Re: The structure of a self-conscious mind
- From: curt@xxxxxxxx (Curt Welch)
- Date: 05 Oct 2005 05:40:17 GMT
"Allan C Cybulskie" <allan.c.cybulskie@xxxxxxxx> wrote:
> I've lost the other two posts of our previous explanation (my news server
> dropped them) so if there's anything that you wanted to know, just repost
> them (I'm trying to aim for an hour a day to respond to posts now).
>
> One question I WILL answer is about research projects. I was a bit
> unclear when I mentioned research projects, since I'm not employed as a
> researcher in any area (I work as a software designer in a telecom
> company). However, I am working on a Master's degree in Philosophy and
> one of my interests is Philosophy of Mind (I'm currently aiming at doing
> a thesis on emotions and rationality, oddly enough relying on an argument
> about conditioning techniques and their use in controlling emotional
> reactions, based on the Stoic model). So my comments related to that.
Yeah, it makes sense that you have an interest in Philosophy. I don't tend
to relate well to people that have an interest in Philosophy.
> "Curt Welch" <curt@xxxxxxxx> wrote in message
> news:20050929232152.276$bW@xxxxxxxxxxxxxxxxx
> > "Allan C Cybulskie" <allan.c.cybulskie@xxxxxxxx> wrote:
> > > "Curt Welch" <curt@xxxxxxxx> wrote in message
> > > news:20050925005459.452$Rq@xxxxxxxxxxxxxxxxx
> >
> > > Look, both types of talk can be misleading. I prefer mentalistic
> > > talk because it seems to capture our own experiences and makes
> > > internal
> events
> > > important, unlike behaviourism which attempts to eliminate them. For
> > > your work in AI, you really want to eliminate internal events because
> you
> > > have no way of ensuring that a computer has them.
> >
> > Actually, just the opposite, I MUST understand, explain, and build, the
> > correct internal events into a machine in order to solve AI. It's all
> > about the internal events. If I don't correctly understand what's
> > inside, I have no hope of solving AI.
>
> The issue is that I don't think we are talking about "internal events" in
> the same way. When I said internal events, I meant internal experiences
> -- conscious thought, images, and so on and so forth -- and it seems to
> me like you are talking about some sort of hardware event.
Those are one and the same to me. You can't have an internal thought or
image without having an internal hardware event.
> But it's
> always hard to figure that out when behaviourists talk because they
> always get upset when you try to draw a line and differentiate internal
> and external experiences/behaviour, and so they either end up ignoring it
> or end up building it in and not telling you.
You don't seem to understand the difference between behaviorsim and the
radical behaviorism of Skinner. Skinner includes all internal events (like
thoughts and images) as part of the environment which shapes our behavior.
It's nearly impossible to study internal events (well, without the help of
special hardware like functional MRIs), but it's not impossible to both
acknowledge their existence and their impact on our behavior, both of which
Skinner and everyone that knows anything about Behaviorsim and humans does
as well.
The problem is not that they deny internal events, they deny that the names
you give to them are useful or correct (like goals, and intentions).
> But, back to my point, there is no requirement for internal experiences
> to produce behaviour that looks like that experience had occurred.
Right, many internal experiences are well isolated from creating direct
effects externally.
> Think
> about pain. If a computer/robot was built with a sensor that merely
> toggled a flag when you stuck a pin in its "arm", it could react to that
> as IF it had actually experienced pain, but in this case we should all be
> able to agree that it would not have experienced pain.
Now hold on there.
I actually believe a computer can experience pain. But only if you program
it correctly. And I agree that the type of program you are talking about
is probably not "correct". It's a program made to look like a machine in
pain, instead of being a machine in pain.
I belive that our pain is nothing more than the physiological reaction of
our body to the correct type of stimuls. And nothing more.
And I believe if you duplicate that type of reaction in a computer, it's
pain will be just as real as ours. The only problem is that no one really
knows for sure what it takes to "build it correctly" yet. So even people
that belive human pain is just a result of human physiology, they choose to
believe tha thumans will never be able to experience pain because the
physiology is too different. SInce no one has the clear answer to what
type of physilogy we need, no one can know the correct answer to whether
computers might be within the range of "correct class of machine with the
right physiology". But, as I said, I believe they are.
