Re: The Extinction Challenge

Robert J. Bradbury (
Tue, 27 Jul 1999 14:27 PDT

> In a message dated 99-07-27 15:37:42 EDT, Billy wrote:
> > I hereby challenge anyone to come up with a disaster, natural or
> > artificial, that could reasonably be expected to render an advanced
> > spacefaring species completely extinct.
> >
> > For purposes of clarification:
> > 1) The target civilization is assumed to have self-sustaining colonies in
> > several solar systems spread across a volume of at least a few dozen cubic
> > light-years.

This isn't necessary. I believe that a "near"-nanotech/AI capable civilization based on a single "home-world" cannot be eliminated if it has a "minimal" defense posture (requires a strong belief or actual knowledge that non-friendly ETs may exist). [We do not currently have such a minimal defense posture.]

> >
> > 2) The target civilization is assumed to have advanced robotics, genetic
> > engineering, non-sentient AI, and other such technologies. Just to give
> > exterminators a sporting chance, we'll pretend that nanotechnology and
> > sentient AI don't exist.

This makes no sense. Nanotechnology has a high probability of development for tech-civs (since it is useful). Sentient-AI should also develop unless actively prevented due to the dangers it may pose. In any case non-sentient AI is better for "reconstruction"/"revenge" than sentient AI since sentient AI can change its program. If you want to impose these restrictions, you probably need a fundamentally different universe with different physical laws (to make nanotech or AI very difficult).

> said:
> Not fair, Billy. If I can't use nanotech and at least minimal AI, I can't
> wipe out the target. WIth them, I think I can.

You can probably wipe out the "primary" target(s), but unless you want to remain on the run for the next ump-teen billion years, you have to do atomic disassembly, radiation inactivation or temporal-spatial removal of *ALL* possible revenger-bots in the system. Since a revenger-bot can probably be on the order of a few microns cubed and you easily have trillions of them in suspend mode. Eliminating them is very difficult. Atomic disassembly/irradiation take lots of energy and time. Temporal-spatial removal (by importing black holes) probably generates gravity waves that provide advance warning.

The challenge does lead to another interesting question:

What is the most "defensible" architecture for an SI?

I would argue that it is the most "dispersed"/"stealthy" configuration. That would seem to imply that the most "natural" (unengineered) systems are those that would provide the greatest protection. There would be nanotech micro-SIs scattered all over the solar system being very careful to make all energy absorption & emissions look natural (e.g. heat from the gravitational collapse of Jupiter or a planet's molten core). Communications would have to be low-energy, directed and as white-noise-like as possible. It is clear that there is a trade-off between maximal intelligence and minimal detectability.

There must be an interesting divergence point in the development of SIs/ETCs where they decide to go the visible/stealth route. If you are "visible", you want to have available huge energy stores, massive seek-and-destroy capabilities, revenger-bots for may parsecs around your stars/worlds, etc. If you are "stealthy", you want to develop slowly, leaving things just as they would normally appear to an outside observer.