From: mez (mez@mezziah.org)
Date: Sat Jan 20 2001 - 20:30:33 MST
From: Max More [mailto:max@maxmore.com]
> If I were Kurzweil, I'd be annoyed. You say his math is
> sloppy then you
> sloppily misrepresent his numbers. Someone borrowed my copy
> of his book, so
> I'm going by an early pre-publication draft, but it only took
> me a couple
> of minutes to find the correct numbers. You are off by four order of
> magnitude. He gives a "conservative" estimate of 2 x 10^16, not
10^12.
Whoops. Mea culpa. Thank you for pointing out my sloppiness and
misrepresentation of Kurzweil.
That having been said, I still take issue with his math, actually on a
couple fronts:
1) Brain computational power. My math is as follows:
~10^14 synapses
10^2 maximum firings per second
10^1 conservative oversamping (analog data)
~10^1 operations to calculate the effect of potential firing
Results in a very conservative number of 10^18. In all likelihood
this number is higher, as the above math:
a) Doesn't factor any potential impact of glial cells
b) Doesn't leave many computations for each neural firing
c) Doesn't deal with synaptic plasticity
d) Doesn't deal with neurotransmitter reuptake,
competition at binding sites, etc..
e) Has a very low margin of oversampling the analog firings of
neurons
So my personal bet is that the actual number is more like 10^20
operations per second or higher.
2) Rate of computational power growth. In _Age of Spiritual Machines_
(my copy is also on loan so I will again go out on a limb) Kurzweil
projects computational power growing by a factor of 10^3 every decade.
The industry rule of thumb, however, as well as current data, support
a more modest but still extraordinary growth rate of 10^2 every
decade.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:05:08 MST