Re: type writers and tooth aches II

"grayden" <gsolman@xxxxxxxxx> wrote:
> Curt Welch wrote:
> > "grayden" <gsolman@xxxxxxxxx> wrote:
> > > Curt Welch wrote:
> >
> > > But you are forgetting that that whole interpretation is a result of
> > > your experiential set.
> >
> > I don't believe I'm forgetting that. Though I'm not exactly sure what
> > you think an "experiential set" is.
>
> Sorry - that was pretty ambiguous on my part. By experiential set, I'm
> refering to that subset of environmental data which an individual
> actually experiences directly. IOW, a generalized term for an
> organism's collection of sensory inputs.
>
> > > There is no justifiable reason to assume any
> > > hierarchical precedence of 'reality'.
> >
> > I'm not exactly sure what you are trying to make reference to there.
> >
> > But at the atomic level, all interaction is linked into causal chains
> > (atoms pushing their neighbors around, which in turn push their
> > neighbors around) which create a hierarchy of interaction which leads
> > to the "virtual" ideas at the higher level I was talking about. It's
> > not just two levels, it's millions and millions of levels however in
> > this complex and long causal link of motion in the universe.
>
> My issue is with the fact that there is no actual precedent to adopt
> the belief you just stated. For starters, the billiard ball model has
> been out of date for close to century - but even allowing for that, my
> issue is with the assumption that there are any levels at all. Top-down
> structure modulates the action of bottom-up elements which determine
> top-down structure. No interaction has absolute causal sway over
> another. The system must necessary develop as a whole. If you were to
> cut up an orange and look at consecutive slices, you would be in a good
> position to predict what one slice would look like based on the
> previous one - but to say that that slice was caused by the first one
> is clearly a misinterpretation. So what I was refering to was the fact
> that although our subjective experiences are almost certainly a result
> of a complex configuration of energy - and that configuration a result
> of some particular history - there is no legitimacy in claiming that
> that experience is 'less' real, anymore than one could say that the
> frequency of a wave is less real than its wavelength. They are two
> aspects of one structure. To state that our experience is less real
> than the 'outside' world - is to claim that if we were to dispense with
> consciousness, the world would remain essentially unchanged. But this
> is impossible. We *know* that the way our universe works has resulted
> in consciousness. The structure and function of the system implies
> consciousness, so if you want to get rid of consciousness you have to
> look at a different system. How then can experience be somehow 'less
> real' than the system that implies it?

I think my use of "virtual" has you going off on a tangent that I don't
really understand. I have no desire to get rid of consciousness so I don't
know why you are talking about that. I only want to understand what it is.

But, the odd things about all this, is that our brain (we) only "know"
about everything indirectly. That is, it's as if we are forced to see and
interact with the world through remote control. Because of this, we can
never know if the world we think we see, is the "real" world, or only what
our the "media" allows us to see.

Now, for AI, I don't think this fact that the system has the power to hide
reality from us is an issue. I think with our current understanding of the
"real world", we will have no problem building conscious intelligent
machines.

But once we have mastered an understanding of AI, I suspect it will give us
new insight into the tool we are trying to use (our brain) to study the
universe we live in, and that insight will allow us to understand how our
very ability to "understand" is limited by the tool we are using to do the
understanding with. This might open some new interesting insights in
physics in regards to some very basic ideas like time and space.

> > > Your brain 'locked inside' your
> > > skull, isn't really locked inside anything. The neurons themselves
> > > may happen to be encased in bone, but their functioning is explicitly
> > > reliant on a continuous chain of dependencies. They rely on blood
> > > flow, nutrition, a body to interface them with the environment,
> > > sensory input to interpret. All of these elements are logically and
> > > physically enmeshed. If you try to define a physical boundary between
> > > 'you' and the environment 'out there', you'll find that nothing you
> > > come up with is complete.
> >
> > It's not hard to do at all.
> >
> > It's the signal processing that creates our consciousness.
>
> Though in *belief* I agree with you entirely, it's premature to state
> it as fact.

