The First World Takeover

Be­fore Robin and I move on to talk­ing about the Fu­ture, it seems to me wise to check if we have dis­agree­ments in our view of the Past. Which might be much eas­ier to dis­cuss—and maybe even re­solve… So...

In the be­gin­ning was the Bang. For nine billion years af­ter­ward, noth­ing much hap­pened.

Stars formed, and burned for long pe­ri­ods or short pe­ri­ods de­pend­ing on their struc­ture; but “suc­cess­ful” stars that burned longer or brighter did not pass on their char­ac­ter­is­tics to other stars. The first repli­ca­tors were yet to come.

It was the Day of the Stable Things, when your prob­a­bil­ity of see­ing some­thing was given by its prob­a­bil­ity of ac­ci­den­tal for­ma­tion times its du­ra­tion. Stars last a long time; there are many he­lium atoms.

It was the Era of Ac­ci­dents, be­fore the dawn of op­ti­miza­tion. You’d only ex­pect to see some­thing with 40 bits of op­ti­miza­tion if you looked through a trillion sam­ples. Some­thing with 1000 bits’ worth of func­tional com­plex­ity? You wouldn’t ex­pect to find that in the whole uni­verse.

I would guess that, if you were go­ing to be stuck on a desert is­land and you wanted to stay en­ter­tained as long as pos­si­ble, then you should sooner choose to ex­am­ine the com­plex­ity of the cells and bio­chem­istry of a sin­gle Earthly but­terfly, over all the stars and as­tro­physics in the visi­ble uni­verse be­yond Earth.

It was the Age of Bore­dom.

The hal­l­mark of the Age of Bore­dom was not lack of nat­u­ral re­sources—it wasn’t that the uni­verse was low on hy­dro­gen—but, rather, the lack of any cu­mu­la­tive search. If one star burned longer or brighter, that didn’t af­fect the prob­a­bil­ity dis­tri­bu­tion of the next star to form. There was no search but blind search. Every­thing from scratch, not even look­ing at the neigh­bors of pre­vi­ously suc­cess­ful points. Not hill-climb­ing, not mu­ta­tion and se­lec­tion, not even dis­card­ing pat­terns already failed. Just a ran­dom sam­ple from the same dis­tri­bu­tion, over and over again.

The Age of Bore­dom ended with the first repli­ca­tor.

(Or the first repli­ca­tor to catch on, if there were failed al­ter­na­tives lost to his­tory—but this seems un­likely, given the Fermi Para­dox; a repli­ca­tor should be more im­prob­a­ble than that, or the stars would teem with life already.)

Though it might be most dra­matic to think of a sin­gle RNA strand a few dozen bases long, form­ing by pure ac­ci­dent af­ter who-knows-how-many chances on who-knows-how-many planets, an­other class of hy­pothe­ses deals with cat­alytic hy­per­cy­cles—chem­i­cals whose pres­ence makes it more likely for other chem­i­cals to form, with the ar­rows hap­pen­ing to fi­nally go around in a cir­cle. If so, RNA would just be a crys­tal­liza­tion of that hy­per­cy­cle into a sin­gle chem­i­cal that can both take on en­zy­matic shapes, and store in­for­ma­tion in its se­quence for easy repli­ca­tion.

The cat­alytic hy­per­cy­cle is worth pon­der­ing, since it re­minds us that the uni­verse wasn’t quite draw­ing its ran­dom pat­terns from the same dis­tri­bu­tion ev­ery time—the for­ma­tion of a longlived star made it more likely for a planet to form (if not an­other star to form), the for­ma­tion of a planet made it more likely for amino acids and RNA bases to form in a pool of muck some­where (if not more likely for planets to form).

In this flow of prob­a­bil­ity, pat­terns in one at­trac­tor lead­ing to other at­trac­tors be­com­ing stronger, there was fi­nally born a cy­cle—per­haps a sin­gle strand of RNA, per­haps a crys­tal in clay, per­haps a cat­alytic hy­per­cy­cle—and that was the dawn.

What makes this cy­cle sig­nifi­cant? Is it the large amount of ma­te­rial that the cat­alytic hy­per­cy­cle or repli­cat­ing RNA strand could ab­sorb into its pat­tern?

Well, but any given moun­tain on Pri­mor­dial Earth would prob­a­bly weigh vastly more than the to­tal mass de­voted to copies of the first repli­ca­tor. What effect does mere mass have on op­ti­miza­tion?

