RE: SI Comparative Advantages [was Re: Free Will]

Robert J. Bradbury (bradbury@www.aeiveos.com)
Tue, 17 Aug 1999 09:03:33 -0700 (PDT)

> O'Regan, Emlyn <Emlyn.ORegan@actew.com.au> wrote:

> ah, if you have gone to the trouble of becoming an SI, why would you
> then snooze on the job?

The same reason that predators (lions/cats) at the top of their evolutionary niche sleep 12-16 hours/day -- because there is nothing better to do!

> There's always something you can do with CPU/brain time.

Yes, you can always run the VR for the uploaded minds. The trick of it is, since all of their senses are fed in anyway, they have no way of knowing whether they have been suspended for 1000 years and are just being fed 1000 year old pictures of the universe! I envision an interesting game of cat & mouse -- the SI "scheduler"/ sensory management subroutine attempting to convince you that you have been awake the last 1000 years, while you (an uploaded mind in a small region of the SI) attempts to prove that you have only been given 3 minutes of CPU time in the last 1000 years!

> And wouldn't a permanently awake SI have an advantage over
> a strategically sleepy one, just because it's thinking more?

I think it depends on the rate of "challenges". Remember what dictates our current longevity -- tradeoffs between energy devoted to reproduction vs. energy devoted to maintenence & repair. If the rate of challenges/hazard function is "high", then thinking more has its advantages. If the rate of challenges/ hazard function is low, then sleeping may be the preferential strategy for survival (think of the mouse vs. the lion). I think this works out so that SIs in high density regions (stars, other SIs, black holes, etc.) prefer to think, while SIs in low density regions prefer to sleep.

> Or do you mean that doing "nothing" means doing entirely internal
> thinking, ie: not interacting with the outside world, but CPUs at 100%?

No, if Robert Freitas is correct (in his communications to me), that SIs will dismantle Jupiters/brown dwarfs/stars for long term fuel requirements, then your rate of fuel consumption is entirely self-determined (so you can sleep "between" gas stations). In other words you have an entirely "self-controlled" rate of fuel consumption, thought rate, heat production/signature.

If my thinking is more correct, that fuel management is "expensive" (from a material resources perspective) for "young" SIs, then you want to run "full-out" using all of the material/energy resources available, until you get to the point where you either (a) have computed future paths with a high degree of confidence or (b) have computed everything of interest to you and are willing to sacrifice the computational resources to energy management (fuel tanks, star atmosphere harvesters, etc.)

You always have the ability to turn the CPU clock rate to zero, but you don't have the ability to turn off the star (and therefore the energy production). So in even a moderately engineered system you will radiate heat (and waste energy). Only if you go to the trouble of constructing what I call Externally Powered SIs (EPSIs), that use materials harvested from Jupiters/Brown Dwarfs/Stars, can you completely control both the though rate and the waste heat production rate.

> I just don't like the idea that SI's work like an office photocopier
AH, but an office photocopier doesn't rate the "entertainment value" or the "survival value" of the material it is copying. If, you as an SI had the ability to judge these factors, then presumably you would optimize your operations to maximize these values.

Robert