SV: Capacity of the brain

Robert Herndon rmhj at rmi.net
Thu Sep 2 17:34:46 EST 1999


In comp.arch Russell Wallace <manorsof at iol.ie> wrote:
> Lawrence Woods wrote:
>> 
>> I recall an old Reader's Digest space-filler from my youth (late 50's or
>> early 60's) that claimed a computer with the capacity of the human brain
>> would occupy a building the size of the Rockefeller Center in New York,
>> and require the entire East River to cool it.
>> 
>> I wonder how big that Rockefeller Center computer would be using
>> today's technology. Any ideas?

> They were being excessively optimistic; back in those days, the
> complexity of the brain wasn't fully appreciated.

> In round numbers (assuming our current understanding of how neurons work
> is more or less correct), here are the raw figures for the human brain:

> Storage capacity: 10^15 bytes
> Processing speed: 10^18 calculations per second

The original quote (from memory) is attributed to Vannevar Bush, and 
involved the Empire State Building and 3 Niagaras to cool it.

The empire state building has a volume of about 37 million cubic feet.
Assuming an acorn tube or similar requires 5ml volume (err on the small
side), that yields about 1.05*10^12 gates, neglecting power cables and
supplies, coolant channels, pumps, etc..

Top-end processors these days seem to run 10-20 * 10^6 transistors,
so 50-100k CPUs should cover it.  If you want to include memory
chips, you can divide by up to another 10 or so (DRAMs have lots of
transistors).

As Lawrence Wood wrote, a lot of the complexity of the brain was
underestimated.  For many years, and still often today, claims are
made that the human brain has about 10^10 neurons.  To paraphrase
a joke among doctors:  Yes, the brain has about 10^10 neurons.
And 10^11 of them are in the cerebellum.  (A section of the brain
near the brain stem that deals with coordination.)

I would presume that Vannevar Bush was familiar with the idea that
the brain had about 10^10 neurons, and was presuming equivalence
between neurons and tubes/gates.  This isn't a fair comparison,
however, as neurons often have up to ~1000 connections, and
these connections are neither simple, symmetric or equivalent to
each other, nor immutably fixed in character.

/R



More information about the Neur-sci mailing list