Dan Clemmensen wrote of assorted ways to depopulate Earth using nanotech.
I'll reply to his suggestions in another post - I've got a related point to
make here:
I believe I have detected an interesting implicit assumption in your
scenarios, which is common to nanotech doomsday theories but is not at all
realistic. You seem to assume that your nanotech is controlled by a
computer system that can obey commands like "Go grab 2x10^6 comets from the
Oort cloud, merge them to make a giant projectile, and then smash it into
the Earth." The computer obligingly does so, with no muss or fuss.
Now, the AI that runs that computer can already do everything an SI could.
Even if the AI can't be made sentient for some reason, this is still an instant-SI scenario. "Upload me" isn't any harder a command than the comet project, after all. There are several other easy paths to SI as well - you could build a neural interface and integrate your own mind with the genie machine's design ability, for example.
If you want to talk about nanotech sans SI, that necessarily means there aren't any genie machines. That in turn means that many types of construction still involve large numbers of people and a lot of money - your nanofab simply builds whatever designs you can buy for it. The resulting world is very different than our own, and it is more unstable, but it isn't nearly as bad as the one you fear.
Billy Brown, MCSE+I
bbrown@conemsco.com