Re: A request for information please.

Stephen Harris <cyberguard-1048@xxxxxxxxx> wrote:
Curt Welch wrote:
Stephen Harris <cyberguard-1048@xxxxxxxxx> wrote:
Don Geddis wrote:
Stephen Harris <cyberguard-1048@xxxxxxxxx> wrote on Sat, 19 May 2007:

In other words, the answer to why the "mind" is needed, only comes when
you accept the physicalist point of view. The physicalist answer is
that the mind is the brain, so what I believe you are really asking
(even though you may fail to believe this is what you are asking), is
why is the brain needed to make a human body act like a human? And the
answer to that is obvious.


I really don't understand why more people haven't gotten hooked on the
physicalist view. Once you understand it, and break your brain free of
all the old dualistic ideas, it's so powerful a view, and so obvious,
you can never go back to believing that the mind is anything other than
the brain. And once you see that, then the idea of consciousness as
meaning "having a mind separate from the brain" becomes meaningless and
silly. The word reduces to meaning nothing more than, "having a
brain".


I wasn't objecting to Physicalism. I was asking where is the evidence
for Computationalism, where is a coherent explanation?

10.5 "A Taxonomy of Computationalist Theses
The functional account of computing mechanisms can be used to formulate
a taxonomy of computationalist theses about the brain, in order of
increasing strength. The theses range from the commonsensical and
uncontroversial thesis that the brain processes information to the
strong thesis that it is a programmable, stored-program, universal
computer. Each thesis presupposes the truth of the preceding one and
adds a further assumption to it:

1) The brain is a collection of interconnected neurons that deal with
information and control (in an intuitive, formally undefined sense).
2) Networks of neurons are computing mechanisms.
3) Networks of neurons are Boolean circuits or finite state automata
(McCulloch and Pitts 1943).
4) The brain is a programmable computer (Devitt and Sterelny 1990).
5) The brain is a programmable, stored-program computer (Fodor 1975).
6) The brain is a universal computer (Newell and Simon 1976).34"

SH: Behaviorism and Physicalism cover 1) and 2). But where is the
evidence that you talk of that supports 3) and 4)? as an extension
of evidence for and arising out of Physicalism?

This issue of "computation" in the context of AI is just stupid. I'm not
sure why people waste such time thinking about.

Computation models (like Turing's computation model known as the Turing
Machine) are finite state models with fixed state transition rules. As an
abstract conceptual model, these are very useful in helping us design,
build, and program, computers and other such machines. But as it happens,
we don't have the power to actually build a finite state machine with fixed
state transition rules. The machines we try to build don't have 100%
probability of making the correct state transitions. We build them in such
to make the probability of the expected state transition extremely high
(99.99999% or higher?). But no machine we build will do what we expect it
to do every time, so to start with, we can't even build a machine that
actually implements the abstract concept known as "computation".

Then, to add to this, we have another huge problem. In AI, as has been
recently argued by Joe Durnavich, the problem we address is not just that
of building a robot. The problem we face is building an agent that
interacts with an environment. So the system we are trying to create is
larger than just the machine we build - it's the machine and the part of
the universe the machine is able to sense and interact with. So the real
"thing" we are creating is the larger system which consists of the
environment and the robot combined. We will know if this larger system has
equaled human level behavior when our AI agent and it's environment
produces behavior similar to what a human does in it's environment. And
this larger system, the environment plus the agent, is so far from being a
finite state machine as to make the idea silly to even consider. The
movement of atoms in this system and they way they interact with each other
(billions and billions and billions of atoms all interacting at once) is so
much not a finite state machine as to make the issue of whether concepts of
"computation" has something to do with this problem a non starter.

If we use a digital computer as the brain of the robot, then we can use
finite state machine models to help us understand and predict what that
part of the machine is doing. But it's impossible to build sensors which
are finite state machines, and it's impossible to build effectors which are
finite state machines. So the abstract finite state models we use to help
us, as humans, build and program the computer brain, can never fully
explain the larger system, because the sensors and effectors which connect
the finite state part of the system to larger environment are not finite
state machines themselves.

