So You Want to Colonize The Universe

Epistemic Sta­tus: Mix of facts, far-fu­ture-spec­u­la­tion with the in­evitable bi­ases from only con­sid­er­ing tech­niques we know are phys­i­cally pos­si­ble, fermi calcu­la­tions, and an ac­tual space­craft de­sign made dur­ing a one-week re­search ben­der.

(this is a se­quence. 2, 3, 4, 5)

Part 1a: Gotta Go Fast, Astron­omy is a-Wasting


Once a civ­i­liza­tion or agent grows up enough to set it sights on coloniz­ing realms be­yond its host planet as its first pri­or­ity (in­stead of avert­ing ex­is­ten­tial risks), there is a very strong con­ver­gent in­stru­men­tal goal which kicks in. Namely, go­ing as close to light­speed (ab­bre­vi­ated as c) as pos­si­ble.

This is be­cause the uni­verse is ex­pand­ing, so there is a finite sphere of reach­able galax­ies, and more pass out­side the hori­zon ev­ery year. IIRC (60% prob­a­bil­ity), about half of the galax­ies in the Hub­ble Ul­tra-Deep Field are un­reach­able even if we trav­eled at light­speed.

Ar­riv­ing at a galaxy even one year faster nets you a marginal gain of (one galaxy of stars)*(av­er­age stel­lar lu­minos­ity)*(1 year) of en­ergy, which for our Milky way comes out to about Joules. As­sum­ing en­ergy pro­duc­tion on earth stays con­stant, that’s enough en­ergy for a billion years of earth civ­i­liza­tion, 130 trillion times over. And I’d ex­pect a tran­shu­man civ­i­liza­tion to be quite a few or­ders of mag­ni­tude bet­ter at get­ting value from a joule of en­ergy than our cur­rent civ­i­liza­tion. And that’s just for a sin­gle galaxy. There are a lot of galax­ies, a one-year speedup in reach­ing them has tremen­dous value.

This is ba­si­cally Bostrom’s as­tro­nom­i­cal waste ar­gu­ment, ex­cept Bostrom’s ver­sion then goes on to ap­ply this loss in to­tal form (which is far larger) in­stead of marginal form, to ar­gue for the value of re­duc­ing ex­is­ten­tial risk.

Now, there are a few cor­rec­tions to this to take into ac­count. The first and most im­por­tant is that, by the Lan­dauer limit, the amount of (ir­re­versible) com­pu­ta­tions that can be done is in­versely pro­por­tional to tem­per­a­ture, so wait­ing un­til the uni­verse cools down from its cur­rent 2.7 K tem­per­a­ture nets you sev­eral or­ders of mag­ni­tude more com­pu­ta­tional power from a given unit of en­ergy than spend­ing it now. Also, if you do re­versible com­pu­ta­tion, this limit doesn’t ap­ply ex­cept to bit era­sures, which nets you a whole lot more or­ders of mag­ni­tude of com­pu­ta­tion.

Another cor­rec­tion is that if there are aliens, they’ll be rush­ing to colonize the uni­verse, so the to­tal vol­ume a civ­i­liza­tion can grab is go­ing to be much smaller than the en­tire uni­verse. There’s still an in­cen­tive to go fast to cap­ture more stars be­fore they do, though.

There’s also a cor­rec­tion where the to­tal en­ergy available from coloniz­ing at all is more like the mass-en­ergy of a galaxy than the fu­sion power of a galaxy, for rea­sons I’ll get to in a bit. The marginal loss from tar­ry­ing for one year is about the same, though.

And fi­nally, if we con­sider the case where there aren’t aliens and we’re go­ing up to the cos­molog­i­cal hori­zon, the marginal loss is less than stated for very dis­tant galax­ies, be­cause by the time you get to a dis­tant galaxy, it will have burned down to red dwarfs only, which aren’t very lu­mi­nous.

Put­ting all this to­gether, we get the con­clu­sion that, for any agent whose util­ity func­tion scales with the amount of com­pu­ta­tions done, the con­ver­gent strat­egy is “go re­ally re­ally fast, cap­ture as many galax­ies as pos­si­ble, store as much mass-en­ergy as pos­si­ble in a sta­ble form and en­sure com­peti­tors don’t arise, then wait un­til the uni­verse is cold and dead to run ul­tra-low-tem­per­a­ture re­versible com­put­ing nodes.”

Part 1b: Stars For the Black Hole God! Utils For the Util Throne!

