Sufficiently Advanced Technology

[Yakk]

So I was reading some stuff on Charlie Stross's blog, and it got me to thinking about sufficiently advanced technologies.

Our civilization is young, small and hot. I suspect that old, large and cold civilizations might be nearly undetectable by us.

We are young, hence we think on scales of seconds, years and centuries. We have no sense of time scale as a species of an individual.

We are small, so we are used to the idea that there is plenty of room to grow into, and getting there sooner is a significant advantage.

We are hot, in that we are amazingly inefficient. We waste lots of energy. We also expect other civilizations to do the same. We have cooled off rapidly over even the last few years of this technological revolution, with the amount of radiation we leak to space per member of the highest technology civilization on the planet falling rapidly.

An old civilization wouldn't have to think on the scale of seconds, years or centuries. It could easily think on the scale of the universe. Ie, how many cycles of computation do you get before the universe ends?

An old civilization could be a large one. Expanding, and especially expanding rapidly, doesn't matter all that much -- you are limited to c anyhow, and the amount of volume of the universe you can expand into is small relative to the size of your existing civilization, possibly even on end-of-universe time scales.

An old civilization could be a cool one. One idea is that high end civilizations would turn matter into computronium, and we'd be able to detect it by looking for hot spots of high entropy in the sky. But hot spots are wasteful. If you make your computronium sphere larger, you get more computation out of the energy, and the temperature of the waste black body radiation (high entropy radiation) drops. At maximal efficiency, the computronium sphere waste products approaches the background temperature of the universe -- the cosmic microwave background, if you hang around in intergalactic space at least. A civilization that shows up as a thermal, radar, radio or other "hot spot" is a hot civilization who is failing to use its energy at optimal thermodynamic efficiency.

We look around, and we don't see such hot spots. We do the math, and it isn't hard for a slightly more advanced civilization to colonize an entire Galaxy over a period of a gigayear or so (exploring via self replicating probes is "easy" for a slightly more advanced civilization -- and if a civilization becomes a machine one, exploring with self replicating probes is colonization). However, if such an event happened in concert with the civilization getting old, cold and large, it might fade out from view.

We could have computronium solar systems that are light years in diameter that look like background radiation. They might even "cool" the star in order to make it last longer. They might pilot the star out of the galactic disk entirely in order to reduce the background heat level, and allow them to swell the size of their sphere and cool off the surface even further.

The only thing that would leak out, and be detectable to us, would be gravity.

The expected symptom we'd see is a large amount of extra mass that we couldn't account for.

(And that is my punchline).

[WarDaft]

I like how you propose moving a star as a means to make a more efficient computer.

Spoilered for much too seriousness..

Spoiler:

Moving a star 10,000 LY out of the galactic plane over 1 billion years is wasting about 10³⁷ Joules in kinetic energy. You could take even longer to do it, but much more and your star will burn out, and you'd probably want it further away for a colder background anyway. There's no way to recover that energy either.

And, if there are even vaguely efficient ways to convert mass into energy, then the most effective computer possible won't be near CMB temperatures. They will be will be "much" hotter, dozens, maybe even over a hundred Kelvin depending on ratios, because the small amount of computation the outer shells do (having very little energy difference to work with) will not even vaguely compare to fueling the entire structure with some of their mass - remembering of course that it will take energy to make them into computronium in the first place. Finally, there's no way you can consider the outermost shells as part of the same effective computation as the innermost shells, the relative time lag is too high... if this matryoshka-brain is running an intelligence, the latency and effort of communicating with the outer shells must be considered. Think networking between two computers on opposite ends of the earth, compared to adjacent gates in a CPU, multiplied by a billion or so.

The most likely and efficient way I can see of practically making such a structure is to have a massive fusion reactor and assembly plant acting as the 'sun', located somewhere in a stellar nursary at just the right density (probably somewhere just before a star is forming, 'just' in astronomical timescales of course.) The fusion reactor builds compution cells and copies of itself out of the surrounding pre-star cloud over time, with the computer cells starting close in and gradually migrating further out. This also allows the hottest cells to be the newest, taking maximum advantage of any improvement in hardware design. Furthermore, the whole thing is virtually gravitationally neutral for a very long time, so little energy is wasted in orbits. However, there is also a matter of diminishing returns on the cells. As they drift out, they will receive less and less energy, and over a sufficiently long time scale will follow an inverse square drop-off in usable energy passing though. This means we can calculate a precise bound on the amount of useful computation each cell will perform, and thus pick a date to recycle their mass into some more energy - should such things be possible. Furthermore, there is an overall time limit, as the pre-star cloud will not wait around forever to collapse into a star, particularly with our density increasing meddling. Such a time limit would seem bad yes, it is easy to see how the ease of construction and availability of materials (it will just drift in naturally!) could make this one of the most powerful (as in, does the most calculations over its life span) computers physically constructable.