Right. We just don't have the data to know if this is true or not. Mostly
what I've been trying to talk about lately is that we have no data to show
this is not possible. The reason so many people see problems with a
physical explanation for consciousness is just a problem of our culture and
not a problem of science.

> > Even though the
> > correct functioning of a neuron depends on the rest of the body (and
> > the environment) doing it's job, the neurons don't fire in response to
> > blood flow. They fire in response to other neurons firing.
>
> But without blood flow and the associated systems, A: this firing would
> never happen, and B: it wouldn't be able to do anything. If I dump a
> mass of neurons in a bathtub or on my back porch, they will not form
> networks and process information. They will die. There will be nothing
> supporting them. No nutrients to draw energy from. No chemicals to
> modulate their expression patterns and signalling properties. No
> delicate ion concentrations to allow membrane depolarizations. No
> precursors to allow construction of neurotransmitters and
> neuromodulators. There is a prevalent assumption that neurons are more
> important than the rest of the body just because they process
> information. But in reality, your stomach processes a vast amount of
> information, with very real and noticeable effects. It is as reasonable
> (if not more so) to say that the stomach grew a brain to get more food
> than to say that the brain grew a stomach to support itself. (Of course
> they both developped together - but that's really my point anyway).

Right, I think I agree with all that.

But, the question at hand for AI is what do we have to build to create a
conscious machine? My belief is that we don't need a stomach or blood. A
battery will work just fine.

My belief is that any signal processing machine with the correct function
will work just fine. And by signal processing, I'm just talking about a
physical system with the correct type of limited causality links (signal
paths) between the components. It makes no difference if those signal
paths are created by water flow in pipes, or electron flow in wires, or
chemical systems like the brain uses. If it's got the correct type of
signal processing hardware, linked to the world with sensors and effectors
which are compatible with the signaling system, it will be a conscious
intelligent machine with powers equal to humans.

> >By tracing this
> > causal flow of information (the signal processing), we can identify
> > multiple loops which I believe is the essence of what creates human
> > consciousness and what defines the boundaries between "us" and "out
> > there". There is no single hard boundary, there are multiple loops, but
> > along with the complexity of the loops comes the complex ideas of self
> > which humans have.
>
> Though I have a rough idea of what you mean, and is quite a sensible
> interpretation - I can't really say much without a more explicit
> definition of what constitutes a 'loop'.

A feedback loop in a single processing system - like a microphone in a PA
system picking up the sound from the speaker causing a feedback loop. Or
the intentional negative feedback loops built into control systems.

Signals propagate over time no faster than the speed of light, and if you
create a casual path that forms a closed loop in a signal processing
system, you that's the type of loop I'm talking about.

In humans, we have external loops through the environment. The brain can
make us speak, and then we hear what we said. We move our hand, and we see
the hand move. The movement of the hand causes us to touch something which
causes a sensory input signal.

Loops are normally problems in control systems. Most control systems are
designed not to have loops because they tend to destabilize the system.
The human brain on the other hand is a control system that exists in a
constant and unpredictable feedback loop - and must learn to deal with it
anyway.