Now, bank­ing mass-en­ergy for sex­til­lions of years in a way that doesn’t de­cay is a key part of this, and for­tu­nately, there’s some­thing in na­ture that does it! Kerr black holes are spin­ning rapidly, and warp space around them in such a way that it’s pos­si­ble to re­cover some en­ergy from them, at the cost of spin­ning down the black hole slightly. For a max­i­mally spin­ning black hole, 29% of the mass-en­ergy can be re­cov­ered as en­ergy via ei­ther the Pen­rose Pro­cess (throw­ing some­thing near the hole in a way that in­volves it com­ing back with more en­ergy than it went in), or the Bland­ford-Zna­jek Pro­cess (which in­volves set­ting up a mag­netic field around the hole and this in­duc­ing a cur­rent, and this is a ma­jor pro­cess pow­er­ing quasars). I’m more par­tial to the sec­ond be­cause it pro­duces a cur­rent. Most black holes are Kerr black holes, and we’ve found quite a few black holes (in­clud­ing su­per­mas­sive ones) that are spin­ning at around 0.9x the max­i­mum spin, so an awful lot of en­ergy can be ex­tracted from them. So, if we sac­ri­ficed the en­tire Milky Way galaxy to the black hole at the cen­ter by nudg­ing stel­lar or­bits un­til they all went in, we’d have 5x10^57 joules of ex­tractible en­ergy to play around with. Take a minute to ap­pre­ci­ate how big this is. And re­mem­ber, this is per-galaxy. Another or­der of mag­ni­tude could be got­ten if there’s some way for a far-fu­ture civ­i­liza­tion to in­ter­face with dark mat­ter.

So, the dom­i­nant strat­egy is some­thing like “get to as much of the uni­verse as fast as pos­si­ble, and sac­ri­fice all the stars you en­counter to the black hole gods, and then in the far far fu­ture you can get the party started, with an ab­solutely ridicu­lous amount of en­ergy at your dis­posal, and also the abil­ity to use a given unit of en­ergy far far more effi­ciently than we can to­day, by re­versible com­pu­ta­tion and the uni­verse be­ing re­ally cold”

(It’s a fuzzy mem­ory, and Eliezer is wel­come to cor­rect me on this if I’ve mis­rep­re­sented his views, and I’ll edit this sec­tion)

Due to the ex­pan­sion of the uni­verse, these mega-black-holes will be per­ma­nently iso­lated from each other. I think Eliezer’s pro­posal was to throw as much mass back to the Milky Way as pos­si­ble, and set up shop there, in­stead of cut­ting far-fu­ture civ­i­liza­tion into a bunch of ab­solutely dis­con­nected is­lands. I don’t think this is as good, be­cause I’d pre­fer a larger civ­i­liza­tion over the whole uni­verse (from not hav­ing to throw mass back to the milky way, just throw it to the near­est hole), cut into more dis­con­nected is­lands, than a much smaller civ­i­liza­tion that’s all in one is­land.

Part 1c: The Triple Tradeoff

In un­re­lated news, there’s also a un­set­tling ar­gu­ment that I came up with that there’s a con­ver­gent in­cen­tive to re­duce the com­pu­ta­tional re­sources the com­put­ing node con­sumes. If you switch a simu­lated world to be lower fidelity, and 80% as fun, but now it only takes a fifth of the com­pu­ta­tional re­sources so 5x as much lifes­pan is available, I think I’d take that bar­gain. Tak­ing this to the end­point, I made the joke on a Dank EA Memes poll that the trillion trillion heav­ens of the far-fu­ture all have shitty Minecraft graph­ics, but I’m ac­tu­ally quite un­cer­tain how that ends up, and there’s also the ar­gu­ment that most of the com­pu­ta­tional power goes to run­ning the minds them­selves and not the en­vi­ron­ment, in which case there’s an in­cen­tive for sim­plify­ing one’s own thought pro­cesses so they take less re­sources.

