> From: "Gina Miller" <nanogirl@...>
> My estimates come from Rob Freitas:
>Consider that a nanostructured data storage device measuring ~8,000
>micron3, a cubic volume about the size of a single human liver cell
>and smaller than a typical neuron,
>could store an amount of information equivalent to the entire
>Library of Congress. If implanted somewhere in the human brain,
>together with the appropriate interface
>mechanisms, such a device could allow extremely rapid access to this
Gina, I've always used 1 Library of Congress (LOC) ~ 40 million volumes x 150,000 words/volume x 6 bytes/word x 8 bits/byte ~ 3 x 10^14 bits. Philip Morrison [SciAm(July98):115] claims the 1998 LOC has only 20 million volumes but that there are also ~2 x 10^15 bits of sound recordings -- assume 10:1 data compression of the sound for nano-storage, and our total LOC figures agree. Or make your own estimate. Below, I'll use mine.
Assuming fluorocarbon tape information density of ~26 bits/nm^3 [NMI:178], 1 LOC can be stored in ~12,000 micron^3 (~one large-ish human liver cell volume). Using Drexler's slightly more conservative ~5 bits/nm^3 for tape storage [NS:366], 1 LOC can be stored in 60,000 micron^3 (~one very fat macrophage cell volume).
Assuming rod logic register storage with information storage density of ~0.025 bits/nm^3 [NS:357], 1 LOC requires a storage volume of 0.012 mm^3, a cube ~230 microns on an edge.
>>A single nanocomputer CPU, also having the volume of just one tiny
>>human cell, could compute at the rate of 10 teraflops (1013
>>floating-point operations per second),
>>approximately equalling (by many estimates) the computational output
>>of the entire human brain.
> On page 370 of Nanosystems, Drexler talks about a 10^15 MIPS (roughly
> 10^21 operations per second) machine with water cooling as being a
> cubic cm in volume. Merkle estimates the computational power of the
> human brain as "It seems reasonable to conclude that the human brain
> has a raw computational power between 10^13 and 10^16 operations per second."
> Thus the 1 cubic cm computer is roughly a megabrain computer.
> Since 1 cubic cm is 10^12 cubic microns, there is one brain per 10^6
> cubic microns. Using an estimate of 10^4 cubic microns for a larger
> human cell, that would be one brain per 100 cells. Thus there is a
> difference of 100 between Drexler and Freitas, probably because
> Freitas is using less conservative assumptions, or assuming molecular
> electronics, etc.
Drexler's 10^15 MIPS/cm^3 figure [NS:370] equals 10 teraflops per 10,000-micron^3 (mammalian cell size) volume. Ten teraflops is simply Merkle's lower 10^13 ops/sec brain-capacity figure. If instead you use Merkle's upper 10^16 ops/sec brain-capacity figure, then you need 1000 tissue-cell volumes, a cube ~200 microns on an edge, roughly the size of a human ovum (also a mammalian single cell). Take your pick. AFAIK, there is no "difference of 100" involved; our assumptions are identical.
Robert A. Freitas Jr.