johnswentworth

• johnswentworth 18 Jun 2019 19:21 UTC

  2 points

  on: Par­tial prefer­ences and models

  How does this no­tion of par­tial prefer­ences differ from say­ing “prefer­ences are de­ter­mined by a causal net”? I.e., the y’s would be the di­rect causal par­ents of a de­ci­sion, and the z’s ev­ery­thing else.

• johnswentworth 18 Jun 2019 17:02 UTC

  34 points

  on: Is the “busi­ness cy­cle” an ac­tual eco­nomic prin­ci­ple?

  Economists tend to view the busi­ness cy­cle as one of the main stylized facts which any the­ory must ac­count for, rather than a prin­ci­ple in it­self.

  In terms of statis­ti­cal ev­i­dence, mar­ket price move­ments are definitely not a vanilla ran­dom walk—ran­dom walks don’t have long tails un­less you add some­thing else into the the­ory. The usual baseline model is a ran­dom walk in which the var­i­ance is also gov­erned by a ran­dom walk; a quant trad­ing shop will aug­ment this with oc­ca­sional jumps and a few small terms to ac­count for higher mo­ments of the dis­tri­bu­tion. And even that is just the baseline to rea­son­ably model a sin­gle as­set—mod­el­ling mul­ti­ple as­sets is far more com­pli­cated, since they usu­ally be­come more cor­re­lated dur­ing a crash (an­other phe­nomenon which wouldn’t hap­pen in a vanilla ran­dom walk). If you want to learn more about this stuff, look for a math­e­mat­i­cal fi­nance class.

  Get­ting back to the eco­nomics side, ev­ery ma­jor class of macroe­co­nomic the­o­ries has an an­swer to the ques­tion “What the heck is up with these busi­ness cy­cles? Why do mar­kets some­times crash way harder than they should with ran­dom walks?” Real busi­ness cy­cle (RBC) the­ory posits that they’re the re­sult of shocks to the econ­omy, e.g. hur­ri­canes or wars. Mar­ket mon­e­tarism at­tributes crashes to tight money, as mea­sured by NGDP. Key­ne­si­anism fo­cuses on vari­a­tion in ag­gre­gate de­mand. Mone­tarism fo­cuses on vari­a­tion in the money sup­ply and/​or money ve­loc­ity. Etc, etc. If you want to see the fancy stuff, pick up a book on “re­cur­sive macroe­co­nomic the­ory”.

  But to the de­gree that the busi­ness cy­cle is a sep­a­rate un­der­stand­able phe­nomenon, can’t in­vestors use that un­der­stand­ing to place bets which make them money while damp­en­ing the effect?

  This is part of what the var­i­ous the­o­ries try to ex­plain. Some re­quire ir­ra­tional­ity in or­der for busi­ness cy­cles to ex­ist—usu­ally in the form of “sticky wages”, mean­ing that peo­ple don’t like wage cuts even if the al­ter­na­tive is get­ting fired and ac­cept­ing lower pay el­se­where. In this case, traders can’t nec­es­sar­ily coun­ter­bal­ance the whole effect, no mat­ter how clever they are—so mar­kets will crash as soon as the traders know that a sticky wage prob­lem is on the hori­zon. On the other hand, RBC ex­plains (some) busi­ness cy­cles with­out any ir­ra­tional­ity at all: if a big hur­ri­cane sig­nifi­cantly low­ers GDP this year, then so­ciety as a whole will eat into their sav­ings to cover the costs—which means eat­ing into the cap­i­tal stock, and low­er­ing GDP over the next few years. Traders can’t price that in un­til they know about the dam­age, at which point there’s a mar­ket crash.

ISO: Au­to­mated P-Hack­ing Detection

• johnswentworth 14 Jun 2019 0:17 UTC

  4 points

  in reply to: Dagon's comment on: Real-World Co­or­di­na­tion Prob­lems are Usu­ally In­for­ma­tion Problems

  “From Per­sonal to Pri­son Gangs” is my main foun­da­tion here. I say real-world co­or­di­na­tion prob­lems “usu­ally” look like this, be­cause these are the kinds of prob­lems we’d ex­pect to in­crease over time, based on the ideas in that pre­vi­ous post.

