skaggs at bns.pitt.edu
Tue Mar 11 11:23:56 EST 1997
flefever at ix.netcom.com(F. Frank LeFever) writes:
> (1) no one knows how info is stored in the brain (I repeat: no one), so
> one cannot calculate it rationally--e.g. by estimating the number of
> cells and their interconnections and their dendritic spines and their
> multilevel (non-discharge) graded states, etc., etc., etc.
It is commonly assumed that information is stored by changing the
strength of synaptic connections. There is no proof of this, but the
case is strong enough that many of us are willing to bet our lives
(our scientific lives, anyway) on it. If it is true, then the most
straightforward approach is to ask how much information can be stored
in a single synapse, and how many synapses there are. Counting
synapses is relatively easy -- there are about 10^14 of them in the
human brain. Estimating the information content of a synapse is more
difficult, but the plausible range of possibilities is not enormous.
So, all in all, I don't think this is an unapproachable problem.
> (2) no one, so far as I know has approached this problem "empirically",
> e.g. by extraplating from behavioral samples, except in a very limited
> realm. Last week, at the NY Academy of Sciences (Linguistics Section
> meeting), Steven Pinker of MIT reminded us of studies in which people
> sampled a large dictionary (e.g., every 10th word on every 5th page) to
> form vocabulary tests, yielding estimates of the total vocabulary of
> high school students. If you can think of a way of randomly sampling
> EVERY POSSIBLE KIND of info, maybe you can come up with an estimate of
> how much your subjects have "stored"..
Well, there was a guy at IBM, back in the 1950's or 1960's, who tried
to do precisely that. I wish I had written down the reference,
because this question comes up surprisingly often, but unfortunately I
didn't. If anybody knows it, I would be delighted to have it. In any
case, as I recall, the value he came up with was something on the
order of 10 megabytes (which I think must be a radical
underestimate). He did, however, assume optimal compression.
-- Bill Skaggs
More information about the Neur-sci