>From: hanson@hss.caltech.edu (Robin Hanson)
>Date: Tue, 15 Oct 96 10:41:50 PDT
>Subject: Great Filter, Low Profile, Cryptocosmology
>
>[...]
>
>If one looked at the inards of a very advanced computer, then if one
>intercepted a "long" communications channel, where the cost to
>transmit was large compared to computation costs at either end, then
>yes it should look like random noise relative to a high entropy state
>of that channel. But this observation is a long way from showing that
>the innards of this computer should look like the stars and dirt we
>see.
Well, right, that's even less than circumstantial. Motive and method
are also needed. I'm saying the motive is just that hiding is necessary.
Whether there's a possible method...because we're talking
about civilizations more advanced than our own, we can only ask, can we
prove that it's *not* possible to be computing with (what looks like)
dirt?
By the way, in case I wasn't clear, it isn't that dirt, or any of our
particular surroundings, are what I would expect to see. "Dirt" is just
my code word for whatever science would lead us to expect are natural
things to see at this stage in the evolution the universe.
If looking like "dirt" is a great advantage, and it's possible,
then I would expect advanced civilizations to look like dirt.
>As you point out, all that sunlight streaming away seems
>inefficient.
Yes. But maybe that's just the cost of hiding. Our civilization is
certainly paying the same price without much complaining. Or, and this
is something I've been thinking since, sunlight may be the dumping of
waste heat. A little more on thermodynamics below.
>I'd expect an advanced computer to notice if some external force
>started to muc[k] around with its innards, and to do something about it.
Sure, "route around it," in John Gilmore's immortal words. "To see
the internet in a grain of sand..."
>So if the world around us is an advanced computer, it knows about us
>and chooses for some reason to allow our activities to continue.
>Either we are part of this computation, or it works around us.
My assumption is that our activities are so gross and slow that they
pose no problem. Even we humans know how to do distributed computing,
redundancy, moving running processes between processors, etc. Anyway,
I think we would be part of the computation, whether intended
or noise or bugs. We don't seem much more troublesome than, let's say,
rainstorms, though. How to compute with dirt seems much harder than,
"Given I can compute with dirt, what if it gets up and walks?"
>How much physics do you know Steve?
"...Clever people may learn as much as they wish of the results of
science--still one will always notice in their conversation, and
expecially in their hypotheses, that they lack the scientific
spirit; they do not have that instinctive mistrust of the
aberrations of thought which through long training are deeply
rooted in the soul of every scientific person. They are content
to find any hypothesis at all concerning some matter; then they
are all fire and flame for it and think that is enough... If
something is unexplained, they grow hot over the first notion
that comes into their heads and looks like an explanation..."
--Neitzsche
Say, college freshman level, polluted by popular accounts. Including
doing my best to understand things like thermodynamics, information
theory, and reversable and quantum computers.
>I'm not sure you realize how very
>much we do know about all that information streaming around us.
Well, not enough, but tell me the fastest-reproducing human cell and
I can compare it to a writable CD-ROM drive...
>We
>know which systems are exchanging bits with each other, and how fast,
>and where this arrangement changes, etc.
Wow, do we? How does it compare to the computing cost of *simulating*
events at the quantum level, or even at the classical-physics level of,
say, folding proteins? I know no single number completes the picture--
my point is, it's the *amount* of interaction at various scales that's
interesting, and I was under the impression that our computers have many
orders of magnitude to go before we even start to dent that. Plus, our
computers are a pretty small fraction of the mass of the solar system.
> We also know a lot about
>designing computers, carefully arranging the routes and speeds of bit
>exchanges.
Whoa. Computers I know. It is 1996 and almost all computers have single
processors with clocked logic. One thing we know is that neither of
those design choices are necessary. Sure, we know "a lot," but how much
do we know about what all computers *must* look like, under all design
constraints--especially the dirt constraint?