Gen­er­al­iz­ing this, there seems to be a three-way trade­off be­tween pop­u­la­tion, lifes­pan, and com­pu­ta­tional re­sources con­sumed per mem­ber. Pick­ing the pop­u­la­tion ex­treme, you’d get a gi­gan­tic pop­u­la­tion of short-lived sim­ple agents. Pick­ing the lifes­pan ex­treme, you get a small pop­u­la­tion of sim­ple agents liv­ing for a re­ally re­ally long time. Pick­ing the com­pu­ta­tional re­sources ex­treme, you get a small pop­u­la­tion of short-lived re­ally re­ally posthu­man agents. (note that short-lived can still be quite long rel­a­tive to hu­man lifes­pans) I’m not sure what the best trade­off point is here, and it may vary by per­son, so some­thing like “you get a finite but ridicu­lously large amount of com­pu­ta­tional re­sources, and if you want to be sim­pler and live longer, or go to­wards ever-greater heights of posthu­man­ity with a shorter life, or have a bunch of ba­bies and split your re­sources with them, you can do that”. How­ever, that ap­proach plus would lead to most of the pop­u­la­tion be­ing de­scended from peo­ple who val­ued re­pro­duc­tion over long life or be­ing re­ally tran­shu­man, and they’d get less re­sources for them­selves, and that seems in­tu­itively bad. Also maybe there could be merg­ing of peo­ple, with as­so­ci­ated pool­ing of re­sources? I’m not quite sure how to nav­i­gate this trade­off, ex­cept to say that the pop­u­la­tion ex­treme of it seems bad, and that it’s a re­ally im­por­tant far-fu­ture is­sue. I should prob­a­bly also point out that if this is the fa­vored ap­proach, in the long-time limit, most of those that are left will be those that have fa­vored lifes­pan over be­ing re­ally tran­shu­man or re­pro­duc­tion, so I guess the last thing left liv­ing be­fore heat death might ac­tu­ally be a min­i­mally-re­source-in­ten­sive con­scious agent in a world with low-qual­ity graph­ics.

Part 1d: An Ex­ploitable Fic­tion Opportunity

Also, in un­re­lated news, I think I see an ex­ploitable gap in fic­tion-writ­ing. The elephant in the room for all space-travel sto­ries is that space is in­com­pat­i­ble with mam­mals, and due to ad­vances in elec­tron­ics, it just makes more sense to send up robotic probes.

How­ever, Burn­side’s Zeroth Law of Space Com­bat is:

Science fic­tion fans re­late more to hu­man be­ings than to sili­con chips.

I’m not a writer, but this doesn’t strike me as en­tirely true, due to the ten­dency of hu­mans to an­thro­po­mor­phize. When talk­ing about the goal of space travel be­ing to hit up as many stars and galax­ies as pos­si­ble as fast as pos­si­ble, and throw them into black holes, the very first thing that came to mind was “aww, the civ­i­liza­tion is act­ing just like an ob­ses­sive speedrun­ner!”

I like watch­ing speedruns, it’s ab­solutely fas­ci­nat­ing watch­ing that much op­ti­miza­tion power be­ing di­rected at the task of go­ing as fast as pos­si­ble in defi­ance of the lo­cal rules and find­ing cool ex­ploits. I’d to­tally read about the ex­ploits of a civ­i­liza­tion that’s over­joyed to find a way to make their dust shields 5% more effi­cient be­cause that means they can reach a few thou­sand more galax­ies, and Vinny the Von Neu­mann probe strug­gling to be as use­ful as it can given that it was sent to a low-qual­ity as­ter­oid, and stuff like that. The stakes are mas­sive, you just need to put in some work to make the marginal gain of ac­cel­er­ated coloniza­tion more vivid for the reader. It’s the ul­ti­mate real-life tale of munchk­inry for mas­sive stakes and there’s also am­ple “I know you know I know...” rea­son­ing in­tro­duced by virtue of light-speed com­mu­ni­ca­tion de­lays, and ev­ery­one’s on the same side vs. na­ture.

I think Burn­side might have been refer­ring to sci­ence fic­tion for a more con­ven­tional au­di­ence, given the gap be­tween his ad­vice and my own re­ac­tion. But hard-sci-fi fans are already a pretty self-se­lected group, and Less Wrong read­ers are even moreso, and be­sides, with the ad­vent of the in­ter­net, re­ally niche fic­tion is a lot eas­ier to pull off, it feels like there’s a Hard-SciFi x Speedrun­ning niche out there available to be filled. A dash of an­thro­po­mor­phiza­tion along with Suffi­ciently In­tel­li­gent Probes feels like it could go a long way to­wards mak­ing peo­ple re­late to the sili­con chips.

So I think there’s an ex­ploitable niche here.

Put­ting all this to the side, though, I’m in­ter­ested in the “go fast” part. Really, how close to light speed is at­tain­able?