On Mon, 28 Aug 2000 18:01:13 +0100, "Theophilus Samuels"
<theophilus.samuels at btinternet.com> wrote with regards to binary or
analogue processing in the brain:
>The world at such subatomic levels is quite bizarre, as I'm sure you know -
>the simple idea of superposition is a good example. You see, the problem
>faced by saying that the brain processes information in an analogue manner
>is that this by definition, this depends on a continuously variable
>quantity. What is this? Is it the number of ions that participate in the
>graded responses or the total number of dendrites that are affected by the
>electric fields set up by such conditions? Where does the level of
>processing lie?? Of course, nobody as yet knows the answer (if this were to
>be true) and I suspect it will be a very long time until we do.
I don't actually understand what you mean by the phrase 'depends on a
continuously variable quantity' and how that relates to neural
computation.
As far as I know there is processing at all levels of the neuron. From
the ion channels in the synapses to the ion channels in the nerve
fibres, the protein machinery as well as the grading distances at
synapses and neurotransmitter diffusion rates, and even glia. In
essence each neurone is a distinct computer performing many different
calculations and passing it on in a digital and analogue fashion. I
don't see how this would prevent a silicon-based brain simulation
since you could have each neurone function as a virtual program with
protein, ion channel, conductance, etc subroutines passing information
to another virtual neurone according to various modifiable rules
(representing synapse-buoton size, synapse distance, glial
interactions, receptor density, etc). The information passed on is
graded, but the activity on the other virtual neuron may be any
combination of digital and neural depending on the rules.
To simulate a small area neural network you will need billions of
these virtual neurones simulated within a computer in real time. And
for a brain a much much larger virtual network. But why would you want
to simulate the brain function in a digital computer? Instead I think
computational neuroscience is about drawing lessons from the neural
processing and applying it to computational understanding, rather than
trying to copy the brain in a computer.
>I think by now, we all accept that digital states are present within the
>brain, just as analogue ones are. As pointed out before, inputs from the
>periphery can only travel to the brain in trains of action potentials (ON
>and OFF events) and somehow these must be intrepreted - processed - in some
>manner unknown to us all. Also remember, that the OFF states may just be as
>equally important as ON states. For example, the absence of AP could also
>contribute to the final level of processing - as the composer Webern said
>'the rests are just as important as the notes'.
>> To conclude, the brain may not just process information by either analogue
>or binary means, but in fact there may exist a continuum between the two
>states where the brain can utilise X amount of analogue and Y amount of
>binary processing, such that X + Y = 1. The possibilities are amazing when
>one thinks about it - it is one of the reasons I enjoy Neuroscience so very
>much.
Well, that is quite a succint way of describing it, although I think
you might be mixing some terms occassionally. I sometimes get the
impression you mean neural processing when you say brain function, and
vice versa.