> And these two really look nothing alike! We can see
>nature's circuit diagram, even if we don't know what the bits
>"encode". In particular, we can see the "short" communications
>channels, where we have no strong reason to expect noise-like bits.
Unless (sorry to repeat) it's strategically necessary to design things
that way. It might be fun to try to come up with a method for designing
logic circuits where every signal looked as noisy as, say, a feedback
shift register. By the way, it's looking natural that's the constraint,
not looking like noise. Orderly or stereotyped patterns--say, the
recurrance of atoms in the periodic chart--are part of looking natural.
>To me, the idea that this circuit diagram is actually an
>optimum design for an advanced computer seems completely crazy.
So I hope I've made my point that you have to look at the driving game
theory, not just 20th century notions of computer design. If hiding is an
absolute constraint, then efficiency is secondary. If the most important
measure of "optimum" and "advanced" is whether something looks natural,
dirtlike, then whether it looks "crazy" is also secondary. Heck, almost
necessary. Just look at the way bits flow through an encryption algorithm.
[I will not say, "Yeah, crazy like a fox!" as this would spoil the
paragraph.]
I think the right point of view is to look at how much information
is sitting, colliding and moving free at each scale level, and then take
that as the given and design a computer and software (if the two must be
distinguished) to make use of it. When you're done you know how much
computing you can do--and you (the great intergalactic empire) live with
it. Them's the breaks, we can't compromise security. In computing there
are often ways of making use of speed *or* storage space, according to
what's available. Also ways of trading speed for thermodynamic efficiency.
I imagine aliens with billion-year patience would have extra slack here.
There are also strange tradeoffs in crypto. Like Chaum's cash protocol
where you can spend money once anonymously, but if you spend it twice you
reveal yourself. Imagining much more advanced crypto calls into question
what we know about the kinds of computing that need to be done, and also
simple game theory. For instance, imagine it was possible to cause
something to happen, but not know where it happened. What would that do
to notions of territory?
The thing I wanted to say about thermodynamics is that calculations of the
kind of computing and communicating you can do depend on the background
noise, or temperature. When trying to figure out how much computing dirt
or stellar plasma is doing, there is a problem: are those particle
motions really random? To what extent? If someone could throw off the
motions they didn't want and replace them with motions they intended, then
the effective temperature for the users and outside observers would be
different. That would change how much computing would seem possible.
(The other cryptocosmology quip: "God is in the least significant bits.")
I don't know whether physics would let this kind of temperature
difference exist for long, or at what cost. Imagine (metaphorically)
extremely cold parts moving extremely fast in an extremely well-designed
machine. Does physics let you keep that up without friction "melting the
parts" quickly, even with refrigerators? Even with, say, big systems in
known quantum states (setups too perfect for us to imagine making)?
The sun's light might plausibly be refrigerating the sun. Not so plausibly
the planets. Of course sunlight can power planetary refrigerators, but so
few of those BTUs are falling on planets. This really makes suns
look like the place to compute, except that plasma is even more unruly than
people (or is it?). Maybe planets are very very slow mass storage, doing
processing only in order to store and retrieve and refresh. But then there's
that nice single crystal of iron that forms the solid core of the earth...
Planets and suns both seem to have it over interstellar gas and dust because
they're compact, but considering what I said about speed/storage tradeoffs,
and also all that mysterious free-range sunlight, there could be something
going on there.
If we want to think about advanced civilizations, I think we need
to recognize what a wildcard that "advanced" is.
Cryptocosmology says They might need to evade probes like science.
At best we can ask whether computing is possible in "dirt" *as we know it*.
If not, the Great Filter has more plausibility.
If so, there's a giant "OR...NOT" thrown into the picture, with no
probability assigned at this point, but it would make some questions
in crypto and game theory more interesting.
--Steve
P.S., Do you think it would be clearer if I said "Stegocosmology?"
--
sw@tiac.net http://www.tiac.net/users/sw
"See, you think you're on a cruise ship but someone's moving the ocean."
--Patricia S. Sullivan