> > > > The only reason you think it's better is because you were
> > > > conditioned with operant conditioning to think that is better.
> > > > Yes, you were brainwashed into believing you acutally understand
> > > > what your mind is doing. This is the long standing problem about
> > > > humans tring to talk about the _cause_ of human behavior. They get
> > > > it all wrong most the time.
> > >
> > > But the problem is that the rejection of the operant conditioning
> > > theory is NOT because I think I know the source, but that the theory
> contradicts
> > > my personal experiences of mind. You cannot simply insist that I am
> > > brainwashed to believe something when my personal experiences seem to
> > > lead us to the conclusion that reasoning plays a part in learning and
> > > isn't conditioned.
> >
> > Well, I can insist that you are branwashed. I just seem unable to
> > prove
> it
> > to you. That is the problem of behaviorist. They understand how blind
> the
> > rest of the world is but no matter how hard they try to explain it,
> > they can't get through to many people. It's not just you that is
> > branwashed, it's our entire culture. Some people seem able to learn to
> see
> > past it, others can't.
>
> You dodged what I said. I say that you cannot insist that I am just
> brainwashed when my personal experiences seem to lend themselves to that
> explanation. Note MY PERSONAL EXPERIENCES lead to that conclusion. How
> can that be brainwashing? How can it be just a side effect of the words
> we use?
It just is. Brainwashing is just "learning". We are all learning machines
and we are all shapped by our environment. We are all brainwashed to
believe the things we believe. All our "beliefs" are just behavior. And
our behavior is shaped by conditioning. Your personal experiences don't
just lead you to a conclusion, they force you to that conclusion just like
my personal experiences force me to my "conclusions" (aka behaviors).
For example, my personal experiences forced me into becoming an english
speaking machine. I had no say in the matter. I was an english speaking
machine long before I learned the concept of "free will". I have no option
to undo that brainwashing now. Without doing severe physcial modifications
to my body, I am now stuck being an english speaking machine for the rest
of my life.
Whatever you believe, is a result of your brainwashing by definition of how
I believe we work.
By writing all these messages, I'm attempting to change a small part of
your "brainwashing" to match my "brainwashing". I'm not getting very far.
Your mind is strongly implanted with these ideas. Maybe you are too old to
be taught this new trick just like I'm too old to be taught not to speak
English.
> And yes, I am aware of how it might be. But you need more proof than
> just the idea that we all talk that way, because you have no reason to
> assert that it is just a factor of how we talk when the more reasonable
> conclusion is that we talk the way we do because that's how it appears to
> us.
Yeah, I don't have the proof. It's only how all my beliefs add up.
"proof" would be a magic set of words to brainwash you into thinking like
me. I don't have those magic words.
> > > For example, I can learn to multiply by either a) memorizing
> > > multiplication tables (the conditioned way) or b) by learning HOW to
> > > multiply (the reasoned way). Since it looks like I can do both, it
> seems
> > > odd to insist that it's all conditioned, and your standard reply of
> > > "there's no evidence against it all being operant conditioning" seems
> > > quite hollow at this point.
> >
> > I've taken a lot of time to try and explain it.
>
> Then you could have actually dealt with my example instead of going on a
> rant about how wrong I am, even though you can't in any way demonstrate
> it.
I have explained it all to you many times alrady. None of it stuck. I
don't expect this to stick so I didn't explain it again. But since you
seem to want an answer, I will.
You talked about learning HOW to multiply. What exactly do you sense
happening in your head, when you multiply 123 * 934? Do you not produce a
long stream of actions, something like, "123 * 4 is four eighty, ah ninety
6, 123 * 3 is 369, 123 times 9 is 9 18, 108, etc. etc".
What I believe, is that long string of behaviors is just a large set of
learned reactions to the environment. We were conditioned to say/think
"three times 9 is twenty seven" by a long processes of conditioning. And
we were conditioned to string that together with the other things, in
response to the question, "what is 123 * 934". All our beahvior can be
explained as nothing more than complex reactions to the current environment
(where the "current environment" includes everything happening in our head
as well).