  That said, Per­sonal to Pri­son Gangs at­tributes both the in­for­ma­tion prob­lems and the trust prob­lems to the same root cause: I in­ter­act with a larger num­ber of peo­ple, with fewer in­ter­ac­tions per per­son. On the one hand, fewer iter­a­tions means less penalty for “defec­tors”, and com­mon knowl­edge of this fact means less trust. On the other hand, more peo­ple + fewer in­ter­ac­tions per per­son means both less time and less men­tal re­sources to cus­tomize my in­ter­ac­tion with each in­di­vi­d­ual per­son. Thus, in a large com­pany, peo­ple are forced to rely more heav­ily on job ti­tles—and in a larger so­ciety, peo­ple are forced to rely more heav­ily on iden­tities more gen­er­ally.

  All the ex­am­ples listed in the OP are the sorts of things you’d ex­pect in a world with more peo­ple, more spe­cial­iza­tion, and fewer in­ter­ac­tions be­tween any given pair. (In some cases, this means zero in­ter­ac­tions be­tween a pair which would re­ally benefit from in­ter­act­ing, as in sev­eral of the ex­am­ples.)

  I don’t dis­agree that al­ign­ment & trust have a role here. But I do think that the large ma­jor­ity of real-world co­or­di­na­tion prob­lems could be solved by stick­ing the right two or three peo­ple in a room and just let­ting them talk for a full day. And in most cases, I think the rele­vant peo­ple would ac­tu­ally like to talk! The prob­lem is find­ing the right two or three peo­ple and build­ing that com­mu­ni­ca­tion chan­nel.

Real-World Co­or­di­na­tion Prob­lems are Usu­ally In­for­ma­tion Problems

• johnswentworth 13 Jun 2019 16:25 UTC

  2 points

  on: Some Ways Co­or­di­na­tion is Hard

  Is “shilling point” some new thing I’ve never heard of, or is this just an­other spel­ling of “Schel­ling point”? I as­sume the lat­ter, but it sounds like a name some­one would come up with for a con­cept similar-to-but-slightly-differ­ent-from a Schel­ling point.

• johnswentworth 5 Jun 2019 22:52 UTC

  40 points

  on: Ma­jor Up­date on Cost Disease

  I looked through Tabar­rok’s book, and my gen­eral im­pres­sion is:

  • He does a de­cent job go­ing through the ed­u­ca­tion data. (At least, he finds the same stuff I did when I went through the data a few years ago.)

  • He to­tally whiffs on health­care data. In par­tic­u­lar, I saw no men­tion of de­mo­graphic shifts, which are far and away the biggest driver of growth in US health­care spend­ing.

  • He then takes a hard left turn and goes off talk­ing about the Bau­mol effect, with­out ground­ing it very well in the data. He gives a bunch of qual­i­ta­tive ar­gu­ments that things look con­sis­tent with Bau­mol, but never makes the quan­ti­ta­tive ar­gu­ments that Bau­mol ex­plains all the growth, and never prop­erly rules out al­ter­na­tive hy­pothe­ses.

  A good ex­am­ple is on page 50: “it is ev­i­dent that de­spite higher costs Amer­i­cans have cho­sen to buy more health­care out­put over time. Once again, this is con­sis­tent with the Bau­mol effect but in­con­sis­tent with a purely cost-driven ex­pla­na­tion for ris­ing prices.” It’s also con­sis­tent with the de­mand curve shift­ing up over time, and Bau­mol hav­ing noth­ing to do with it. Which is ex­actly what we’d ex­pect in an ag­ing pop­u­la­tion.

  He does do a de­cent job rul­ing out “purely cost-driven ex­pla­na­tions” as an al­ter­na­tive hy­poth­e­sis, but that does not im­ply Bau­mol.