There are a lot of interesting things that can be proven about finite state
machines models but do do so requires us to limit our scope to finite state
machines. This is the work done in all the theory of computing work with
Turing machines. But this work has little to do with the limits of robots,
even robots with digital computer brains, since the finite state brain is
linked to, and interacting with, a very non-finite state environment.

The only interesting question to be asked here, is whether a machine built
to follow a finite state model (such as any of our digital computers or
other digital devices) be used as the central controller of a robot which
is able to produce full human behavior? And this is a question we can't
answer with the data we currently have. Not until someone produces a
machine with what we can agree is full human behavior, will we be able to
say anything meaningful on the subject. If the machine uses digital
systems as the core technology, that will pretty much answer that question.
But we might have to build it out of a lot of analog technology that we
don't know how to duplicate with digital technology. That would leave the
question of computation unanswered.

Though I happen to work with computers on my AI work, I also build robots
that are clearly not finite state machines (their location in the
environment as they bump into walls and spend the wheels are anything but
finite state in nature). When I build machines that have a wheel that can
turn, I've build a machine with an infinite number of states (the
rotational position of the wheel, and it's path through temporal space is
anything but finite). If I were to every believe that I had hit the limits
of what I could do with finite state machines as the core information
processing technology, I would switch in a second to analog systems if I
could envision them giving me an advantage. Our cars and motorcycles and
electric motors are not discrete digital computation systems and I already
use those to build AI. If I figured I could build AI out of pure analog
parts, I'd do it. For the moment, the very high speed digital systems seem
to work much better than building information processors out of analog
components like gears and shafts, or ropes and pulleys, or vacuum tubes, or
analog systems (one who's behavior can not be easily understood using
finite state machine models) out of transistor.

'The right program instantiates a mind.' 'In order to pass the TT,
a mind will need to be instantiated, no program simulation of the
mind will be sufficient to pass the TT unless a mind is instantiated.'

These are Computationalist conjectures which would require an
underlying assumption around 4) IMO.

I think everything from 2 on is nonsense and has nothing to do with
building AI machines or the limits of what we can build. I build physical
machines. Because I'm a human, I have to create conceptual models of the
machines I try to build.

Well this is not really true but without a valid conceptual model
I can't understand what I built and I think any building we do
without understanding what we have built is not very interesting. We
can build conscious machines by having sex but that's not really what we
are talking about here - that's not engineering. And we might be able
to evolve a conscious machine without understanding what we ended up
with - but that too is not really what we are trying to do here - we are
trying to engineer AI - and that assumes we understand the design of the
machine we have created.

Without a conceptual model of what the machine is, I can't build it.
Computation however is just one very narrow type of conceptual model which
is very useful in building all machines (we use computation concepts when
we build a bridge to make sure it will be strong enough for what we need to
do). Since our brain is not limited to thinking in only computational
models, it silly to even suggest that the AI machines we try to build (and
understand) will be limited to computation based models as well.

In other words, I would not in the least call my self a computationalist.
I'm an engineer. I take hunks of atoms and mold them into shapes that
behave in interesting ways. It just happens that the hunk of atoms I most
often work with are computers. As an example however, I also happen to be
in the middle of getting a college degree in welding. I'm doing this to
help improve my robot building skills. Sticking hunks of metal together
with the help of a a lot of heat in order to create AI certainly doesn't
seem to make me a "computationalist".

These conjectures are best
called open questions I suppose, because there isn't enough
physical evidence to prove it one way or another.

Yes, I think that's true. But I have little to no interest in any of the
conjectures on your list other than 1) and though I can see why some people
who like to think about the limits of computational models might explore
the ideas on your list, I think they have very little to do with what I see
as real AI work.

My real interest is in trying to create a workable conceptual model to
duplicate the interesting behaviors that a networks of neurons in our brain
is creating for us. My current best conceptual models are network of
signal processing devices which sort real time asynchronous pulse signals.
Real time asynchronous pulse signals are a conceptual model which is
outside the scope of "computation" to start with. So I start with your 1),
and take a turn away from "computation" at the very start.

My current conceptual processing nodes make pulse sorting decisions by
measuring temporal pulse spacing and comparing the spacing of one set of
pulses to another. The abstract concept in my model here is that the
"measurements" are analog with infinite precision (nothing like a finite
state machine at all). Though every one of these machines I've built was
done with a finite state computer, it wasn't a requirement by any means. I
could easily design the nodes using transistors and capacitors to store
analog charges to represent the pulse spacing.

