Re: To the writers of the Talk.Origins website

On 5 May, 22:52, "R. Baldwin" <res0k...@xxxxxxxxxxxxxxxxxxxx> wrote:
"someone2" <glenn.spig...@xxxxxxxxxxxxxx> wrote in message

news:1178390699.784759.96980@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx





On 2 May, 11:11, n...@xxxxxxxx (Eric Rowley) wrote:
From: "R. Baldwin" <res0k...@xxxxxxxxxxxxxxxxxxxx>:RB> "someone2"
<glenn.spig...@xxxxxxxxxxxxxx> wrote in message

RB>news:1178042952.112760.270940@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx >
RB> On 1 May, 14:59, "R. Baldwin" <res0k...@xxxxxxxxxxxxxxxxxxxx>RB>
wrote: >> "Eric Rowley" <n...@xxxxxxxx> wrote in message >>

RB> >>news:WinLMSG.46370052.2698.7a4@xxxxxxxxxxxxxx
RB> >>
RB> >> > From: someone2 <glenn.spig...@xxxxxxxxxxxxxx>: >> > <snip>
RB> >>
RB> >> >> The question that you keep avoiding is what behaviour can
RB> an >> >> artificial network display which can't be explained in
RB> terms of >> >> node interactions, and the nodes giving the same
RB> outputs they >> >> would in the lab, being in the same state,
RB> and given the same >> >> inputs? >>
RB> >> > None whatsoever!
RB> >>
RB> >> Actually, that is not true. It is not possible to contrain
RB> >> the stimului to exactly what the network saw in a lab, and
RB> >> there is also noise.

RB> > I think you misunderstood. It is only the individual nodes in
RB> > the lab, being given inputs that the nodes in the system were
RB> > retrospectively shown to be given by the log. Are you
RB> > suggesting that from the log nodes would be shown to be
RB> > giving unexpected outputs, differing from those given by
RB> > nodes in the lab given the same inputs?

RB> Yes. They might not differ much, depending on the signal
RB> strength of the noise and the noise rejection of the circuit;
RB> but even if you constrained the stimuli exactly to very
RB> specific lab conditions from the initial conditions to the
RB> point of measurement (which is not possible), noise will
RB> occasionally result in unexpected outputs.

But noise can be added in the lab, so if you've done your homework
properly you would expect unexpected outputs. ;-)

I think (occasionaly, as a last resort) ;-) that the point isn't that
the behaviour of the network can be predicted but just that it wouldn't,
from a materialist standpoint, be having consciousness imposed on it from
the outside and therefore there would be no consciousness, since
everyone,
or Someone in any case, knows that the merely physical is incapable of
producing anything as wonderous and complex as consciousness.

I could be wrong of course, Someone _has_ confused materialism with
determinism before, but as far as I'm concerned, if he wants to argue
about
determinism he has to say so.

RB> You'll also find that signal acquisition in circuits varies
RB> with supply voltage, temperature, and device parameter
RB> variation. Take two neural network circuits built to identical
RB> specifications, and they will be behave slightly differently
RB> under identical environments, even more differently
RB> under different environments.

RB> There will be an error between the expected behavior of the
RB> system and the actual behavior of the system.

I doubt that God Himself knows what Someone's argument is, mine is merely
that, as far as anyone knows, following the laws of physics doesn't
disqualify
one from possessing consciousness (or useing consciousness as a basis for
behavioral decisions.)

I was not talking about adding noise in the lab, and nor was I talking
about recreating the emergent behaviour of the system in the lab. I
was simply suggesting that if the nodes wrote out all the information
such a current weightings on each connection, the node inputs etc,
then this could be recreated in the lab for a single node. You seem to
be suggesting otherwise am I correct?

Well, you previously made no argument about a single node that I saw, but
the argument works just the same. If you have a weighted sum node, the
output will be a weighted sum of the inputs plus an error term that depends
on the physical implementation of the node.

The difference with a network is that the error sources increase
geometrically with the number of nodes.

You can constrain the error sources so that a system works within an
acceptable margin of error, but your notions about exact knowledge of a
system are mistaken.


An exact knowledge of the system would be attainable via the system
log each node writes out to. I am not saying that artificial neural
networks write out to a log usually, but the is no reason they
couldn't. If you are suggesting that there were sources of error for
node behaviour, in signal reading for example (if so changes in design
which slow the system could eradicate these I should think), then the
errors could be seen in the log, and changes in the expected frequency
for the number of nodes checked. So you are still back at the same
place. The behaviour of the artificial neural network can be explained
with no knowledge of whether it was subjectively experiencing or not.
There would just be some errors in the system which can be explained
along with the node interactions.

.

Relevant Pages

  • Re: To the writers of the Talk.Origins website
    ... RB>>> there is also noise. ... RB>> the lab, being given inputs that the nodes in the system were ... differing from those given by ... the outside and therefore there would be no consciousness, since everyone, ...
    (talk.origins)
  • Re: To the writers of the Talk.Origins website
    ... RB>> the lab, being given inputs that the nodes in the system were ... differing from those given by ... RB> occasionally result in unexpected outputs. ... be having consciousness imposed on it from ...
    (talk.origins)
  • Re: To the writers of the Talk.Origins website
    ... RB>>> there is also noise. ... RB>> the lab, being given inputs that the nodes in the system were ... differing from those given by ... the outside and therefore there would be no consciousness, since everyone, ...
    (talk.origins)
  • Re: Implausibility that we can be explained by evolution
    ... in the lab, which gives the same outputs given the same inputs (i.e. ... the lab gets the same inputs as it would in the network, ... lab is influenced by consciousness. ... the reality was such that it wasn't consciously experiencing. ...
    (talk.origins)
  • Re: Implausibility that we can be explained by evolution
    ... in the lab, which gives the same outputs given the same inputs (i.e. ... the lab gets the same inputs as it would in the network, ... lab is influenced by consciousness. ... the reality was such that it wasn't consciously experiencing. ...
    (talk.origins)
Loading