• johnswentworth 2 Jun 2019 17:58 UTC

  5 points

  in reply to: Paperclip Minimizer's comment on: The Fun­da­men­tal The­o­rem of As­set Pric­ing: Miss­ing Link of the Dutch Book Arguments

  Great ques­tion. The setup here as­sumes zero in­ter­est rates—in par­tic­u­lar, I’m im­plic­itly al­low­ing bor­row­ing with­out in­ter­est via short sales (real-world short sales charge in­ter­est). Once we al­low for nonzero in­ter­est, there’s a rate charged to bor­row, and the price of each as­set is its dis­counted ex­pected value rather than just ex­pected value. That’s one of sev­eral mod­ifi­ca­tions needed in or­der to use this the­o­rem in real-world fi­nance. (The same ap­plies to the usual pre­sen­ta­tion of the Dutch Book ar­gu­ments, and the same mod­ifi­ca­tion is pos­si­ble.)

• johnswentworth 2 Jun 2019 7:08 UTC

  11 points

  in reply to: Said Achmiz's comment on: The Fun­da­men­tal The­o­rem of As­set Pric­ing: Miss­ing Link of the Dutch Book Arguments

  We’re not ac­tu­ally talk­ing about the VNM for­mal­ism here. That’s why the “in the ab­sence of un­cer­tainty” part is im­por­tant.

  We have a finite set of world-states and prefer­ences over those world-states. We do not care about prefer­ences over ran­dom mix­tures of world-states, we don’t even have a no­tion of ran­dom mix­tures, just the de­ter­minis­tic states them­selves. We want a util­ity func­tion which en­codes our prefer­ences over those de­ter­minis­tic world-states.

  In the ab­sence of un­cer­tainty, we don’t ac­tu­ally need the con­ti­nu­ity as­sump­tion or the in­de­pen­dence as­sump­tion for any­thing. They don’t even make sense; we need a no­tion of ran­dom mix­tures just to state those as­sump­tions. VNM util­ity needs those be­cause it’s try­ing to get ex­pected util­ity max­i­miza­tion right out the door. But we’re not start­ing from VNM util­ity, we’re start­ing from de­ter­minis­tic util­ity.

  Whether we need com­plete­ness or not is more de­bat­able. It de­pends on how we’re in­ter­pret­ing miss­ing prefer­ences. If we in­ter­pret miss­ing prefer­ences as “I don’t know”, then it seems nat­u­ral to al­low the util­ity func­tion to give any pos­si­ble prefer­ence for that pair. In that case, lack of com­plete­ness may mean our util­ity func­tion isn’t unique, but it won’t pre­vent a util­ity func­tion from ex­ist­ing.

  It’s ex­actly the same in Eliezer’s post. His cir­cu­lar prefer­ences ar­gu­ment comes be­fore ran­dom out­comes are even in­tro­duced. There’s no no­tion of ran­dom­ness at that point, no no­tion of lot­ter­ies, so he’s not talk­ing about VNM util­ity. The cir­cu­lar prefer­ences ar­gu­ment is not the VNM util­ity the­o­rem, it is a sep­a­rate thing which makes a differ­ent claim un­der weaker as­sump­tions. That does not make it in­cor­rect.

• johnswentworth 2 Jun 2019 3:16 UTC

  2 points

  in reply to: Mary Chernyshenko's comment on: When Ob­ser­va­tion Beats Experiment

  Right, we need to use ex­per­i­ments to figure out that Y is needed in the first place. That’s the “figur­ing out the struc­ture” part—figur­ing out what the rele­vant gears are and how they fit to­gether.

  Now, ex­per­i­ments also in­her­ently in­volve some kind of ob­ser­va­tion. You make a change, then ob­serve the effect of that change. In some cases, the ob­ser­va­tion built into the ex­per­i­ment may be enough to figure out the sys­tem’s state—that’s what hap­pens in your small doses idea. But this is a very in­di­rect (and likely er­ror-prone) way of figur­ing out that Y is high in our rat strain.

• johnswentworth 1 Jun 2019 20:40 UTC

  5 points

  on: The Fun­da­men­tal The­o­rem of As­set Pric­ing: Miss­ing Link of the Dutch Book Arguments

  I’ve tried to min­i­mize the tech­ni­cal pre­req­ui­sites for this post, but it’s still very ab­stract and mathy. If you un­der­stand it and can write well, please con­sider writ­ing up a more hu­man-read­able ver­sion which builds around a con­crete ex­am­ple or two rather than keep­ing ev­ery­thing ab­stract. Alter­na­tively, if you are Eliezer Yud­kowsky, con­sider in­te­grat­ing the FTAP into that great in­tro I linked above.

