[ExI] Von Neumann probes for what?

[Anders Sandberg]

On 2010-12-31 23:47, Samantha Atkins wrote:
> Eugen and others,
>
> What exactly do we expect these probes to do when they reach a workable planetary system?

The model I have been looking at would use a laser-powered solar sail to 
accelerate and deaccelerate, find suitable Kuiper belt objects to mine, 
build an industrial infrastructure, construct laser-launchers and more 
probes, and then do whatever other things the designers wanted.

In many ways they are a Swiss army knife: they can be used to set up 
discreet listening stations, wipe out potential competitors, build 
defenses against malicious probes, industrialize the whole system to a 
M-brain, terraform or do anything else. Depending on the goal different 
levels of smarts are needed. I think the basic propagation system 
doesn't have to be smarter than an animal, but obviously the probe might 
contain AGI for bootstrapping more complex projects. However, probe 
programs could be made unalterable or unevolvable (consider good 
checksums to check mutations, and unintelligent probes building 
infrastructure and moving on before turning on the local intelligence).

Listening posts are interesting: we found that even if there is a finite 
failure rate per year, if you have self-replicating installations they 
can keep the probability of all surviving *indefinitely* positive by 
reproducing at a logarithmic rate. So you would have a few extremely 
sparse and hard to find installations out there, lasting billions and 
billions of years.

Probes are also great for making buffer zones. If advanced warfare 
follows the Lanchester equation (a big if), then you want a numerical 
advantage. So you convert some resources into hidden depots of defense 
<whatevers>, and wait for invaders. The real problem is when you get 
conflicts between expanding replicator clouds. I haven't finished my 
work on this, but it looks like there are endless war solutions where 
resources get used up but the conflict never ends. Also, in space buffer 
zones don't work well since you can always aim straight at the center 
without going by intermediate systems (some caveats here on survival 
probabilities of probes and weapons on long flights).

I think a probe infrastructure could be something that just looks like 
added value to a civilization. It launches the probes, they spread and 
set up waystations that can receive instructions and mindstates, as well 
as send back observations. If they want to use the system they can. The 
key limiters are whether the cost of the initial probe is high relative 
to the civilization GDP and whether the time horizons of *every* entity 
within it are so short there is no value in getting a fraction of the 
galaxy in the far future.


-- 
Anders Sandberg
Future of Humanity Institute
Oxford University