problem. All this is already
solved, and I'm just waiting for you guys to figure that out
so we can move on to talking about the interesting stuff that
isn't solved.

JC:
Ok. You have a belief. Until you work out the "interesting stuff"
it will remain a belief.

The details have not been worked out. But what I write here, and what I
say about how "subjective experience" being nothing more than a smoke
screen, has already been "worked out". I just can't get enough people to
put down their pacifiers and see the truth.

I'm the one in the debate with the evidence. You have none. Not only do I
have the evidence (all the work of science behind me), I have the answers
to why you and others get so seduced by these smoke screen ideas.

What I don't have, is the power to recondition you to be rational instead
of being seduced by the drugs of the smoke screen ideas.

Until then all we have is your beliefs
and statements.

Right. If I made a machine that duplicated full human behavior, I'd have a
huge club to hit you with. I don't have that big club. I only have
rational logic to work with. But since you are wiling to ignore rational
logic and go with what feels good at times instead (you WANT to believe you
have something the machines don't have), I'm at a loss. You have been
brainwashed (aka conditioned) in a way that I can't undo with Usenet posts.
Maybe another 1000 of them will fix it, but I don't this one will do any
more than the last 1000.

You want others to see the pulses going left or
right as "making decisions", "forming concepts" and so on but
we need to see its external behaviour as doing just that in the
same way we can see an animal doing just that.

Yeah, making it act like an animal even a little bit would be a good club
to work with. But I'm not there yet. My frustration is not with how much
I still have to do, my frustration is that that there seems to be no one
here that can understand what I've done - and a huge part of why they
can't, is because like you, you are still under the influence of these
seductive idea drugs. You are seduced into believing what they suggest
even though they have no rational foundation of truth to them - the only
truth is that they make people feel better and because of that, people
believe.

I know I've told you this in the past, and it didn't do any good. You said
the ideas don't make you feel better.

But if that is true, why is it you believe in the idea that humans have a
special power called subjective experience and machines don't have this
when there is ZERO evidence to support that believe? Why are you so
invested in an idea that has not foundation?

Maybe, for you, it's because there are so many smart people that are saying
this stuff about subjective experience and you trust what smart people say
instead of figuring out the truth for yourself? And you don't see me as
being smart enough to counteract the huge number of very smart and well
educated people that keep saying humans have subjective experience and
machines don't? Maybe you would only understand it if I had that big club
- the working machine which duplicates human intelligence?

Maybe for me, it's because I don't trust what anyone says and I've always
liked to figure stuff out for myself. So just because there are some
"smart" people who believe this, that doesn't hold much weight with me.
I've spent my life correcting the errors of well educated (but typically
not that smart people). And I've done it by trying to understand the true
cause of everything where as many people never try to understand the cause,
but instead, just spend their time memorizing what is "right".

Now that I wrote this, I see how obvious it is. Most people don't figure
things out for themselves. They trust in the wisdom of others. We all do
it in our modern society. Most of what we have learned we got from the
wisdom of others. And when there is a disagreement, we develop skills to
analyze who is most likely correct when we don't know the correct answer
ourselves. And a prime rule we all use, is trust what the majority
believes and reject the ideas of the minority. The smaller the minorty,
the less we trust their ideas and the more hard evidence we need to believe
them.

I however happen to have spent my life trying to figure things out for
myself. No matter what someone tells me, I always like try to find a way
to verify the truth of the idea. I didn't develop my ideas of AI by
reading what the experts thought. I developed them by thinking about it on
my own. I've spent years just thinking about what must be true, instead of
reading and picking the ideas that sounded good from others. I worked out
what must be the most rational answer given the facts we have available to
us. I didn't latch on to the most popular urban legends just because they
sounded good or because they were popular. I found the answers on my own,
and then compared my answers to what others had done. Sometimes I found
that my answers were the same, or very close to what others where thinking.
Sometimes I found that people had ides so odd I couldn't at first even
understand how they thought these things (like all this stuff about
consciousness). It's taken me some time to just to backtrack and figure
out where these odd ideas came from. But I've done that as well now.

So I suspect the real issue here is that (1) you really haven't figured
anything out about this stuff for yourself, so all you have to go on, is
the best ideas you can find and understand from others. And you do what we
all do when you have to trust the ideas of others, you use all the standard
measures of worthiness. Is it coming from a well educated, articulate, and
smart person that never seems to say anything stupid and has a large
following of people that trust his ideas? Is the idea accepted by many
people who you would trust their opinion? Does the idea make sense? Does
it sound reasonable? If not, it's an extraordinary claim which requires
extraordinary proof.

So your problem, is that since you have no answer of your own, you have to
look at my ideas as extraordinary since they are not well excepted by the
educated majority. And with no ideas of your own, you are forced to wait
until I can give you extraordinary proof that my ideas are valid (a robot
built using my ideas that acted like a human would do nicely for this - but
just seeing the nets do something you see as extraordinary for a neural
network would be good as well).

I guess you will just have to learn to figure these things out for
yourself, or wait until someone else beats you to it and shows you the
extraordinary proof. And I'll just have to accept that there are few here
that have spent enough time thinking about this problem on their own to
have their own answers. And without that ground work, they will never
understand what I'm saying until I have that extraordinary proof.

CW:
The stuff that isn't solved is understanding how to make a network
of independent decision agents (nodes) learn through reinforcement.
The stuff that isn't figure out is learning how these networks
should process data so that they produce optimal solutions to the
data binding problem. How for example does such a network combine
audio and visual data and recognize that the cause of that data is
one common object (a dog). Binding the data is trivial. Any
function that takes two inputs and produces one output has bound
the data. But we can't use just any algorithm and correctly
produce signals that represent the unique objects in the sensory
environment. The algorithm has to make use of temporal
correlations and temporal predictions to identify how to correctly
bind the data. And because it's a reinforcement learning problem
as well, it also has to make use of predictions of future rewards
in determining how to bind the data.

This is the stuff that needs to be talked about and worked on,
but I can't get the group past it's obsession with consciousness
and the limits of computation and whether we need to make the
machine out of meat. And until people can understand and get past
these basic freshman level problems, I've got no one here to chat
with about the real problems.

JC:

Did I leave you speechless here? :)

--
Curt Welch http://CurtWelch.Com/
curt@xxxxxxxx http://NewsReader.Com/
.