Anders Sandberg (
Fri, 1 Aug 1997 14:04:04 +0200 (MET DST)

On Thu, 31 Jul 1997, Dan Clemmensen wrote:

> True, but I think the original question is: what is the
> smallest computer that is still capable of flight control
> in addition to other useful behaviour?

Small insects have around a thousand neurons. Assuming 10^3 Hz update
frequency and 10^3 synapses each, we need on the order of 10^10 flops
for uploading a real time insect. Not that far from what we can do
today, although it would be impractical. This is a high upper bound
on the complexity of an insect animat; most functions can be done
much more easily as can be seen in robotics labs over the world. So
my guess is that a few megaflops of computing power might be enough.
The size of the computer will depend on the available technology;
today it would be hard to make a clever enough insect-animat, but in
a few years it should be quite doable.

> Now, I don't know how far we can extend this reasoning.
> Just how "smart" can an intracellular probe be? Well, a
> probe the size of a bacterium should be at least as "smart"
> as a bacterium, but is that smart enough to do cellular repair
> an avoid damaging the cell?

Smarts isn't really needed, cunning is better. Insects are stupid,
but they can do great evasive flying and very complex manipulation
based on simple, interconnected responses. A cell repair device could
just move around randomly, avoiding things that could be damaged
by it until it found something that was damaged, then it would fix
it. No more brain than a simple insect; I guess you could get that
power into a 100nm side cube.

Anders Sandberg Towards Ascension!
GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y