From: Eugen Leitl (eugen@leitl.org)
Date: Tue Sep 10 2002 - 10:29:43 MDT
On Tue, 10 Sep 2002, Robert J. Bradbury wrote:
> But following the reproduction vector means making a conscious
> decision to limit the development of the vector you are on. Unless of
It is not obvious whether incremental Lamarck of semimmortal individuals
or discontinuous Darwin or a judicious mix of these two will be the
winning strategy long term. I think we will have to be there to be able to
tell. The costs for reproduction are certainly not high, if you're living
on the edge of the wilderness. Of what use to me personally is an
adjacent-star resource, which is several megayears remote in subjective
terms? You have to be there in person to be able to use it.
> No. You are assuming the offspring make the same choice as the
> parents (to reproduce) -- there is no "law" that says that will occur.
Evidence throughout the animal kingdom would seem to imply there is. The
"law" is in fact too trivial to formulate, offspring that consistently
fails to reproduce lacks the crucial property of existance. We're children
of survivors. No wonder we consider sex and procreation a basic fabric of
being.
> Further you assume that the non-reproducers do not execute a system
> claim and sterilization strategy. One could easily execute a seed
In a Singularity scenario, ascendence to the ceiling (or the first kinetic
bottleneck, at least) is relatively short in wall clock seconds. I.e. if
you're a nonexpansive ceiling civilization which has expansive neighbours
you can't sterilize them without having been lurking in the local system.
By the time you get there there would have about the same technology, and
the logistics advantage. I.e. you would have to pursue a very rapid but
stealthy expansion, leaving sentinels all over the place. The time to
react, using your initial advantage is short enough so that you have to be
already in-system.
This scenario implies that it's probable the sentinels are already
in-system, and will soon squish us, or at least will start using force to
keep us down if we start getting too uppity.
> strategy which populated a system not with "intelligence" but with
> seeds capable of eliminating any incoming not-from-the-same-tribe
> seeds.
You would have to express the probes at a sufficiently high level to
maintain a local advantage, yet not sufficient to wind up as readings on
our instruments. This is clearly possible. But this also looks very
construed to me, especially since we're lacking a motive. We're
invulnerable in our local system. If we're nonexpansive, we don't care
what happens elsewhere. Why, also?
> This approach makes sense if sending resources back to a parent system
> for use is expensive. In that case you want to "claim" the resources
You know very well that resource transport over interstallar distances is
ridiculously expensive, according to known physics. That's the reason
probe design strongly favours molecular-scale components. The only
resources transported will be most likely information (including
information encoding individuals: travellers).
> so others don't walk off with them but leave them where they are until
> one gets sufficiently close to utilize the resources without wasting them.
The parent doesn't consider resources utilized by her child wasted. The
opposite, in fact.
> You don't -- you only take over matter that within your light horizon
> can be ascertained to be "unallocated". Once a system is occupied
You would have trouble trying to annex already allocated territory. In
fact there would be no point for you trying, if the occupants pursue
exactly the same strategy (very rapid but stealthy expansion; active
maintenance of a low sentinel population).
> the energy/matter within trump any incoming seeds so colonization is
> effectively prevented.
>
> > Probably there is only one way to do it. You must replicate.
>
> Yes, but you don't need to replicate much to trump anything that
> can be sent in your direction.
Assuming the development ceiling hypothesis is true, expanding wavefronts
(whether stealthy, or runaway replicators) can't penetrate each other,
cancelling each others out.
> Eugene has claimed this but I don't think the case is that solid.
> The microwave strategy doesn't work well at all (due to beam divergence).
I'm not sure there is much beam divergence over lightmonths distances if
your instrument aperture (whether microwave or optical) is lighthours to
lightdays across, and you have the entire stellar output at your
fingertips.
> A pellet strategy would work better but I'm still questioning his
Are you proposing linear accelerators? It would be interesting to see
whether you can bring up a ~kg packet to relativistic speeds using a queue
of aligned segments. They're in orbits, though. The advantage of a cloud
of active radiators is that you don't have to force their orbits, just
change the direction of reradiated energy.
> shielding approach (the calculations Spike once did suggested this was
> an expensive proposition).
I'm not sure shielding is necessary. Even if, I fail to see what is so
expensive hiding a liter of living circuitry in the shadow of a tungsten
rod pulled by a sail.
> More matter & energy trumps less matter & energy. Sending matter
> and energy at high velocities is expensive. Whomever occupies
> the system first presumably trumps any latecomers.
Yes, I think these are very reasonable assumptions, assuming the ceiling
thing is true.
> Distributed backups don't really work over interstellar distances.
Interstellar distances favour ACK-less protocols, suggesting treating
vacuum as a FIFO, firing bits as tracer bullets, with some encoding
redundance in case you get transmission errors. As such it would be
feasible to maintain a remote backup (as a static, incrementally updated
image).
> Claiming resources seems to make sense. But then you aren't creating
> "offspring" that represent future competitors.
Why competitors? You said yourself they won't be able to enter this system
in hostile intent, due to simple logistics reasons. Why fight, if we can
trade and travel (and have sex)?
> Of course if there are limits to the effective size of intelligences
> then claiming more matter may be relatively pointless. As I pointed
> out in the MBrain discussions -- one wants to get smaller not bigger.
Yes, but a hierarchical model would allow very rapid local activity and
increasingly loose-coupled long-range activity. In a sense, this is a lot
like reflexes (ms) vs. higher order processes, which can take several
seconds. It's a somewhat different architecture, because we don't see
mammals the size of mountains, or covering whole continents. They would
twitch, if you poke them locally, but they would be pretty slow
conversationalists.
> Perhaps our perspective of going out and claiming resources is an
> inherently anthropomorphic perspective that is due to our very
I think so.
> limited size scale perspective (mm to km). Once we become more
> virtual we may discover that it is really rather irrelevant.
I still fail to see us suddenly frozen into stasis as we approach some
virtual circle. 1.25 lighthours. Thus far, and not further.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:16:55 MST