RE: Nanotech Arms Race

Billy Brown (bbrown@conemsco.com)
Fri, 22 Jan 1999 09:04:59 -0600

Michael S. Lorrey wrote:
> You own implicit assumption which I find weak is that there needs to be an
AI to
> run this system. The thing with nanotech is that you can give the little
buggers
> simple instructions to fulfill, and when done, die. You also assume that
if
> there were an AI, that it would, could, or couldn't be restricted from,
> developing enough independence to prevent the mad plot to blast the
planet.
> Computers are, despite their power, amazingly stupid devices when they
don't
> have good software. Even the smartest ones can easily be programmed to
take no
> initiative (or, rather, just don't program it to take initiative).

Let's stick with the same example: "Computer, go gather two million comets from the Oort cloud, make a giant projectile out of them, and smash it into the Earth." What does the computer have to do to carry out that order?

Well, first it has to figure out how to do the job. That means inventing a whole range of spacecraft, construction robots (nanoscale or otherwise), power systems, giant propulsion systems, etc. Then it has to co-ordinate all the work - so we add in a system-spanning communication system, all kinds of sensors, and a huge computer network. Oh, lets not forget - it has to program all those robotic gadgets, plus the computers that will coordinate their work.

Then it builds the initial seed and sends it on its way - oh, wait, we also need to invent an Earth-based launch system and figure out where to aim it (better build some astronomy gear). When the seed finally arrives on site it starts building - and unexpected things happen (the unexpected *always* happens on construction projects). So the software running the project needs to be smart enough to adapt to unforeseen problems, inventing solutions along the way to its goal.

Add it all up, and you have a system that has all the intelligence of a human engineering & construction team. It can do pretty much anything a human could, but millions of times faster. It doesn't really matter how it works internally - it could be an algorithmic AI, a giant neural net, an ecosphere of evolving programs, or a mixlplic of quasitronic gimrods. From the outside, the effect is the same as having a lobotomized SI.

I make no assumptions about the nature of the AI - I just pointed out that if it already does everything a sentient can do *except* take action of its own volition, it won't be long before someone adds that last missing piece to their copy of the program. I do think that you would have to invent a fully sentient AI before you could figure out how to make one that can act sentient without having a sense of self, but that isn't central to the argument.

> A missing factor in these blast calculations is seismic. Get enough
impacts
> going at once and the shock waves will turn the entire lithosphere (the
crust)
> of the planet to lava. Diamonds burn just fine at such temperatures.... We
know
> that this is an effective strategy because the Alvarez asteroid which
killed the
> dinos created a shock wave that went around the planet and focused in the
middle
> of the Indian Ocean (where the island which is now India was at the time)
and
> turned the area which is now the Deccan plateau to lava.....
>
> I would say that 20 asteroids or comets about 20 miles in diameter
impacting at
> once in a dodecahedral pattern would do the trick....

Sorry, no. That's a fantasy. It takes around 10^18 calories to melt one cubic kilometer of rock. Your asteroids carry about 10^21 cal apiece, so the whole flock could melt 20,000 cubic km - which is less than their own combined volume of 81,000 cubic km, and insignificant on the scale of a planet. In practice most of the energy gets dispersed over a large area in the form of flash, shock wave, flying debris, and so on, so the actual melting effect is even smaller.

Billy Brown, MCSE+I
bbrown@conemsco.com