That said, none of these considerations would make the structure visible to our level of detection technologies.

I also recall hearing an amusing fact that said even with all our detection efforts, we couldn't even detect our own civilization from outside our own solar system at the peak of it's signal wasting.

[Yakk]

I though that was a rather fun idea.

While the mass of the computronium could indeed be turned into energy, using it as computronium doesn't prevent you from doing it later.

And similarly, your concerns about latency are also a speed issue. Remember, I'm talking about an old and large civilization who cares more about how much computation can be done before the end of the universe, rather than how much can be done in a billion years. Latency of computation of a few centuries isn't significant when your game is this long.

The energy budget to slowly nudge, or sling-shot, a star out of the galactic plane (or the galaxy itself) might not be ideal. But I should double check:

Cost to do a state change is around 4kT. Intergalactic space is ~3 k. Assuming your E37 J to kick a 1 stellar mass star out of the galaxy in a billion years or so is accurate... The mass energy of the sun is about E47 J (mc^2). So it takes 1/10000000000 of the energy of the sun to get it out of the galactic disk, assuming perfect mass-energy conversion. As the theoretical limit on computation is the background temperature, this cost could be justified if the background temperature outside of the galaxy is 0.000000001% colder than inside the galaxy.

So yes, it can be efficient to move a star out of the galaxy to provide more efficient cooling on your computer, over the long term.

[Soralin]

There is a time limit though on this. I mean, right now, stars are using up huge amount of resources. The Sun is using up 620 million tons of hydrogen every single second, converting it into helium. And even if you have 100% efficient matter->energy conversion, that's still 4.26 million tons of mass gone, every second, converted into energy. For the energy output of the milky way galaxy, it's somewhere around half a quintillion(10¹⁸) tons of matter lost, every single second that you're not making use of it.

So, if you really wanted to get the most cycles of computation done before the end of the universe, you'd want to encase or dismantle every single star as soon as you can, to prevent all that energy from going to waste.

[mfb]

Rotating black holes can convert ~30% of their incoming mass into hot (~keV), intense radiation. If you have access to multiple stars, you can use existing black holes without problems. Fuse and fission everything to something near iron ("small" energy output) and throw it into the accreation disk around the black hole afterwards (large energy output).
If you have a method to destroy a star and use its mass: Fine. Otherwise, do some computation around it until it is cool enough and use the remnant as fuel afterwards.

You can collect the energy and use it far away from the stars in a 3K environment. It does not really matter whether you are in the galaxy or outside, the temperature differences are too small. And as you use (and block) the stellar radiation of nearby stars anyway, it is as cool as it would be far away from the stars.


>> The expected symptom we'd see is a large amount of extra mass that we couldn't account for.
The expected symptom would be an unequal distribution of stars (like areas without them or with different categories of stars) or no stars at all.
Oh, and a lot of microlensing events, or events where a star vanishes due to the large object in front of it.

[Giallo]

So, if you really wanted to get the most cycles of computation done before the end of the universe, you'd want to encase or dismantle every single star as soon as you can, to prevent all that energy from going to waste.

This seems to me an exceptionally good idea for an enemy for sci-fi literature

[AvatarIII]

So, if you really wanted to get the most cycles of computation done before the end of the universe, you'd want to encase or dismantle every single star as soon as you can, to prevent all that energy from going to waste.

This seems to me an exceptionally good idea for an enemy for sci-fi literature

Alastair Reynolds already did it. Sort of, out of control dyson cloud self replicators.

[Sean Quixote]

So this thread is my first encounter with the concept of 'computronium', so I apologize if this is an uninformed/unoriginal/or perhaps even obvious thought. But, given the purely theoretical and thusfar fictional nature of such a substance.... how do we know that WE - sentient, semi-intelligent, biological lifeforms - are not indeed the closest approximation the Universe has yet conceived of, to what we are currently referring to as computronium?