> > Everything inside the loop is what creates the "us", and everything
> > outside the loop is the "out there".
> >
> > Most important however is not just the loops, but the causal links of
> > the learning system that is constantly adjusting the signal processing
> > hardware. That is, what causes us pain, and pleasure, is what is
> > important to us because those things are what the learning hardware
> > uses to shape the behavior of our neurons - to shape our very
> > consciousness. [...]
> >
> > > You, the environment, and the boundary are all
> > > manifestations of the same environment. So to say that there is
> > > something 'out there' and something else that is somehow less real
> > > 'in here' is largely unjustified.
> >
> > Yes, if you ignore the importance of the signal processing, and the
> > learning hardware that adjusts it, there seems to be no place to draw a
> > line. The body and the rest of the environment are one large dynamic
> > system. The neurons depend on energy which comes from the environment
> > so the entire environment, out to the Sun and beyond, are all part of
> > what makes us what we are. So, if you look at this you might wonder
> > why our "consciousness" is limited to only our body? You might wonder
> > why the whole universe doesn't create one large conscious "being"
> > instead of lots of individual beings.
> >
> > The easy answer is that it does. i.e., everything in the universe has
> > some degree of consciousness and our consciousness is all linked
> > together.
>
> Couldn't agree more - as I've argued before in a different thread.
>
> > Human consciousness however is different because of the complex way we
> > interact with our local environment. The complex causal loops created
> > by the structure of our body and brain creates a dense concentration of
> > data flow though our brain and though our local environment. If you
> > were to try and draw these loops, you would end up with a diagram that
> > looks something like the magnetic flux lines which surround and pass
> > through a magnet with a very dense concentration in the brain, and
> > decreasing concentrations as you get further out into the environment.
> >
> > My consciousness extends out into my office because as I move my head,
> > the view of all the walls around me changes. I'm consciously aware of
> > how the light coming from my office changes as I move - or as something
> > moves my walls.
> >
> > I have no current conscious awareness of what is happening in the next
> > room because the causal loops don't currently connect me to the objects
> > in the next room. My "conscious flux" lines don't extend into the next
> > room at the moment. Or more accurately, the connection is so weak as
> > to be "off my radar".
> >
> > This disconnect from the next room happens because the causal links of
> > neurons are discrete/digital - they either fire or don't fire. It's
> > not a continuous/analog causal connection. Signals too weak to make
> > the neuron fire are ignored so they can't flow through these loops.
> >
> > What we "experience" in our consciousness is this entire set of
> > conscious-flux lines surrounding us. And these flux lines are created
> > by the causal features of the brain, and the body in how it's connected
> > to the environment and how it forms these local loops.
> >
> > Any machine that can create the same type of causal flux lines with its
> > local environment should have a similar type of consciousness in my
> > view.
> >
> > A rock has a causal link to its local environment, so I think it's
> > conscious as well. But where it has 1 path through the loop, we have
> > 100 billion paths. That's what makes our causal loops create a type of
> > consciousness not found in rocks.
> >
> > And the fact that each human body creates it's own set of closed loops,
> > is why each human seems to be an independent consciousness.
> >
> > When we interact with another humans, we create larger loops that pass
> > through both of us. The better "connected" we make our causal loops,
> > the closer we become to being one consciousness instead of two.
> > However, our ability to connect our loops is limited by the bandwidth
> > of our sensory channels and by our ability to link our sensory
> > channels.
>
> I have no complaints with this, and it is in fact almost perfectly in
> line with my own beliefs. However, I maintain that you still can't
> actually assign preferential importance to any part of the system. No
> matter how many steps removed, if you take away a clump of apparantly
> peripheral 'loops', then the whole chain falls appart.

That's why the brain is a learning controller and not hard wired. When we
gain weight, or grow taller, and typical controller would have to be
reprogrammed to deal with the new body size and shape. A learning
controller however is built to adapt to a changing environment - including
when the body changes because to the brain, the body is all part of the
unknown environment it has to deal with.

> > > The relation of consciousness to its
> > > instantiating substrate can only be analyzed by that same
> > > consciousness,
> >
> > I don't really grasp what you are getting out there. Humans have no
> > problem analyzing each other. I don't think I grasp what you mean by
> > "analyze" if you think only a person can analyze themselves. Humans
> > also have no problem analyzing pigeons.
>
> I just mean that any and all of our observations with respect to the
> brain are limited by the capabilities of that brain. Even using
> microscopes and EEGs and fMRIs, the information that we can ultimately
> interpret must be presented through one of our existing sensory
> modalities. So the question of how consciousness arises from our form
> is made particularly difficult by the fact that our consciousness is
> itself (or at least seems to be) limited to a finite subset
> information.