You in effect "memorized" the sequence "three times 9 is twenty seven" yet
you used that memorized sequence as part of what you just called "the
reasoned way" to do math. What I claim, is that your reasoning behavior,
is all learned by conditioning. It's how our life time of exposure to our
environemnt left us conditioned to respond to the multiplication question.
memorizing 3*9 is no different than memorizing that you can solve 3*9 by
adding 9+9+9. We don't in fact ever menmorize an "answer". We only
memorize procedures. Everything we do is a procedure because it's all a
reaction to the recent past environment.
You were trained by conditioning how to reason.
> > > > The same thing goes for all our metalistic terms. We talk about
> having
> > > > "goals" which makes people think they must build "goals" into their
> > > > hardware. But in fact, the reason we talk about goals is not
> > > > because they allow us to understand what is happening in our brain,
> > > > but
> because
> > > > it allows us to predict future human behavior.
> > >
> > > It's both. You cannot be said to understand how humans behave if you
> > > can't predict future behaviour in at least some general way. But if
> > > goals can allow us to predict future behviour, there must be an
> > > implementation in the brain that allows us to generate and react to
> > > goals. Your system is GOING to have goals in it in some way.
> >
> > The problem is that you have no real understanding of what you are
> > talking about when you say "a system HAS goals". You don't seem to
> > have the foundation of understand needed to understand what I'm talking
> > about here.
>
> No, the problem is that you don't understand what I mean when I say
> "goals", which is why you so rant against them at the end of the post for
> no particular reason. When I talk about "goals", I'm talking about them
> as "functionality", not as entities. So you have to implement the
> functionality of goals into your system, or you won't have a system that
> does what you want it to.
OK, I'll buy that. As long as you don't take that idea of "functionality"
to the wrong conclusion.
> > > This was my reaction when people said that connectionism would
> > > eliminate beliefs and desires. IF we act according to beliefs and
> > > desires, THEN
> my
> > > beliefs and desires will be encoded into the connectionist theory and
> > > analysis of my brain ... and I'll be able to read them out from
> > > there. You cannot eliminate mentalism by hiding it in the
> > > implementation.
> >
> > Well, yes, I agree with that. If I build it correctly, you must see
> > everything in the machine I build, that you also see in a human, or
> > else I built it wrong. But, to me, this means, "if it looks like a
> > human to you, it is a human", but to many people, and I suspect you are
> > one as well,
> what
> > that means is that it must have the right stuff inside, or else it is
> > not human.
>
> Now that you know about my philosophy background, allow me to get
> philosophical on you here: what do you mean when you say "human" here?
> Because I'll assert that you will never, ever be able to make a computer
> that is "human". Why? Because a human is a specific biological
> organism, and computer chips cannot duplicate that organism. If you
> built a computer that mimicked the exact biology of a human, then it
> would indeed be a human ... but then it wouldn't be AI, but would be a
> biological human.
Right. That's all true. I can't build a human out of chips because humans
are biological.
But what I believe we will build is exactly what you see in some of the
sci-fi movies. That is, machines which act like humans, and which will
also sense "real" pain, and pleasure, and have "real" fears and emotions,
and have a mind just like a human.
> I bring this up because the fact that you don't really understand what
> the terms you are striving for mean is, I think, driving you to the wrong
> conclusions and to fight battles that either were never being fought or
> that you don't need to fight, as I mentioned in other posts and in the
> question I asked in this post that you reacted rather badly to. Before
> you can build AI, you have to know what you are aiming for ... and
> behaviourism, by rejecting mentalistic terms outright, leaves you with
> only behaviour to aim at.
Do you understand that my dislike of mentalistic terms is not a rejection
of the fact that we have private behavior and that private behavior is just
as important as public behavior?
The problem I have with mentalistic terms is that they don't mean what most
people believe they mean. I don't mind people using them (they are very
useful in our communication), but when they think our use of mentalistic
terms like "goals" defines what in fact is happening inside the brain,
that's where they have missed the boat.
And when they argure that my descriptions of what is probably happening
inside the brain is wrong BECAUSE they conflict with the mentialistic terms
they have been taught to use, that's where I start to agure. And this
seems to be where we are. My ideas conflict with the mentialistic terms
you have been taught to use and you can't get past that because you believe
your mentalistic terms do in fact accurately describe what is happening in
your brain.
You for example argued above that reasoning and memorizing were in fact
different functions in your brain when I know for a fact they are not. You
agure that position because the mentialistic terms of "memorizing" and
"reasoning" taught you to belive they were different.