[Copper Bezel]

Computronium would refer to the theoretical limit of computing power you could get out of a substance. Our nerve cells would be the closest we have right now, and any time the term is used in this thread, it really means the closest thing the hypothetical civilization would have to computronium.

We do know that human nerve cells don't come close to the theoretical limits, of course, and we don't know whether or not the highly advanced civilizations discussed here might approach them. That is, we know at least that their nerve cells will be better than ours. = ) (And thus closer to computronium than ours are.)

[Quizatzhaderac]

I think moving the star out of the galactic plane might have a significant energy cost just in increasing the star's gravitational potential. I don't actually have any of the numbers, but assuming a strong correlation between "stuff producing radiation" and "Stuff with mass", I'm not sure moving stuff around would get you more computation, even assuming you can stasis a star, and spend a trillion years moving it.

[niky]

So, if you really wanted to get the most cycles of computation done before the end of the universe, you'd want to encase or dismantle every single star as soon as you can, to prevent all that energy from going to waste.

This seems to me an exceptionally good idea for an enemy for sci-fi literature

Alastair Reynolds already did it. Sort of, out of control dyson cloud self replicators.

Minds me of Greg Bear's Jarts from Eternity. Who dismantle civilizations and digitize them for more efficient storage downstream (or downtime, or where/whenever).

-

And of Baxter's Manifold Time, where the end result of the Universe is a Universal Mind, that, unfortunately, can only have a limited number of thoughts due to energy constraints... which then sends messages back into the past to create an alternate future that isn't quite as dead-end as that...

-

Eventually, if it is possible, we could all become digitized life-forms living on the radiation from evaporating black holes.

Not very sexy or fun... as it would naturally be a very cold and slow existence.

[mathmannix]

So, if you really wanted to get the most cycles of computation done before the end of the universe, you'd want to encase or dismantle every single star as soon as you can, to prevent all that energy from going to waste.

This seems to me an exceptionally good idea for an enemy for sci-fi literature

Alastair Reynolds already did it. Sort of, out of control dyson cloud self replicators.

Reminds me of the Masters of Krikkit...

It'll have to go.

[Xanthir]

Eventually, if it is possible, we could all become digitized life-forms living on the radiation from evaporating black holes.

Not very sexy or fun... as it would naturally be a very cold and slow existence.

Given that you'd be living in simulation, it would be as sexy and fun as you wish. The cold doesn't matter, and the slow just means you run the simulation at a lower clockrate and don't notice the difference on the inside.

[niky]

Slower clockrate means far fewer eons of virtual life while waiting for entropy to take you all. Then again, you could go into an energy saving loop for a few billion years and never know the difference.

[WarDaft]

Hmm, I think I read somewhere that there was no reachable limit on how entropy efficient certain reversible processes could be made (that is, it is perfect in the limit, and anything less than perfect is theoretically achievable in a finite amount of time)... the question then is, can we make self-improving processes which gradually produce less and less entropy while doing calculations, such that in the infinite limit it sums up to a finite amount of total entropy produced and thus allows us to have effectively infinitely many cycles over the lifetime of a flat universe? If such is the case, we don't have to go around gobbling up all the stars we can, we only need a fixed amount! Beyond that, it won't actually increase the amount of things we can calculate, and we can all live forever using a finite amount of energy.

[idobox]

isn't there a theorical limit on computing efficiency? something like minimal entropy increase per bit switch?

[eSOANEM]

isn't there a theorical limit on computing efficiency? something like minimal entropy increase per bit switch?

IIRC this (the Landauer limit) only applies to irreversible computations (such as bit switches) and so, if you're cunning with your programs, it should be possible to compute in such a way that you avoid it.

[Yakk]

isn't there a theorical limit on computing efficiency? something like minimal entropy increase per bit switch?

For irreversible computing, yes there is. But it is a function of the temperature at which the computation occurs at, and (IIRC) the speed at which it happens.

[eSOANEM]

isn't there a theorical limit on computing efficiency? something like minimal entropy increase per bit switch?

For irreversible computing, yes there is. But it is a function of the temperature at which the computation occurs at, and (IIRC) the speed at which it happens.

I don't think it is a function of the speed of computation (at least, the lower limit on the entropy isn't affected by this). The formula given on wikipedia (kTln(2)) is dependent solely on temperature.