Ed Porter wrote:
Richard,
Any ideas on what do such 10^5 neuron cortical columns represent? For
example, how do they relate to the multiple topological maps in the brain?
They all represent just one concept, or do they represent families of
related concepts? Or perhaps do the neurons of each maxi-column act in
unison as one cell assemblies, so that subsets of the roughly one million of
them can be selected to create unique coding for any one of millions or
billions of episodic or pattern memories?
Is there relatively clear demarcation between such columns (i.e., a
surrounding surface were the number of interconnects are much lower than
similar surfaces within the column), or is it such that any given location
on the surface of the cortex the neurons directly below it would be
connected to the nearest 10^5 neurons?
These are big questions.
I think it is fair to say that nobody knows what they represent or do.
The columns are all over the cortex, so in some places they will be in
regions where we know there are topological maps. Beyond that, you just
have to remember that there all all kinds of different things going on
in these different areas, but the columnar organization seems to be
everywhere (although with significant column differences in different
locations... hence some of the disputes about what exactly is a "column").
The units do not act in unison (they are not synchronized).
Each columns is marked by strong connectivity within it, with less
significant numbers of connections to surrounding columns. So it is not
the case that connectivity is homogeneous across the cortex.
I have a hypothesis about what they do, but it is still tentative. I
believe that each one transiently represents a single concept ("atom" in
my terminology), with each column having a large repertoire of atoms
stored in a passive state, but only one that (at any moment in time) is
in the active state.
Also, I believe that the columns do several other things associated with
that basic feature.
FWIW, in the recent neuroscience paper I coauthored, I use this
tentative model to give an account of several unrelated pieces of data
from brain imaging studies. One of these is, I believe, quite
significant, but I am working on a more detailed examination of that
one: it might be that I can actually predict the results of a study
that Koch et al made (not sure yet, but it looks hopeful).
There: quick brain dump.
Richard Loosemore.
Ed Porter
-----Original Message-----
From: Richard Loosemore [mailto:[EMAIL PROTECTED]
Sent: Wednesday, November 28, 2007 3:25 PM
To: agi@v2.listbox.com
Subject: Cortical Columns [WAS Re: [agi] Funding AGI research]
Benjamin Goertzel wrote:
ED>>>>>I must admit, I have never heard cortical column described as
containing 10^5 neurons. The figure I have commonly seen is 10^2 neurons
for a cortical column, although my understanding is that the actual
number
could be either less or more. I guess the 10^5 figure would relate to
so-called hypercolumns.
The term "cortical column" is vague
http://en.wikipedia.org/wiki/Cortical_column
There are minicolumns (around 100 neurons each) and hypercolumns
(around 100 minicolumns each). Both are called columns..
Ben,
It is true that there is enormous ambiguity and confusion about the
definition.
However, a number of people believe that the best compromise values for
column size and number are 10,000 neurons per column and 1 million columns.
I came to this conclusion after looking into it for my own purposes,
then was gratified that Henry Markham quoted the same numbers:
http://video.google.com/videoplay?docid=-2874207418572601262
Henry Markram, EPFL/BlueBrain: The Emergence of Intelligence in the
Neocortical Microcircuit.
Richard Loosemore
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?&
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?&
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?member_id=8660244&id_secret=69878670-45c202