X-Message-Number: 3038
Newsgroups: sci.cryonics
Subject: Re: An irreverent introduction to Cryonics
Message-ID: <>
From:   (Joe Strout)
Date: 28 Aug 94 11:55:49 -0500
References: <>

In article <>,  (Brent Fox) writes:

>         Have you seen an estimation of what the total information storage
>  capacity of the human brain is?   In Carl Sagan's _Cosmos_, he stated
>  that the brain has on the order of 10^14 neural connections in the cerebral
>  cortex.  Assuming each connection would be equivalent to 1 bit, 
>  taking this further would yield a capacity of 12.382Tb.  

>               ... (calculations omitted, but thanks for including)

>         I would be interested in hearing your thoughts on this.

Your arithmetic looks good, but I'd say that 1 bit/connection is a
gross underestimate.  Consider that there are many different types of
connections, which differ in their natural (unstimulated) firing rate,
effect on the postsynaptic (next) cell, in their directed or diffuse
effects, etc.  If there are, say, 256 functional classes of synapses,
then that's 8 bits already.  Then there's connection strength -- i.e.,
how much influence the connection has on the postsynaptic neuron. 
That's an analog value, so the question would be: into how many
discrete steps do you want to break the continually varying values? 
It's the same question that the music industry faced when inventing CD
players: sound pressure is a continuous variable, approximated by
32-bit values (if I remember correctly).  So we're up to 40 bits or
so.  Then comes the real problem: it is extremely relevant what these
connections connect.  It's not a uniform pattern; it's circuitry, and
so you have to take into account some way of specifying where the
synapse "comes from" and "goes to".  That is likely to increase your
bits per connection by an order of magnitude or two.

Of course, this is mostly loosey-goosey speculation at this point, but
I hope to be doing research soon to help determine exactly how much
information needs to be stored to represent the functional properties
of real neural networks.  Keep an eye on the neurosci groups for more
details as they arise.

-- 
|o|    /////    Joe Strout                   |o|
|o|   | @ @     Neuroscience                                               |o|
|o|   C  _)     U. C. San Diego                                            |o|
|o|    \ o      La Jolla, CA                                               |o|

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=3038