  I’ll prob­a­bly get around to writ­ing a more con­crete ver­sion of this post even­tu­ally, but I wanted to get the idea out there, since hardly any­one seems to know about it.

The Fun­da­men­tal The­o­rem of As­set Pric­ing: Miss­ing Link of the Dutch Book Arguments

• johnswentworth 1 Jun 2019 18:23 UTC

  4 points

  in reply to: Mary Chernyshenko's comment on: When Ob­ser­va­tion Beats Experiment

  Yeah, that gets into the tech­ni­cal de­tails brushed un­der the rug. There’s two rele­vant types of equa­tions gov­ern­ing the equil­ibrium value of PP:

  • The ther­mo­dy­namic equil­ibrium equa­tion, which fixes a product/​ra­tio of the con­cen­tra­tions of the 3 species (lin­ear in log-con­cen­tra­tions)

  • The sto­i­chio­met­ric con­straints, which fix a cou­ple lin­ear com­bi­na­tions of the con­cen­tra­tions of the 3 species (lin­ear in con­cen­tra­tions)

  I’m effec­tively as­sum­ing that the ther­mo­dy­nam­ics fa­vor X + Y over PP, so that the sto­i­chio­met­ric con­straints can be ap­prox­i­mated as “X and Y con­cen­tra­tions are each fixed”—there’s never enough PP pro­duced to sig­nifi­cantly de­crease them. That way, we can ig­nore the sto­i­chio­met­ric limit (so long as X and Y are abun­dant), and just pay at­ten­tion to the equil­ibrium equa­tion. Then log-con­cen­tra­tion of PP is a pos­i­tive lin­ear func­tion of log-con­cen­tra­tions of X and Y, so ev­ery­thing is easy to think about. In­creas­ing/​de­creas­ing ei­ther X or Y by enough can always shift the equil­ibrium PP above/​be­low the thresh­old.

  The prob­lem with sep­a­rate re­ac­tions (X → PP and Y → PP) is that, if Y is high, then in­creas­ing or de­creas­ing X does noth­ing—PP will always be above thresh­old re­gard­less of the X level. It’s an or-gate, rather than a lin­ear func­tion. Similarly, if PP were mainly de­ter­mined by sto­i­chio­met­ric limits in my origi­nal re­ac­tions, we’d have an and-gate.

• johnswentworth 1 Jun 2019 4:05 UTC

  4 points

  in reply to: Ruby's comment on: Feed­back Re­quested! Draft of a New About/​Wel­come Page for LessWrong

  I definitely think it’s fine for the short term. I don’t want to push pre­ma­ture perfec­tion­ism here—this will not make the site worse than it is, and may make it bet­ter.

  I wouldn’t want it to go up and then for­get about it, and have sev­eral years of new­com­ers drop­ping off be­cause the en­try point didn’t grab them. (I’m less con­cerned about perfect­ing a page whose pur­pose is not en­try-point.) That said, if I’m ever re­ally un­happy about it, I can always just draft some­thing up my­self and then pro­pose it to you guys.

• johnswentworth 1 Jun 2019 3:25 UTC

  11 points

  in reply to: Raemon's comment on: Feed­back Re­quested! Draft of a New About/​Wel­come Page for LessWrong

  As writ­ten, it feels like this is try­ing to mix some as­pects of a mis­sion state­ment and some as­pects of an en­try point, and do­ing nei­ther one very well. A lot of it comes out sound­ing like bland gener­ispeak—not all of it, but a lot. It would be easy to make it more en­gag­ing if that’s what we’re go­ing for, or more in­for­ma­tive if that’s the ob­jec­tive, etc—it needs some goal that says what read­ers are meant to get out of it, and then more fo­cus on achiev­ing that goal.

  (Sorry if that sounds overly harsh. I’m view­ing this thread as a round of edit­ing, so cri­tiquing it as a piece of writ­ing seems right.)

