RE: SI Comparative Advantages [was Re: Free Will] (fwd)

Robert J. Bradbury (bradbury@www.aeiveos.com)
Wed, 18 Aug 1999 00:46:38 -0700 (PDT)

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

As a complete aside to this thread -- why are so many of the people in the extropian commmunity in Australia???

At any rate:
> >
> A whole universe, an M-brain, and there's nothing better to do? Come on!
> Felines sleep 12-16 hours a day because they are bozos.

Emlyn, I'll assume "sarcasm" on your part here. Predators, in general are *highly* adaptive and efficient hunting machines. Once you are "fed" and have done your reproduction "duties" what else is there to do? They were at the top of their ecological niche until the screwy primates came along. Even then the interpretations are clouded because the smaller ones seemed to have convinced a significant number of primates to spend money to put food on the floor in front of them.

The question comes down to *survival* in your ecological niche. If SIs know all the laws of physics, how the universe works, are able to watch & simulate any potential dangers, know that the "magical fantasy worlds" are nothing but senseless diversions for the uploads, etc. then *why not sleep*?

You, as a human, may have "curiosity", but the real question you have to answer is -- is there an endless supply of unanswered questions to feed that curiosity?

> OK, while we are constructing the SI's as predators, look at it this way.
> Imagine that they are all in fierce competition - there can only be one! -

You are assuming there are no "costs" to predation/competition, and when you are talking about relocating stars or transporting star sized fuel resources, I assure you, that is not the case. You should not expend more thought/material/energy than you can hope to derive from that expenditure.

There can't be only one! The most efficient one, puts the matter into the smallest possible physical space (to reduce propagation delays) and collapses into a black hole (thus there are zero). I only hope it broadcasts its results before termination. :-) Those "ones" with a more pragmatic perspective will stop slightly short of that, so there can be many of them.

> which there may be reasons for. Even if a small number of them think this
> way, the rest must at least adopt a strong self defense stance (I would if I
> had an SI gunning for me!!!).

You need to think of this in terms of the solution to a set of linear equations. There may be multiple solutions, but the solutions are "locally" optimal. So long as an SI has manipulated/optimized his local resource base, the ROI on expansion/conquest/etc. are highly marginal. Think about this for a minute -- what happens to your "thought" process when the delay between computational nodes on opposite sides of your power source (usually a star) takes days! Now, expand this "handicap" to *years* or *decades*. There has to be a tradeoff -- when the investment in time/material/energy to expand to "double" your thought capacity exceeds the 2x the time/material/energy required to compute the same result using your existing capacity, that is the point where you stop expanding.

>
> Say you stop thinking, and rely on automated (or smaller, slightly less
> intelligent) subsystems to keep an eye out. Meanwhile, another killer SI
> (KSI?) is thinking, thinking, thinking... it models your alarm subsystems,
> works out a way to attack you without alerting you (they are quite clever by
> definition), and good bye sleepy head! In general I think proactivity is
> more successful than a reactive stance in a competitive situation.

Maybe. It depends on how far away the killer SI is. If they are actively computing, you will see them, attempt communication and be able to determine whether they are "trustable". If they are "not thinking" (emit low levels of IR), then they are probably implicitly trustable/non-threatening. I will admit that in these situations I would consult the galaxy history books to attempt to verify the trustability of an individual SI.

You have to think of it like "The Transparent Society" by David Brin. Since everyone can watch your behavior, if you "misbehave" everyone knows about it and you become "untrustable". This creates a huge incentive to "behave" in a socially acceptable manner.

>
> Why would it do this? Maybe it can examine the corpse, steal the energy
> reserved for when you next woke up, and continue on it's high-energy
> strategy, off to nuke the next SI. Meanwhile it's also been able to devote
> substantial resources to some long term goal, like the fundamental meaning
> of the universe, or some similar guff.

You have to destroy all of my nodes and crack all of my security beacons to other SIs. If other SIs lose the "security" messages, then the entire region in which you are operating becomes fair game for radiation sterilization.

>
> Is it possible that SIs will discover that a lower level of processing
> (dumber self) which is sustainable over longer periods of time, such that
> the SI is constantly "awake", is more advantageous than short periods of
> mega-brain activity punctuated by long periods of sleep? Maybe some
> combination is possible; maybe that is what you might also mean by periods
> of inactivity - you would instead have periods of comparitive stupidity.

I definately think you can always be "awake" in the sense that you monitor external astronomical observations (in great detail) and determine if any of those changes represent a threat. I would say that survival favors the elderly. Looking up a survival strategy in a table has to be less expensive than computing a strategy around those that are already in the tables.

> I still don't like it. I can sleep 12-16 hours a day now if I am so inclined
> (not really, but that's out of choice too). Why would I transcend to
> (S^N)I-hood, just to go for a trans-galactic snooze?

Simply because you want to see if it "all turns out". You may not believe any of the Quantum Computing hype, but wouldn't it be interesting if it contained an element of truth.

>From my perspective, I'm intensely curious whether we manage to plot
a course through this swamp.

Robert