I don't know how to break you free of your mentialistic langauge
programming. (notice I didn't say brainwashing this time). I do know it
took you many, many, years to learn to use it correctly, so a few hundred
posts here are not enough to re program you unless you have a foundation
that is close to allowing you to change already.
> And you'd have no problem if you didn't then
> attempt to make claims about what your AI would mean or prove if it
> worked, when you don't know what the debates are that you are claiming to
> disprove.
That triple negative sentence is beyond my power to understand. :)
> > > > We don't need to build goals into our hardware. We need to build
> > > > hardware which exhibits goal-seeking behaviors. These are two very
> > > > different
> > > things
> > > > that far too many people have failed to grasp.
> > >
> > > I agree that some people take mentalistic terms too literally,
> >
> > Well, then you are not totaly lost and maybe one day you will
> > understand.
> >
> > > but to
> > > ignore things like goals in the implementation is just as bad,
> >
> > The fact is, when people talk about goals, they sound like they are
> talking
> > about something inside us, but in fact, they never are. They are only
> > using the langauge of goals to describe human behavior.
> >
> > > and that's
> > > what behaviouristic talk leads to.
> >
> > No, behaviorist talk makes it clear that we are talking about behavior
> when
> > we talk about behavior without pretending to be talking about something
> > inside us when we are not talking about something inside us.
> >
> > To talk about behavior as if we were talking about something inside us
> only
> > brainwashes people into thinking that stuff is actually inside us.
>
> But the problem is, of course, that it leads people into ripping things
> out that they don't consider behaviour. You avoid classifying behaviour
> (even "goal-directed behaviour" types of descriptions have angered
> behaviourists in the past) and so don't really know what a behaviour is
> doing or what it means. And this can lead people to ignore things like
> goals entirely, even as functional descriptions. And it also can lead
> people to do things like we did talk about in one of the other replies
> where they claim that brainwashing is the same as a reasoned decision,
> and that if someone is conditioned or trained to do something it's as
> intelligent as a reasoned decision to take that action -- and it isn't
> clear that that's the case. At least there's a question, but
> behaviouristic talk skirts the whole issue and so leads people to
> ignoring the critical question as well.
Well, let me make this point.
Human perception is a job of clasifictaion. When we turn our perception
powers towards understanding behavior, we naturally clasify behavior into
clasifications. That's what the brain does. So that's how we end up
talking about behavior, in terms of our perceptual clasifications of what
we are looking at.
These clasifications are then given names, and those names are what we keep
talking about as mentialisism. It's just a name for a perceptual
clasification of the behaviors that come out of us.
Now, just because our perceptions has clasified a set of behaviors into
something we call "goal seeking", why is it that people want to insist that
the perceptial hardware in their brain which created the behaviors must
also work by "goal seeking"?
How we perceive behavior is no proof of how the system that created it
works.
However, how you were _taught_ to perceive behavior, seems to be your
strongest "proof" of how my description of what creates behavior, is wrong.
I don't "rip things out and ignore them" (like goals). I find a better way
to describe them.
If you are a programmer, maybe you have learned both structured and OO
programing techniques? If you have worked in both systems, you might have
learned how it's possible to write the same program, once in a structured
style, and once in an OO style, and end up with the same functionaliy, yet
the implemenation looks nothing the same. It's too completly diferent
ways, to descripe the same function.
Mentialisms, and behaviorist-talk, have much the same relationship. It's
two completely different languages to describe the same effect. However,
because of the shere complexity of human behavior, most the mentialisms in
fact describe things that aren't even there half the time. Where as the
langauge which behaviorists use very carefully examine and describe only
the things which are there, and which can be formalized.
> > > > Right. If I build a machine that acts like a human, which you can
> say,
> > > "it
> > > > seems to understand because I can explain to it how to stack lumber
> and
> > > > then it spends the day doing it - expect for the few mistakes it
> > > > made but which it quickly corrected by focusing its attention on
> > > > the work", then I've done my job. So, it makes no difference how I
> > > > talk about
> the
> > > > inner workings of my machine, as long as in the end, you see that
> > > > it understands concepts like humans do. I agree completely with
> > > > you on that.
> > >
> > > Well, let me get into a major issue here: Do you want an intelligent
> > > machine, or a machine with a mind?