I did watch
an IBM sponsored round panel discussion group composed of people
with decent credentials in areas of brain study, Gerald Edleman
for example. They didn't think the brain was a computer.

I might have well talked about the brain as being a computer at times, but
if I use that term, I use it very loosely and don't mean formally mean that
the brain performs a function that can be fully described using a finite
state machine conceptual model. Normally I refer to it as a signal
processor. And the signals I normally think of are pulse signals for the
sensory data flow, and otherwise unspecified signaling systems (analog
systems are quite reasonable) for passing training (aka reward) information
through the net.

Anyway, I don't see how your post about the value of Physicalism
directly supports Don's point about the boundaries of AI TT passing
program simulation.

My point was simply that the question of whether an AI machine has a "mind"
depends on what you think a "mind" is. A physicalist defines, and
considers a "mind" to be a very different thing than anyone who has some
type of dualist belief.

I'm also a little confused here on the use of the word "simulation" and
maybe I missed some important points from previous posts about this. The
concept of "simulation" can be as confusing as the concept of
"computation". It's an abstraction we use to help us talk about and
understand the behavior of physical machines. The behavior of one physical
machine is seen as a simulation of the behavior of another physical machine
with the second machine duplicates some interesting properties of the first
machine. Typically, only a small subset of the properties are duplicated
and most are not.

A waterfall is one machine. A computer is another. The computer is
"simulating" a waterfall when we as humans sense that some properties of
the physical computer match the physical waterfall. If the computer
simulation were to create an dynamic image of a waterfall on the screen,
then the properties which we see in common are some visual image
properties. The screen of the computer shares in common some visual
properties with the waterfall. We call it a "simulation" and not just,
similar properties, because we build the second machine for the purpose of
duplicating these similar properties.

A painting of a waterfall is a simulation of a waterfall because it
duplicates some of the features we care about. One waterfall is not
normally called a simulation of a second waterfall because the second
waterfall was created with the intent of duplicating some properties.

If we create an AI machine with the intent of duplicating some properties
of humans, we could call call it a simulation of a human.

But if you were to duplicate enough of the properties in your "simulation"
we would no longer call it a simulation. If you build a channel and a drop
for water to flow in, and pump water though it, this "simulation" of a
waterfall would no longer be called a simulation. It would be called a
man-made waterfall.

And likewise, once the AI machine duplicated enough human intelligent
behavior properties, it would no longer be called a simulation, it would be
called a man made intelligent machine.

But enough on the use of the word "simulation". I assumed before, and I
assume now, that "simulation" was not the point of confusion here. It was
"mind" - and whether the man made machine "had one" or not.

DG: "Just to be clear, the computationalists (including Curt and
me on this newsgroup) assert that it is not possible to "simulate
consciousness" without actually being conscious."

I agree with the major point Don is making here. I think it doesn't even
make any sense to think of a machine acting like a human and not having
every major quality that humans have which cause humans to call themselves
and others conscious.

Though I believe it will be possible to build such machines using digital
computer technology, I don't know for sure if it's practical. It might
require analog processing which, if we were to try and do it with digital
technology, might require a computer the size of the sun to equal what a
human brain sized analog machine could do. But this issue of having to use
analog machine technology is pure speculation and conjecture which I bring
up only to make it clear I'm not locked into the belief that we must use
digital processing technology which operate according to finite state
machine conceptual models.

SH: My understanding is that consciousness and mind are used pretty
much interchangeably because presumably one doesn't have a mind
without having conciousness also (first?).

I think that's very true, but I can't seem to get anyone that likes to use
these terms to admit to this. But that's what I sense motivates their
ideas on what "consciousness" is.

And to me, when they say "mind", I believe this idea comes from the fact
that we can sense things happening in our brain, which are not directly
correlated to the activity happening in the "outside" world. People
believe they have a "mind" simply because they are able to sense private
brain actions.