Sup­pose the first repli­ca­tor had a prob­a­bil­ity of for­ma­tion of 10-30. If that first repli­ca­tor man­aged to make 10,000,000,000 copies of it­self (I don’t know if this would be an over­es­ti­mate or un­der­es­ti­mate for a tidal pool) then this would in­crease your prob­a­bil­ity of en­coun­ter­ing the repli­ca­tor-pat­tern by a fac­tor of 1010, the to­tal prob­a­bil­ity go­ing up to 10-20. (If you were ob­serv­ing “things” at ran­dom, that is, and not just on Earth but on all the planets with tidal pools.) So that was a kind of op­ti­miza­tion-di­rected prob­a­bil­ity flow.

But vastly more im­por­tant, in the scheme of things, was this—that the first repli­ca­tor made copies of it­self, and some of those copies were er­rors.

That is, it ex­plored the neigh­bor­ing re­gions of the search space—some of which con­tained bet­ter repli­ca­tors—and then those repli­ca­tors ended up with more prob­a­bil­ity flow­ing into them, which ex­plored their neigh­bor­hoods.

Even in the Age of Bore­dom there were always re­gions of at­trac­tor-space that were the gate­ways to other re­gions of at­trac­tor-space. Stars be­got planets, planets be­got tidal pools. But that’s not the same as a repli­ca­tor beget­ting a repli­ca­tor—it doesn’t search a neigh­bor­hood, find some­thing that bet­ter matches a crite­rion (in this case, the crite­rion of effec­tive repli­ca­tion) and then search that neigh­bor­hood, over and over.

This did re­quire a cer­tain amount of raw ma­te­rial to act as repli­ca­tor feed­stock. But the sig­nifi­cant thing was not how much ma­te­rial was re­cruited into the world of repli­ca­tion; the sig­nifi­cant thing was the search, and the ma­te­rial just car­ried out that search. If, some­how, there’d been some way of do­ing the same search with­out all that raw ma­te­rial—if there’d just been a lit­tle beep­ing de­vice that de­ter­mined how well a pat­tern would repli­cate, and in­cre­mented a bi­nary num­ber rep­re­sent­ing “how much at­ten­tion” to pay to that pat­tern, and then searched neigh­bor­ing points in pro­por­tion to that num­ber—well, that would have searched just the same. It’s not some­thing that evolu­tion can do, but if it hap­pened, it would gen­er­ate the same in­for­ma­tion.

Hu­man brains rou­tinely out­think the evolu­tion of whole species, species whose net weights of biolog­i­cal ma­te­rial out­weigh a hu­man brain a mil­lion times over—the gun against a lion’s paws. It’s not the amount of raw ma­te­rial, it’s the search.

In the evolu­tion of repli­ca­tors, the raw ma­te­rial hap­pens to carry out the search—but don’t think that the key thing is how much gets pro­duced, how much gets con­sumed. The raw ma­te­rial is just a way of keep­ing score. True, even in prin­ci­ple, you do need some ne­gen­tropy and some mat­ter to perform the com­pu­ta­tion. But the same search could the­o­ret­i­cally be performed with much less ma­te­rial—ex­am­in­ing fewer copies of a pat­tern, to draw the same con­clu­sions, us­ing more effi­cient up­dat­ing on the ev­i­dence. Repli­ca­tors hap­pen to use the num­ber of copies pro­duced of them­selves, as a way of keep­ing score.

But what re­ally mat­ters isn’t the pro­duc­tion, it’s the search.

If, af­ter the first prim­i­tive repli­ca­tors had man­aged to pro­duce a few tons of them­selves, you deleted all those tons of biolog­i­cal ma­te­rial, and sub­sti­tuted a few dozen cells here and there from the fu­ture—a sin­gle al­gae, a sin­gle bac­terium—to say noth­ing of a whole mul­ti­cel­lu­lar C. el­e­gans earth­worm with a 302-neu­ron brain—then Time would leap for­ward by billions of years, even if the to­tal mass of Life had just ap­par­ently shrunk. The search would have leapt ahead, and pro­duc­tion would re­cover from the ap­par­ent “set­back” in a hand­ful of easy dou­blings.

The first repli­ca­tor was the first great break in His­tory—the first Black Swan that would have been uni­mag­in­able by any sur­face anal­ogy. No ex­trap­o­la­tion of pre­vi­ous trends could have spot­ted it—you’d have had to dive down into causal mod­el­ing, in enough de­tail to vi­su­al­ize the un­prece­dented search.