• johnswentworth 1 Jun 2019 1:50 UTC

  9 points

  on: Feed­back Re­quested! Draft of a New About/​Wel­come Page for LessWrong

  A few other things ra­tio­nal­ity is not:

  • An aes­thetic prefer­ence for square grids

  • Asser­tion of the su­pe­ri­or­ity of Western culture

  • The be­lief that cre­den­tialed ex­perts always know better

  • Aban­don­ing your ethics/​morals

  • Always defect­ing in the pris­on­ers’ dilemma

  I would guess that mak­ing it clear what we’re not talk­ing about is more im­por­tant to hook­ing new peo­ple than pre­cisely defin­ing ra­tio­nal­ity. Also, I would avoid us­ing the word “truth” ex­plic­itly in the “what is ra­tio­nal­ity” sec­tion.

  More gen­er­ally, if the pur­pose of this page is be an en­try point, I would front-load it with more hooks, ex­am­ples, and links, and push less hook-y things to­ward the end. On a meta-level, if it’s go­ing to serve as an en­try point, then it’s also a key page to in­stru­ment with track­ing, a/​b test copy, and all that jazz. On the other hand, if the main pur­pose of the page is to serve as a mis­sion state­ment or some­thing along those lines, then parts ex­plic­itly aimed at new­com­ers could be di­aled back, es­pe­cially things like “What is ra­tio­nal­ity” or “Why should I care” that are ad­dressed within the se­quences.

• johnswentworth 1 Jun 2019 0:37 UTC

  8 points

  in reply to: mr-hire's comment on: When Ob­ser­va­tion Beats Experiment

  I was read­ing The Biol­ogy of Aging. Sev­eral times, the au­thor says some­thing to the effect of “ex­per­i­ment beats ob­ser­va­tion, there­fore the gold stan­dard in ag­ing re­search is to show that adding X ac­cel­er­ates ag­ing and re­mov­ing X slows it.” For many val­ues of X, ex­per­i­men­tal ma­nipu­la­tion definitely makes or­ganisms live longer/​shorter, but it’s not clear that X ac­tu­ally changes sig­nifi­cantly dur­ing ac­tual ag­ing.

  So in this in­stance, we don’t fully un­der­stand how the sys­tem works ei­ther, but a mea­sure­ment of X could still tell us whether it’s rele­vant (based on whether it changes).

When Ob­ser­va­tion Beats Experiment

• johnswentworth 25 May 2019 0:27 UTC

  2 points

  on: How would you take over Rome?

  Gen­eral notes, be­fore I ac­tu­ally pro­pose a solu­tion:

  • A lot of pro­pos­als so far in­volve things like “use my ed­u­ca­tion and pre­dic­tive tal­ents to achieve a high po­si­tion in so­ciety”. Given how quickly smart peo­ple with amaz­ing pre­dic­tive tal­ents walk into the White House to­day, com­bined with the poli­ti­cal tal­ents of LWers, I doubt that has any hope of work­ing.

  • A lot of pro­pos­als in­volve in­tro­duc­ing tech­nolo­gies like mills, can­non, etc which would be mas­sive cap­i­tal out­lays by the stan­dards of the time. You’d need to already have an em­pire’s worth of re­sources on hand, and early-mod­ern tech prob­a­bly still wouldn’t be enough to make the econ­omy more effi­cient right away.

  • In terms of real eco­nomic value, prob­a­bly the largest chunk of po­ten­tial is corn and pota­toes—both are New World plants, corn has caloric yield an or­der of mag­ni­tude higher than wheat (in terms of both land and la­bor re­quire­ments), and pota­toes don’t get burned when an army comes by. If I could man­age a round trip to the New World, bring­ing back those crops would be huge—al­though that still leaves the ques­tion of how to cap­ture the value. Con­versely, it would be hard for any an­cient econ­omy to ac­cu­mu­late cap­i­tal—and thus be­gin the trek to au­toma­tion—with­out higher-yield crops. Un­for­tu­nately, a New World round-trip would it­self take a mas­sive cap­i­tal out­lay.

  Think­ing about low-cap­i­tal-in­vest­ment tech which could be im­ple­mented with­out ad­vanced man­u­fac­tur­ing...

