Modelling the human brain by modelling its evolutionary emergence

Frans van der Walle fw.novoware at
Fri Feb 22 08:44:29 EST 2002

2nd reply to Mat:

Re-reading once more my last response and your preceding comment, I realized
that I omitted
to answer a very valid remark of you on the need for investigating on a
detailed level the
'neural hardware'. You mentioned even the subcellular level. I stated in
response only that we
have to keep away from that level in order to avoid that we don't 'see the
forest anymore
through the trees'. That is in itself true but you are right, I should show
also how the link is
made between modelling the virtual image handling and at least the neuronal
We came to the conclusion in our research that we can define this link for a
small number of
standard neuronal assemblies from which the whole brain structure is
constructed by evolution.
An important aspect of this link is the following set of modelling
1  The brain comprises some 10^12 neurons.
2  Each neuron is, on average, connected to some 10,000 other neurons,
mostly in the
   immediate vicinity.
3  The cortex is organized in columnar entities of 0.75 mm diameter,
containing somewhat
   more than 10,000 neurons. The cerebellum is likewise organized in
4  The modelling procedure states / assumes / speculates, as the result of a
detailed analysis,
   that each columnar entity represents the neural equivalent of a standard
Hopfield network,
   as studied and defined in AI. Each Hopfield network comprises n nodal
points, each one
   connected to all other ones. This is a structure remarkably similar to
the columnar entity.
   For cerebellar stripes similar equivalencies are found.
5  Following the design data as derived in AI for such a BHN (basic Hopfield
network), it is
   found that each BHN can store 1400 vectors of 9000 bits each.
6  Together with the found 'chopping time' of 200 msec. (see earlier
response) this leads to
   the conclusion / assumption that in the primary sensory cortex these BHN
memory units
   function as 'shift registers' of the FIFO type (first in first out),
storing a 'short film' of most
   recent observations of environmental events of some 1400 x 0.2 sec = 5
minutes duration.
   This information system is therefore 'conscious' of the most recent
events, enabling it to
   pursue a course with a certain goal, such as chasing a prey, following a
debate, etc.
7  A further analysis leads to the modelling statement that a large number
of these BHN's (the
   cortex contains some 150,000 per brain half) is organied in some
hierarchic structure, that
   store together the information items of the virtual image.
8  Once you have reached this stage, you can then model the virtual image
handling processes
   by referring only to storage- and recall facilities of these BHN's and
its superstructures.
   One can ‘forget' then all more detailed information handling at neuronal
level in the same
   way as a programmer does. He also considers only his programs and files
and does not
   bother to consider the much more detailed instructions at processor level
or in read/write
   operations on disc.

I hope this clarifies this aspect,
Frans van der Walle

More information about the Neur-sci mailing list