Not that I’m say­ing I would have guessed, with­out benefit of hind­sight—if some­how I’d been there as a dis­em­bod­ied and un­re­flec­tive spirit, know­ing only the pre­vi­ous uni­verse as my guide—hav­ing no high­falutin’ con­cepts of “in­tel­li­gence” or “nat­u­ral se­lec­tion” be­cause those things didn’t ex­ist in my en­vi­ron­ment, and I had no men­tal mir­ror in which to see my­self—and in­deed, who should have guessed it with short of godlike in­tel­li­gence? When all the pre­vi­ous his­tory of the uni­verse con­tained no break in His­tory that sharp? The repli­ca­tor was the first Black Swan.

Maybe I, see­ing the first repli­ca­tor as a dis­em­bod­ied un­re­flec­tive spirit, would have said, “Wow, what an amaz­ing no­tion—some of the things I see won’t form with high prob­a­bil­ity, or last for long times—they’ll be things that are good at copy­ing them­selves, in­stead. It’s the new, third rea­son for see­ing a lot of some­thing!” But would I have been imag­i­na­tive enough to see the way to amoe­bas, to birds, to hu­mans? Or would I have just ex­pected it to hit the walls of the tidal pool and stop?

Try tel­ling a dis­em­bod­ied spirit who had watched the whole his­tory of the uni­verse up to that point about the birds and the bees, and they would think you were ab­solutely and en­tirely out to lunch. For noth­ing re­motely like that would have been found any­where else in the uni­verse—and it would ob­vi­ously take an ex­po­nen­tial and ridicu­lous amount of time to ac­ci­den­tally form a pat­tern like that, no mat­ter how good it was at repli­cat­ing it­self once formed—and as for it hap­pen­ing many times over in a con­nected ecol­ogy, when the first repli­ca­tor in the tidal pool took such a long time to hap­pen—why, that would just be mad­ness. The Ab­sur­dity Heuris­tic would come into play. Okay, it’s neat that a lit­tle molecule can repli­cate it­self—but this no­tion of a “squir­rel” is in­san­ity. So far be­yond a Black Swan that you can’t even call it a swan any­more.

That first repli­ca­tor took over the world—in what sense? Earth’s crust, Earth’s magma, far out­weighs its mass of Life. But Robin and I both sus­pect, I think, that the fate of the uni­verse, and all those dis­tant stars that out­weigh us, will end up shaped by Life. So that the uni­verse ends up hang­ing quite heav­ily on the ex­is­tence of that first repli­ca­tor, and not on the coun­ter­fac­tual states of any par­tic­u­lar other molecules nearby… In that sense, a small hand­ful of atoms once seized the reins of Destiny.

How? How did the first repli­cat­ing pat­tern take over the world? Why didn’t all those other molecules get an equal vote in the pro­cess?

Well, that ini­tial repli­cat­ing pat­tern was do­ing some kind of search—some kind of op­ti­miza­tion—and noth­ing else in the Uni­verse was even try­ing. Really it was evolu­tion that took over the world, not the first repli­cat­ing pat­tern per se—you don’t see many copies of it around any more. But still, once upon a time the thread of Destiny was seized and con­cen­trated and spun out from a small hand­ful of atoms.

The first repli­ca­tor did not set in mo­tion a clever op­ti­miza­tion pro­cess. Life didn’t even have sex yet, or DNA to store in­for­ma­tion at very high fidelity. But the rest of the Uni­verse had zip. In the king­dom of blind chance, the my­opic op­ti­miza­tion pro­cess is king.

Is­sues of “shar­ing im­prove­ments” or “trad­ing im­prove­ments” wouldn’t even arise—there were no part­ners from out­side. All the agents, all the ac­tors of our mod­ern world, are de­scended from that first repli­ca­tor, and none from the moun­tains and hills.

And that was the story of the First World Takeover, when a shift in the struc­ture of op­ti­miza­tion—namely, mov­ing from no op­ti­miza­tion what­so­ever, to nat­u­ral se­lec­tion—pro­duced a stark dis­con­ti­nu­ity with pre­vi­ous trends; and squeezed the flow of the whole uni­verse’s des­tiny through the nee­dle’s eye of a sin­gle place and time and pat­tern.

That’s Life.