jbh at anat.UMSMED.EDU (James B. Hutchins) writes:
>Again, in my opinion, there IS no "mind-body problem". That one was
>easier to solve than a brain transplant :-).
>Minsky may want to jump in here (I _hope_ you do!) but I would submit
>that any analog processor with (say) 10^11 elements each connected in
>10^3 (or more) possible ways will look an awful lot like either a) a
>'mind' or b) a horrific mess.
My name is not Minsky, but I wondered why you insisted on ANALOG
processors. I don't see, why also digital processors of sufficent
computational capacity won't do the job of serving as a hardware
platform for a mind comparable to ours.
If you say that it might be possible that our mind software needs
some kind of analog arbitrary precision (i.e. there are
states that can't be mapped on a finite set of integer numbers) hardware
support, what about the following argument: As an analog machine, the brain
is influenced by a lot of noise, i.e. in every numerical model of
a single neural processing unit (no matter, whether its biological
version is called neuron, synapse, neuron cluster, cortex column, or
whatever you like) you'll have a lot of random variables. The quantization
of a digital system is also another type of noise from an analog systems
point-of-view. If you keep the quantization fine enough, so that
the noise caused by it is significantly smaller than the natural
noise (e.g. caused by the brain's temperature), then there should be no
difference between analog and digital brain emulation techniques.
If you say that a brain's function depends on natural noise present in
every analog system, but which is absent in well engineered digital ones, then
what's the problem with digital pseudo random generators? They can easily
do the same job as analog noise sources if carefully designed.
I don't believe that when we will ever be capable of building
artificial brains, we will be limited to analog processing systems.
This makes me very optimistic, because digital techniques are today
much more advanved as far as miniaturization and flexibility is
concerned.
Markus
--
Markus Kuhn, Computer Science student +0o0; University of Erlangen, Germany
Internet: mskuhn at cip.informatik.uni-erlangen.de | X.500 entry available