Yeah, I think the key to solving the consciousness problem is to just
ignore it. When we solve the behavior problem, we will end up with a
conscious machine. Mostly I think the consciousness problem is just that
we were trained by our culture to see it as something other than what it
actually is.

Solving the behavior problem is just an issue of building a reinforcement
learning machine to act as a body controller.

> > > which means the knowledge of the substrate being
> > > analyzed is limited by the perceptual efficacy of the consciousness
> > > it generates. It's a problem akin to Godel's incompleteness. There's
> > > a hole in the system that can be easily patched from a more
> > > encompassing point of view, but this larger system in turn has its
> > > own hole.
> >
> > Yes. But not all self referencing systems need be incomplete or have
> > "holes".
> >
> > Have you have tried to write a computer program, that when you run it,
> > it prints out a copy of its own source code? This is yet another
> > example of self reference that can seem at first to be impossible.
> >
> > If you write program X, it's easy to write another program, Y, which
> > when run produces the source code for X as its output. But program Y
> > is larger than program X. So how can a program print a copy of itself
> > since it can't be larger than itself? It can be done in just about any
> > programming language.
>
> My assumption would be that you would make it read the data from its
> location in memory, however I have not given very much thought to it -
> so I may be missing something.

If the source code for the problem is on the disks, and you write a program
to read the source code and send it to the output, that's "cheating". Any
data you use like that has to be part of the program.

The solution is to read/use the data twice instead of just once.

The trick is that the program has two sections, the code, and the data.
And the code is broken down into two sections as well. The first part of
the code, reads the data, prints all the code. The second part of the
code, reads the data again, and prints the source version of the data.

This way, the data you need, is used twice, so the data can be shorter than
the entire program (half the length), so there's no problem having the data
as part of the program.

> The issue though, is that however your
> clever program works, that very functioning is reliant on the
> implementing system - which cannot be fully described by the program.

Not really true.

The trick is to use abstraction. That's how the above solution worked. If
I describe a NAND gate, I don't have to duplicate that whole description
for every NAND gate used in the computer I'm trying to describe, I simply
make a reference back to my first description for each NAND gate. When I
describe a memory cell, I just say, "repeat that a billion times", and then
I can easily store the full description of the memory, in the memory.

Likewise, it's quite easy to store a complete technical description of a
computer, in the same computer - down to the detail needed to build a
clone.

> If you were to expand its abilities such that it could observe the
> external features of its local environment, it would still be unable to
> explain how that new larger system is implemented. The appearance of
> closed self-referencing is an illusion.

I'm not sure that's true. The power of language to re-use ideas I think
makes it possible.

> > I think we can understand how we work because the knowledge required to
> > understand how we work is far less than the total knowledge we can hold
> > in our signal processing system. Understanding how we work, is easy -
> > understanding everything about ourselves is impossible.
>
> I'm not sure I see what the difference is.

Well, I'm not sure I really understand it either. I'm really just playing
with a lot of vague ideas.

But the power of language makes it easy for the description to be smaller
than the thing described.

I can write a short and simple random number generator with a few lines of
code what produces (a.k.a. describes) a sequence made up of a million
numbers. That is, the source code of the program is a description of
something which is many times larger than the description itself.

Our ability to "understand" is just an inverse of this processes. The
brain (so to say) attempts to create a description of the sensory data
which is as simple as possible. And as such, we can explain and
understand, useful aspects of the behavior of billions of atoms in a
dropping apple with far fewer atoms in our brain.

However, I think the brain is just a reinforcement learning system. Which
means we can describe it in software as a program which is constantly
update a large set of data. To understand how it works(we can build one),
we only need to understand the program, and not the data it produces after
a lifetime of interacting with a complex environment.

Or in biology terms, we need to understand what neurons do, and how they
wire themselves, but we don't need to understand all of the actual
connections in a single human in order to understand how all humans "work".