> >
> > Yeah, I thought you might try to go there.
>
> Problem is that you thought I'd go somewhere I didn't.
That's always a problem. :)
> > > If you want an intelligent machine, I'd submit that we already have
> them.
> > > After all, we have computers that can easily and quickly calculate
> > > complex mathematical formula, something that would make a human a
> > > genius if it could do it. Now, if you want a computer with a mind,
> > > then I disagree that if you could build a machine that acts like a
> > > human, it would have a mind, since having a mind basically means
> > > having internal experience, and simply acting like a human doesn't
> > > prove that at all.
> >
> > Yeah, I was right about you. You just don't get it and you are not
> > alone. You are part of a huge segment of society that are branwashed
> > about this yet they have no clue they are brainwashed - they insist
> > that it's an obvious fact they are not branwashed about it because you
> > know for a fact humans have goals inside them.
>
> I never asserted that. I asserted that we have goal FUNCTIONALITY inside
> of us.
>
> >
> > > If you want a self-motivating, intelligent machine, then that's okay
> > > as well. But it might not have a mind.
> >
> > "mind" is just another one of those mentalistic terms which you have
> > been brainwashed into beliving you have inside you. You don't.
>
> Sorry, but when I talk about mind, I talk about mental experience itself.
> And I KNOW I have that. You can call that internal behaviour if you
> like, but we cannot demonstrate -- currently -- that we can make a
> computer have mental experiences. Not because of any sort of
> mentalistic assumptions, but simply because we currently do not know of
> any external behaviour that would demonstrate that an internal experience
> had occurred. And so you cannot prove that a computer that acted like a
> human being has internal experiences like we do ... and so you cannot
> prove that it has a mind. In order to do this, we need more than a
> behaviouristic view of things ... we need mentalistic analysis to
> determine what actions would indicate that something has internal
> experiences.
You seem to have confused the mentalistic with internal behavior a few
times now. I think a lot of anti-behaviorism folk make this mistake.
Most mentialisms are not in fact making reference to somethign happing
inside the brain. They are just describing external behavior. So the
error is that mentialistic terms teach us to belive that we know something
about what is happening inside us, when we don't. All we know is what is
happening outside us.
What we sense happening inside our head (our thoughts) have the same
problem. We sense the thoughts (what we call our conscious experience),
but what we can't sense is the sub conscious (the hardware at work creating
our conscoius thought sensations). And when we attempt to describe our
private beahviors, we make the same mentialist mistakes we make when we try
to describe our external behavior. We talk about what is causing the
behavior, when we don't know the cause.
We are taught to describe "reasoning" as fundimentially different from
"memorizing" when in fact it's not. Both our a learned procedures. One is
just very simple A->B. where the other is far mroe complex,
A->B->C->D->E->F.
> Your mom just
> > told you you did and you belived her because we all believe our mother.
> > She also probably taught you all about those goals hidden inside you as
> > well and that warm heart you have inside you that makes you such a
> > caring person.
>
> Hey, where did I ever claim to be a caring person [grin]?
>
> > Yeah, all that stuff is inside us and if I don't build it into my
> > machine, it will be a heartless, mindless machine with no goals and no
> > ability to care about anything.
> >
> > Do you grasp at all how silly and pointless that type of talk is when
> > it comes to helping us know what to build in the machine to create AI?
>
> Do you grasp that I never said that at all? I'm talking about
> functionality, not implementation, when I talk about goals.
> Let me explain how I look at this: I see mentalistic talk and folk
> psychology as being algorithms of intelligence. As such, you should be
> able to take those algorithms and implement them in many different ways.
> To continue the computer example, you can implement the same sorting
> algorithms in many different ways. For example, a language with lists
> and no arrays will implement it a lot differently than a language with no
> lists and arrays. Parallel structures can implement some of the same
> algorithms as non-parallel systems, but will implement them differently.
> And mentalism seems to be the best and clearest set of algorithms we
> have.
Ok, I agree with you up into the conclusion.
If you study behavior carefully, and objectively, you find out how totaly
useless our mentialisms are in correctly and accuratley descriping the
human function we call behavior.