  • RSA en­cryp­tion & signing

  • log­a­r­ithms & the slide rule

  • sex­tant & com­pass & as­so­ci­ated nav­i­ga­tion techniques

  • 18th-cen­tury de­sign of chim­neys and stoves (be­fore which they were im­prac­ti­cal for heat­ing)

  • demon­stra­tion-scale tele­graph (two mag­nets and some cop­per wire would suffice for ma­te­rial)

  • maybe very rudi­men­tary ra­dio (mag­nets, cop­per wire and cop­per foil suffice in prin­ci­ple)

  • aniline dyes—rel­a­tively easy once you know to look for them, and would be worth a for­tune in an­cient Rome

  • in­cre­men­tal met­al­lur­gi­cal improvements

  … I’m sure there’s more to think of here, but that should be enough for a vi­able solu­tion.

  
  

  Proposal

  Tech is the big edge, ob­vi­ously. Moderns don’t have a ma­jor edge in poli­tics or war (ab­sent the tech needed for mod­ern mil­i­tary doc­trine). Eco­nom­i­cally, most tech in­volves large cap­i­tal out­lays, and value cap­ture is difficult when you’re not already em­peror. Given all that, gen­eral shape of solu­tion I want is:

  1. ac­cu­mu­late some ini­tial re­sources us­ing a low-bar­rier tech­nolog­i­cal advantage

  2. bring to­gether a rel­a­tively small group of peo­ple (com­pany-sized), in a re­mote lo­ca­tion, for rapid ver­ti­cal de­vel­op­ment of higher tech—not nec­es­sar­ily all the way to moder­nity, but enough to de­ci­sively swing the bal­ance in a war

  3. Go shop­ping for al­lies who don’t like the Romans

  Most of the effort is in step 2 - our rel­a­tive ad­van­tage is tech, so our strat­egy puts most of the work in tech.

  Let’s flesh all that out.

  • Aniline dyes are the hot item for step 1 - they don’t re­quire ad­vanced man­u­fac­tur­ing or ma­te­ri­als, cap­i­tal in­vest­ment re­quired is rel­a­tively low, and given that an­cient Rome was already hem­or­rhag­ing gold to buy Chi­nese silk, they’d definitely sell like hot­cakes. Care must be taken to make sure we paid for the dyes rather than ar­rested or en­slaved some­how, but that’s prob­a­bly tractable.

  • Round trip to the Amer­i­cas is the next big step—we can’t sup­port peo­ple off the trade grid with­out high-yield crops, and fer­til­izer alone won’t get us there. Out­fit­ting a long-range ship should be vi­able for a dye-lord, and mod­ern knowl­edge of sailing and the trade winds should help out. An­cient sea­far­ers did not like leav­ing sight of land, but that’s the kind of prob­lem which can be solved by throw­ing money at it—and we’d want to train them in mod­ern sailing any­way. Pack to trade in both di­rec­tions, the voy­age should be quite prof­itable, and we come back with all-im­por­tant corn and pota­toes.

  • At this point, it’s vi­able to hire a cou­ple hun­dred peo­ple with vary­ing spe­cialties and move to a re­mote patch of land some­where on the edge of the em­pire where no­body will pay much at­ten­tion, with run­ning wa­ter and wood. Coastal would be use­ful, but not if it in­creases visi­bil­ity too much. Get the corn and pota­toes grow­ing, and it should be vi­able to sup­port the whole crowd with a fairly small agri­cul­tural base.

  • Now we have the base to make some cap­i­tal in­vest­ments while safely cap­tur­ing their value, with­out fear of seizure. Let’s say we have half of our two hun­dred peo­ple farm­ing (com­pa­rable ra­tio to early mod­ern England); most of the re­main­ing hun­dred are ar­ti­sans—smiths, car­pen­ters, etc. They’ll build most of our cap­i­tal as­sets. We’ll im­port most raw ma­te­ri­als, like wool and metal ore, plus some pro­cessed ma­te­rial like sail­cloth (un­til we can pro­duce it our­selves), nec­es­saries like salt, and what­ever lux­u­ries our team can en­joy with­out at­tract­ing un­due at­ten­tion.