If humans were different, and the brain didn't produce private behavior
with no short term correlations to the physical world, (aka we didn't have
the power to produce what we call private brain behaviors - aka thoughts
and memories) that most if not all of this confusion over mind and
consciousness never would have gotten started. This is because we would
see ourselves the same way we see others - as bodies interacting with the
environment. Our image of self would not be different in any important way
from our image of others. The only difference is that we would be able to
control our own body and not the body of others - so we would still have a
clear concept of "self" vs "other" but there would be no duality problem of
"mind" vs "body". Because the concept of "mind" comes not from our ability
to act, but from our ability to have private thoughts. But this is pure
speculation.

To generate a computer
program mind they sometimes call hardware analagous to the brain.
Zombies are physically identical to humans; I think it is a
category error to call a calculator a zombie because it doesn't
have consciousness.

I think it's an error to assume a calculator doesn't have consciousness.

Like I recently wrote in another post, even the idea that consciousness is
a property of humans which is separate from our body in some way (aka
created by our body instead being our body) is as stupid as the idea that
there are invisible farting hamsters that cause cars to move. And if you
are willing to even consider the possibility that consciousness is is not
just physical brain behavior, then you should also spend equal time
considering the possibility of invisible farting hamsters - because the
idea has equal merit.

And we might as well assume that calculators work with the help of
invisible farting hamsters which do all the math as well and make sure the
transistors always produce the correct answers.

If you do not think invisible farting hamsters are a useful concept for
understanding cars and calculators, you should also not be wasting any time
considering the possibility that human consciousness and a human mind is
anything other than physical brain behavior.

In which case, it makes no sense to even consider the idea of a human
zombie because it's the same thing as considering the idea of a calculator
without it's invisible farting hamster (IFH). If you don't believe in IFH,
then it's absurd to think about the difference between a computer with, or
without, it's IFH. And since I consider the idea of consciousness as a
property which is separate from the physical body as being as stupid as
IFHs, the entire debate of machines with, or without this separate property
is just as stupid to me as calculators with our without IFHs.

Because a calculator and a human can both
perform some calculation doesn't put them in the same category.

Right. And to me, the only difference between a calculator and a human is
their physical make up and the behavior they produce. The physical
construction and the behavior of those physical parts is all we have to
define what things are. There is nothing else in this universe we have to
work with which we have yet identified. To believe there is, when there is
no data to support the belief, is just as silly as believing in IFHs - it's
just something we made up without any evidence to support it.

Don wasn't talking about zombies, but whether some computer program
that passed the Turing Test had to have a mind in order to do so.

Well, to start with, "computer programs" can't pass Turing Tests. Only
machines can. If you were to build a robot that used a computer brain to
generate its behavior in response to stimulus signals, then that machine
could be given a long and extensive Turing Test.

But to ask the question about whether it must have a mind requires us to
first define what we mean by "mind".

To me, a human mind is nothing more than a brain which is able to generate
internal signals which don't have short term temporal correlations to
external sensory or effector data. In other words, a mind is a brain that
can have private thoughts not directly correlated with the external
environment. (we can think things that we, and no one near us, are able to
sense with our external sensors). So a machine has a "mind" if it's got
internal signals which are not correlated on a short time scale with
external events.

But in the case of qualia, this just means to me that the AI machine has
internal private representations of external events (like the internal
redness signal representing light of a given set of frequency
combinations).

You can't make a machine produce human like behavior if it doesn't have the
power to think without acting, and if it doesn't have internal "redness"
signals. Without internal "redness" signals, it won't be able to correctly
see, and respond to, red things.

Also, let me say something about the Turing Test. Though it's though of
only as a question and answer game, I normally don't think of it that way.
When I talk about a machine passing a Turing Test, I'm really talking about
a machine passing every physical behavior test we care to give it so we can
compare it's behavior to the behavior of humans. This could include
complex pattern recognition and response tests, and IQ tests, and physical
dexterity tests that test how fast we learn new physical skills and how
fast we might forget them. Anything you can think to test, we could test
on the machine.

Now, the machine will not behave the same on many physical behavior tests.
It won't bleed when cut. So it will clearly be very different from humans
in some behaviors. But the point of AI was never to duplicate that
behavior - as well as many others that might require a human meat body.
But it could (and I think should) include many tests like the machines
ability to feel emotions, and feel pain, and show fear and happiness, and
it's ability to learn new things, and it's ability to be creative. What we
are trying to duplicate in AI is the behavior of the central behavior
generator (the brain) not the physical construction of the body. Any human
behavior we can attribute to the brain, we will want to be able to build
into a machine with central controller designed by us.