> So I suspect that you could build sentential goals, beliefs, and
> desires into a system and if you had enough detail and a good enough
> reasoning learning system, it could gain reasonable intelligence
> (inference engines prove this). I also think that you could build a
> system WITHOUT sentential goals, beliefs, and desires and have it work as
> well. I also don't think that humans have sentences in the head that
> refer to goals, beliefs and desires ... but I believe that we have the
> functionality DESCRIBED by mentalism in there somewhere. And you seem to
> ignore the functionality in your rush to operant conditioning. It isn't
> clear to me how the role that goals play in FORMING reinforcement is
> considered in your model.
I've just been talking a lot about this in other posts with Louis and I've
just come to some new undersandnig on how my reward prediction system can
just as easilly be looked at as a goal defining system.
So, the short version is that we recive primary rewards from our hardware
(pain and pleasure), but we also have a mostly hidden and invisible, reward
prediction system at work which attempts to constaly predict the odds of a
future reward. And those preditions, are used to reinforce all our
behaviors (our thoughts as well as our external behaviors). These
predictions constantly shift with every little shift in the state of our
environment, and all those little prediction shifts are shaping all of our
behaviors.
The rediction system is what tells the brain what is "good" and what is
"bad". The "good" stuff are what you call our "goals", and the "bad stuff"
is the "negative goals". This value prediction system is our current
internal "goals" and it's constantly shaping our behavior into goal seeking
behavior.
> And, as you say below, I'd consider behaviourism the use cases for
> intelligence, but no more.
>
> >
> > As you pointed out before, behaviorism speak doesn't tell us what to
> > build into the machine either. It only tells us what the machine
> > should do when we build the right stuff into the machine. So, without
> > someone telling us how the brain works, what we are left doing is
> > guessing at what we need to build into the machine. But, after we
> > guess, we can then test or guess,
> by
> > seeing if the machine we built produces the correct types of behavior
> > as formally and accurately defined by the langauge of behavorism. If
> > it doesn't have that behavior, we have built the wrong stuff inside the
> > machine, and we know to try something different.
>
> > But the problem here with mentalisms is that it causes people to try
> > and build things like "goals" into their hardware, and to totally
> > ignore, our real behavior. It causes poeple to go looking for the
> > "memory hardware"
> in
> > the brain, instead of looking for the systems that explain our actual
> > behavior.
>
> If you are talking about building things into the hardware, you already
> have problems. What we want is the algorithm of intelligence, which can
> then be built onto almost any hardware we want. Intelligence is not
> hardware; intelligence is software.
Software is just a description of hardware. The word "hardware" is just a
description of hardware. The difference is insignificant.
When I write software, and hit "run" I have built new hardware. I caused
the machine to physically reconfigure itself into a new form to match the
machine design I just wrote in the software.
I like building my machines with an IDE instead of with a soldering iron
because I think an IDE is a better tool for building new computers than a
soldering iron.
I do all my work in software, so when I say "build hardware", I'm just
talking about writing software.
But actually, my real "machine designs" are even more abstract than the
software I write to test it. The software I write to test it is just a
protoype machine I build following the abstract design.
> If you want intelligence to be hardware, then how could we guarantee that
> a computer was intelligent if it didn't have a brain?
Until we find the right machine design, we won't know if we can implement
it in a computer. I happen to think I can however. Many people think we
can't. We don't at this point even know if we have the power to describe
the design of an intelligent machine. It might be impossible for us to
ever understand.
I just happen to believe we will be able to and that it won't be all that
far away from now that we find oursleves building machines which are
smarter than humans in all ways. At that point, human intelligence will be
no more special than any other technology, like planes, and telephones, and
cars, and electic motors and space ships and computers.
--
Curt Welch http://CurtWelch.Com/
curt@xxxxxxxx http://NewsReader.Com/
.
- Follow-Ups:
- Re: The structure of a self-conscious mind
- From: Lester Zick
- Re: The structure of a self-conscious mind
- From: JGCASEY
- Re: The structure of a self-conscious mind
- References:
- Re: The structure of a self-conscious mind
- From: Allan C Cybulskie
- Re: The structure of a self-conscious mind
- Prev by Date: Re: What did that thread indicate?
- Next by Date: Re: Temporal Learning
- Previous by thread: Re: The structure of a self-conscious mind
- Next by thread: Re: The structure of a self-conscious mind
- Index(es):
Relevant Pages
|