  • Early cap­i­tal as­sets are mainly ships and wa­ter mills—the former for trade, the lat­ter for au­toma­tion. Note that over­land trade is not a high value prop—ships in any era have ca­pac­ity mul­ti­ple or­ders of mag­ni­tude higher than wag­ons/​trucks, and pre­mod­ern roads were pretty bad any­how. The ships will sail the trade routes, and should provide am­ple profit for any­thing we need to buy. (We won’t be able to keep our nav­i­ga­tion se­crets un­der wraps for­ever, but they should last long enough—an­cient sailors’ fear of the open sea works in our fa­vor here.)

  • With plenty of wood and ar­ti­sans, the mills can provide power for what­ever light in­dus­try we want, but that in­dus­try it­self will prob­a­bly not be prof­itable—la­bor is very cheap rel­a­tive to cap­i­tal in the pre­mod­ern world. The point of the mills is to power any­thing we want to keep away from pry­ing eyes, in our re­mote cor­ner of the world. Me­tal­lurgy will be the first and biggest pri­or­ity—we can im­port ore, but the se­crets of steel need to stay lo­cal.

  • The pre­ci­sion man­u­fac­tur­ing feed­back loop: build more pre­cise tools, and those let us build even more pre­cise tools. High-qual­ity ma­te­ri­als are a limit­ing fac­tor for that loop, thus the im­por­tance of met­al­lurgy.

  • Build­ing a dy­namo shouldn’t be too hard, once we’ve im­ported mag­nets. We’ll want to pro­duce crap-tons of cop­per wire for any­thing elec­tric—that shouldn’t pre­sent any difficul­ties. Elec­tric­ity would mainly be use­ful for chem­i­cal pro­cesses, e.g. split­ting wa­ter—com­mu­ni­ca­tion wouldn’t mat­ter much in our tiny re­mote town, and we don’t want any se­crets get­ting out. We’d need a dam to get use­ful power pro­duc­tion, but dam-build­ing isn’t too hard.

  • Hand-cranked ra­dios are still a maybe, but they’d be pretty cool and a huge mil­i­tary ad­van­tage.

  • Haber-Bosch pro­cess is the next big jump: with a source of am­mo­nia, both chem­i­cal fer­til­izer and ex­plo­sives be­come vi­able. (We could go the old-fash­ioned route for ex­plo­sives, but Rule of Cool.) Against the famed le­gion pha­lanxes, we’re look­ing at tight-packed tar­gets—ideal for things that go boom.

  • At this point, we’d prob­a­bly have the pieces to move on Rome. A de­cent-size trade fleet pro­vides both sup­port in­fras­truc­ture and mo­bil­ity for any troops we ally with, and a trade busi­ness pro­vides con­nec­tions through­out the em­pire and sur­round­ing ar­eas. Ex­plo­sives would, hon­estly, be as much about mar­ket­ing our­selves to po­ten­tial al­lies as about ac­tu­ally win­ning fights, but they’d be awe­some mar­ket­ing. En­cryp­tion and (maybe) prim­i­tive ra­dios would be a less flashy but more prac­ti­cal mil­i­tary ad­van­tage, if it ac­tu­ally came to a fight. All that’s left is to find some­one with a rea­son­ably-sized mil­i­tary who’s more in­ter­ested in beat­ing the Ro­mans than in tak­ing their stuff, and strike a deal.

• johnswentworth 24 May 2019 22:08 UTC

  2 points

  in reply to: Logos01's comment on: How would you take over Rome?

  Ma­jor bar­rier right at the start: buy­ing a lit­tle chunk of land is one thing, but a bunch of the stuff in step 2 will re­quire very large cap­i­tal out­lays by the stan­dards of the time—es­pe­cially the mill.

  There’s a rea­son au­toma­tion didn’t catch on be­fore the in­dus­trial rev­olu­tion: cap­i­tal was scarce and peas­ants were abun­dant. It wouldn’t be easy to ac­tu­ally pro­duce crops more cost-effec­tively than peas­ant la­bor, when the peas­ant la­bor is ab­surdly cheap. Things like fer­til­izer could help, but even that will run into prob­lems—peo­ple don’t like poop on their food, and chem­i­cal fer­til­izer/​in­sec­ti­cide re­quires rel­a­tively com­pli­cated man­u­fac­tur­ing fa­cil­ities.