Now there's the chance that we could produce a machine that didn't have
what we think of as private thoughts (couldn't think to itself). But it
might still have the power to learn and use language like we do. It would
simply have to speak out loud in order to talk to itself instead of doing
the trick of speaking in our head that we do. This type of machine
wouldn't have the same time of mind we have, but yet it would be so human
like, that I doubt anyone would doubt that it was "conscious" (other than
those that just refuse to believe a non DNA based meat machine could be
conscious).

Now, however, when you use the word mind, I suspect you are thinking
something completely different than when I use it. You I assume are using
it in the more traditional dualist sense of "a mind which is separate from
the body but might be directly created by the body". You are talking about
the invisible farting hamster in the head. The only problem is there are a
huge number of being who believe they have invisible farting hamsters in
there head and they believe these invisible farting hamsters are probably
created by their brain - but the IFHs are not the brain, they are just
something created by the brain.

However, we can't test to see if the IFH is there or not. We can't test
others, and we can't test ourselves. There are simply many who believe
(actually that's too weak - they known for a fact) they have an IFH and
they think it's somehow created by the brain. So since they know they have
this IFH, they want to know where it came from. And since we have no way
to test for it's existence, we can't tell if some other brain has their own
IFH or not. So we start to wonder if everyone has their own IFH or whether
they might be machines without IFHs (we call them zombies if they don't
have a IFH).

This problem of course is of immense importance to solve. We call it the
other invisible farting hamster problem. Since we know we have one, and we
assume, but can't tell if others humans have them, and we know calculators
can't have them (we can't test calculators for IFHs but it would just be
stupid to assume they did, so we know they don't have IFHs).

So, when you say "mind" you are are think of, and talking about, the the
mystery of the invisible farting hamsters. I on the other hand, being a
physicalist, don't believe in invisible farting hamsters, so there is no
mystery here for me to consider.

We classify machines by their physical properties and nothing else. The
invisible farting hamster is a fictitious non physical property made up by
a creative human brain, but since they don't exist, it's absurd to me to
act as if the question of whether a person has one or not is a valid
question to spend any time thinking about.

Only people that believe these things we can't test for (invisible farting
hamsters) are real, are the ones with the hard problem of how to identify
something that can't be identified.

Don and I don't believe in invisible farting hamsters (I hope it's ok to
speak for Don on this issue), so we don't see any way to build machines
that act like humans that don't have all the same properties as humans who
believe they have invisible farting hamsters in their heads. We would
expect many of our machines to end up with the same silly behaviors of
talking about the invisible farting hamsters in their heads (if we
allowed our culture to "educate" them instead of educating them ourselves
to the real truth).

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

Relevant Pages

  • The never ending debate ...
    ... An illusion is a persistent difference between ... These are things the brain does. ... is involved in what we might call a Mind Activity. ... concept of consciousness. ...
    (comp.ai.philosophy)
  • Re: Origin of consciousness
    ... reacting to changes in the field organize as a cosmic consciousness? ... my brain get transferred, and how does the new media store it? ... If the ZPF exists, then every molecular change is influencing the ZPF ... Do you have any model in mind, or are you just using words which make ...
    (talk.origins)
  • Re: Does violating the laws of physics require intelligence?
    ... there is insufficient data available to derive ... consciousness exclusively from properties of the brain. ... The mind may not be exclusively in the brain it may exist in the ...
    (talk.origins)
  • Re: Is there a god?
    ... for physicalism. ... and materially co-extensive to the physical state" [of the brain ... I guess) that human consciousness is not reliably a matter ... has conscious experiences or not?" ...
    (uk.religion.christian)
  • Re: Dogma, the false god of audio
    ... > then the origin of consciousness is no longer a problem. ... > emminently reasonable conclusion that the brain is more a hundrance ... the feedback loop is one of the fundamental concepts of control theory. ... feedback loop behaves as if it has a will, but not a mind. ...
    (rec.audio.opinion)
Loading