This is not meant to be some type of proof that it is possible for humans
to "fully" understand humans, but just that there's a chance that it's
possible for us to build machines with the same powers as humans even if we
don't have enough brain power for one of us ever fully understand a single
human and all their complex "behavior".

> > There's no way I
> > can predict what I will do next, but I can know enough to build a
> > machine that is conscious in the same way I'm conscious. At least I
> > believe this will turn out to be true.
>
> I think we will learn the requirements for a consciousness-effecting
> system, but I'm not sure if we will ever know how exactly those
> requirements work together to create that consciousness, nor how
> exactly the subjective experience arises. I do believe we will be able
> to simulate or physically instantiate artificial conscious systems -
> but that does not imply that we will fully understand why or how they
> work.

Yeah, I think that's a possibility. However, I think I already know all
the major answers, so I have less doubt. I think we are just a
reinforcement learning machine and I already have a good idea how to build
a machine with similar powers so I don't see this is being very hard to
understand at all.

> > > That we don't experience everything doesn't imply that what we do
> > > experience isn't real.
> >
> > What we experience is "real" by the very definition of the word. But
> > what we experience is (by the fact of how the universe works) always
> > the shadows on the cave wall and never the objects creating the
> > shadows.
> >
> > If billiard ball A, hits B, which hits C, C only "experiences" B
> > directly, but that experience was caused by A so C has a causal link to
> > A.
> >
> > A -> B -> C
> >
> > C has "knowledge" of A indirectly through the actions of B. B is
> > nothing but a shadow of the actions of A. This happens at the atomic
> > level in our universe. Atoms move mostly in response to their
> > neighbors, but all that motion is shadows of the motion of other
> > neighbors. It's impossible for the action of any single atom to have a
> > direct causal link to all other atoms in this universe so all our
> > "knowledge" must be "second hand". That's if "knowledge" is ultimately
> > nothing but the motion of matter (which I believe it is).
>
> But if knowledge is in fact the motion of matter, then it must
> necessarily have the same reality as that which it is knowledgeable
> about.

Yeah, it all seems to be the same reality for us. But there's no reason it
has to be that way. If I can build a conscious machine using signal
processing, then I could also build a simulation of some world and connect
my machine to that simulation. But if that simulation doesn't have any
"hole" to allow the computer in the simulated world to see our world -
where the computer the simulation and the conscious machine is running on -
it will have no power to see the reality which created it (our universe).
It would only have the power to see, and understand, the reality of the
universe I wrote a simulation for. The universe it could see, and
understand, would be just a small subset of the larger universe it actually
existed in.

We might be caught in the same problem. The larger universe that created
us, might not be visible to us from inside this universe.

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

Relevant Pages

  • Re: antimatter rockets
    ... One of the interesting things about the interstellar environment is ... the development of humanity sees a triumph of cooperation over ... After all if there is an infinite store house (the universe) of raw ... creating conditions to reward selection for competition. ...
    (sci.space.policy)
  • Re: third believe system Re: Does Searles "Chinese Room" argument imply that consciousness is non
    ... causality link to - which is the entire universe. ... rays is different than the way we think about sensory data. ... And we have a direct causal link to cosmic rays. ... any other humans to create context "dog". ...
    (comp.ai.philosophy)
  • Re: Determinism
    ... The double-slit experiment confirms that the universe is neither strictly ... environment according to their wishes. ... strictly existent, or strictly nonexistent. ...
    (sci.physics)
  • Re: third believe system Re: Does Searles "Chinese Room" argument imply that consciousness is non
    ... Since we can think about cosmic rays we ... rays is different than the way we think about sensory data. ... And we have a direct causal link to cosmic rays. ... There is no doubt stuff happening in the universe which we either have no ...
    (comp.ai.philosophy)
  • Re: Determinism
    ... The double-slit experiment confirms that the universe is neither strictly ... These organisms, of which humans are a type, collect information about ... environment according to their wishes. ...
    (sci.physics)