Modelling the human brain by modelling its evolutionary emergence

mat mats_trash at hotmail.com
Tue Feb 26 10:32:28 EST 2002


> ************
> I disagree, we have concentrated already too long on detailed aspects; 

But you talk about modeling neurones?!  You can't have it both ways!
If you want to model them, then you have to do it accurately, not just
'assuming' they are simple logic gates otherwise your results will
have no relevance whatsoever to neuroscience.

>   'There are so many psychologies precisely because there is no one
> psychology'.
> These very early remarks are still valid according to me and should spur us
> to a somewhat different aproach than the usual one.
> **************

But this statement is about psychology, which does not pretend to have
any bearing on neuronal function.  Psychologists are not interested in
synaptic function etc, becuase they study higher level processes. 
However you seem to think you can join the two levels with one fell
swoop.  You simply cannot.
> 
> 2. If the human genome project has shown anything, it's that DNA is not the
> full specification for building an organism. The diff between a fruitfly & a
> human is just 15K genes? Don't think so. Much info is encoded, then, in
> processes which we don't as yet know anything about. So it's an error to
> hang any design argument on DNA alone.
> *****************
> The human genome program considers only that part of DNA that codes for
> proteins. It is only a (small) part of DNA; much more is 'hidden' in it,
> often (I think inappropriately) identified as 'junk DNA'. There is just no
> other genetic 'transportation vehicle' to the next generation.
> ****************

Well you're both wrong here.  Genes specify proteins, envrionment
influences what happens to the proteins.  An increased number of genes
allows greater variation/possibility in the development of an organism
WITHIN ITS ENVIRONMENT.  Therefore relatively few extra genes could
allow for a huge difference in phenotype.  I don't think you can
eliminate the environment from any modeling of evolution/development. 
It is critically important.  As to Frans comment - There is no more
hidden in DNA.  Non-coding protions may of course become coding
protions by mutation, and non-coding portions also have regulatory
roles over the DNA that codes proteins.  But there is no vast resource
of hidden information within DNA that we have yet to access.  What
else would it produce but proteins?


> 4. The assertion that "modelling of the present brain is only possible by
> modelling its evolutionary
> emergence" seems to me as absurd as it would be for the heart, or spleen..
> ***************

I concur.  This evolutonatry modelling is rubbish, but the
developmental modelling (which Frans project is actually doing) is a
little more realistic if much more detail was included in the plans

> 5. Whatever reason is there to imagine that 'conscious' and 'unconscious'
> brain networks (a) exist and (b) are separate!?
> **************

Again I agree, talking about subconscious and conscious processes
prior to you having 'evolved' them in your model is just silly in the
first place.  You would want to see how they became, if of course you
actually knew what it meant for a network to be unconscious or
conscious.

> In my modelling these are interconnected networks; in fact it is only one
> network. However, the evolutionary analyses lead 

Again, what are these 'analyses'?  Are you saying you have examined
the phylogenesis of the brain over the last n-million years?

to the assumption /
> conclusion / postulate that one segement is created by a different recursive
> procedure, making it into the hidden layers = nodal points of a feed
> forward/backward network; it is part of conscious 'seeing/feeling/hearing'.
> The other segment interconnects the hidden layers via a different recursive
> procedure and operates, I assume/postulate/conclude, unconsciously. It
> models/explains the phenomenon of 'sudden revelations', 'AHA Erlebniss',
> Serendipity, etc. Any information transformartion in the second network
> segment will, ultimately, 'arrive' in another nodal point, leading to such a
> conscious revelation. Note: The sole modelling test is: Is it isomorph with
> reality? 

What is the test of isomorphism? In detail.  Ae you going to ask 'it'
questions?


This is all just total speculation (as you acknowledge).  It should be
in the result/discussion of any experiment not in its planning
otherwise you will hopelessly diverge from the real object you plan to
model.

> ***********
> 
> 6. To refer to memory & learning as primarily 'storage' is a fundamental
> error.
> **************

True

Frans, you seem to think that consructing a rather large newtork of
logic gates in a simple neural network paradigm and subjecting to some
learning data, will result in the emergence of an artifical brain the
structure of which will be isomorphic to that of the human brain.  No
matter how much you obfuscate around that point with
philosophy/psychology etc, it just won't work. See the last three
decades of AI research for why.




More information about the Neur-sci mailing list