194
PARTISAN REVIEW
mess of things? The answer is to send the scanners to scan the brain
from inside by way of the capillaries. Now, this is a scenario we can
touch and feel because we could actually build these devices today,
except we can't make them small enough: they need to be the size of
blood cells or smaller. We actually have the technology to build them
pretty small today. For example, there are devices called "smart dust."
About one millimeter in size, these tiny devices have computers in them
and can fly, communicate, take pictures and other measurements, and
can be dropped in an enemy terrain from an airplane. So we're already
building extremely tiny devices, but we still can't build the blood cell
size that I am describing, yet. And if we could, it would be very expen–
sive. But these are exactly the kinds of capabilities we can reliably anti–
cipate in terms of the law of accelerating returns.
As I mentioned, we're shrinking technology at a rate of 5.6 linear
dimensions per decade and we're exponentially raising the price perfor–
mance of computing. Conservatively put, this scenario will be feasible
within thirty years. At that time, we will be able to send billions of lit–
tle nanobots, or nanorobots the size of blood cells, inside the human
brain. They would be programmed to take amongst them cumulatively
every single path through all the capillaries, travel in physical proximity
to every single neural feature, and scan and build up a massive database
of every single thing, every salient neural feature in the brain. They'd all
be on a wireless local area network, so they'd be communicating with
each other, sharing their information, communicating with computers
on the outside that are compiling the database, and we'd be able to get
a dump-a database, a download-of every single salient feature in
the brain, including all the neurotransmitter concentrations and the
results of all our experiences and so on.
Now, what are we going to do with that information? Well, for one
thing, we'll learn a lot about how the human brain works. Having this
massive database doesn't immediately mean we understand how the
brain works; we'd have to begin to understand those processes. But that's
not impossible because we have, in fact very accurately, decoded many
different types of neurons. Carver Mead at Cal Tech has built very com–
pelling replications of different neural systems, including the early visual
system, the early auditory cortex. His vision chip is now being built into
advanced digital cameras. Recently, at the Institute for Non-Linear Sci–
ence in San Diego, scientists actually took a biological set of neurons
from an animal, a lobster, and built a detailed mathematical model from
it-a much more complex model than the simplified one used in con–
temporary neural networks. They then built electronic replications of
