ryan_b

• ryan_b 29 May 2020 15:15 UTC

  2 points

  on: Spec­u­la­tions on the Fu­ture of Fic­tion Writing

  I have long sus­pected that the prob­lem of bad writ­ing in movies is largely driven by ques­tions of com­plete­ness and adap­ta­tion. For ex­am­ple:

  • We rarely see the whole story. Even if it is was shot ex­actly as writ­ten, what we wind up see­ing is an ed­ited-down cut of the film; which chunks of the writ­ing get left out makes a big differ­ence to me in my per­cep­tion of the writ­ing. Con­sider the case of the char­ac­ter that sud­denly sprouts new abil­ities, which is shit writ­ing. Watch­ing a later di­rec­tor’s cut, they of­ten in­clude the scene which in­cludes the cru­cial ex­pla­na­tion of why they have these.

  • The writ­ing doesn’t stay the same over the course of filming. There may be prac­ti­cal im­ped­i­ments to a key scene, like the weather ru­in­ing out­door shots; it may prove in­fea­si­ble to get a good enough set/​cos­tume/​stunt ar­range­ment to drive a part of the story; maybe the ac­tor just can’t pull it off to save their lives. This ne­ces­si­tates re-writes. I strongly ex­pect these to lack the co­herency of the origi­nal screen­play, be­cause now there are lots of peo­ple with in­put rather than the group ac­cept­ing a com­pleted script with a sin­gle au­thor.

• ryan_b 29 May 2020 14:13 UTC

  2 points

  in reply to: Giammi T’s comment on: Open & Wel­come Thread—May 2020

  I feel like there is no con­flict here; in fact it is widely con­sid­ered a deal-breaker for a client to be guilty and lie to their at­tor­ney about it. A client ly­ing to a lawyer is one of the eth­i­cally ac­cepted rea­sons to dump a client they have already agreed to serve. This isn’t even pro-forma; in prac­tice, lawyers don’t blame one an­other for dump­ing clients that lie to them. Nor is it con­sid­ered a black mark for fu­ture hiring with other law firms.

  The im­por­tant vari­ables here are that the lawyer is con­strained by the ev­i­dence, but they have a duty to their client. This is be­cause lawyers are not fact fin­ders; they are ad­vo­cates. The Amer­i­can trial sys­tem em­ploys the ‘ar­gu­ments are sol­diers’ sys­tem speci­fi­cally and de­liber­ately, then it has a lot of rules for set­ting a floor on how bad the ar­gu­ments can be and re­lies on nom­i­nally-neu­tral third par­ties (a judge and/​or jury) to as­sess them.

  Con­sider that a lawyer can rep­re­sent them­selves, their fam­ily, or par­ties in whom they have a fi­nan­cial stake with­out con­flict of in­ter­est. How­ever it is con­sid­ered a con­flict of in­ter­est if they have a fi­nan­cial stake in the other party, or any­thing else that might com­pro­mise their com­mit­ment to ad­vo­cacy of their client.

  So at least in the Amer­i­can sys­tem, I put it to you that ac­cept­ing the case with to­tal cer­tainty your client is guilty is both eth­i­cal and ra­tio­nal.

• ryan_b 28 May 2020 21:20 UTC

  2 points

  in reply to: Aaro Salosensaari’s comment on: Open & Wel­come Thread—May 2020

  I am not versed in eco­nomics liter­a­ture, so I can’t meet your need. But I have also en­coun­tered er­god­ic­ity eco­nomics, and thought it was in­ter­est­ing be­cause it had good mo­ti­va­tions.

  I am skep­ti­cal for an en­tirely differ­ent rea­son; I en­coun­tered er­godic the­ory be­fore­hand in the con­text of ther­mo­dy­nam­ics, where it has been harshly crit­i­cized. I in­stinc­tively feel like if we can do bet­ter in ther­mo­dy­nam­ics, we can em­ploy the same math to do bet­ter in other ar­eas.

  Of course this isn’t nec­es­sar­ily true: er­godic the­ory might cleave re­al­ity bet­ter when de­scribing an econ­omy than gas par­ti­cles; there is prob­a­bly a sig­nifi­cant differ­ence be­tween the eco­nomics ver­sion and the ther­mo­dy­nam­ics ver­sion; the crit­i­cism of er­godic the­ory might be ass-wrong (I don’t think it is, but I’m not qual­ified enough for strong con­fi­dence).

• ryan_b 28 May 2020 20:45 UTC

  2 points

  in reply to: michaelcohen’s comment on: Pre­dicted Land Value Tax: a bet­ter tax than an unim­proved land value tax

  Okay that makes sense, but now I’m con­fused on ex­actly how the real prices re­late to the pre­dic­tions. I ex­pect the de­tails of that mechanism to be the crux of the is­sue; ex­actly how the price up­dat­ing is done will de­ter­mine who the win­ners and losers are rel­a­tive to the de­sired out­come.

  It still feels like solv­ing that prob­lem well would be tan­ta­mount to solv­ing the unim­proved value prob­lem, but I’m perfectly happy to be wrong.

• ryan_b 27 May 2020 18:23 UTC

  2 points

  in reply to: michaelcohen’s comment on: Pre­dicted Land Value Tax: a bet­ter tax than an unim­proved land value tax

  A lens is a more struc­tured form of per­spec­tive, the way we use the term in the com­mu­nity; the em­pha­sis is on be­ing able to move be­tween differ­ent ones. We tend to use it for differ­ent an­a­lyt­i­cal frame­works (CS, econ, en­g­ineer­ing, fi­nance, etc).

  The in­dus­trial fa­cil­ity effects will be vari­able de­pend­ing on what the fa­cil­ity is. For ex­am­ple, a nu­clear power plant or a microchip fac­tory will tend to in­crease prop­erty val­ues be­cause of an in­flux of good pay­ing jobs but a coal power plant or a meat fac­tory will tend to de­crease prop­erty val­ues be­cause they reek and are mis­er­able to live near.

  But the broader point is that the pre­dic­tions are based on the price his­tory of the last 10 years, when there was no such ma­jor price im­pact. I ex­pect the pre­dic­tion soft­ware to con­tinue pre­dict­ing sta­ble prices, which means the peo­ple who live with a new meat fac­tory are pay­ing taxes based on too high a value, and the peo­ple who live next to a new microchip fac­tory are pay­ing based on too low a value. Th­ese would even­tu­ally even out as the ac­tual sales en­ter the price his­tory, but this is a long lead time. I also ex­pect that the price would be dis­torted by the un­der­stand­ing that some places are tax bar­gains, and some tax banes, which would ex­tend the time for the pre­dic­tions to cor­rect back to true value.

  It looks to me like the same mechanism that pre­serves the in­cen­tive to im­prove your land works to ex­clude sig­nifi­cant changes in land value more gen­er­ally; all the di­rec­tions I can en­vi­sion for solv­ing this look sus­pi­ciously like solv­ing the prob­lem of as­sess­ing the unim­proved value of land.

• ryan_b 27 May 2020 14:14 UTC

  2 points

  on: Pre­dicted Land Value Tax: a bet­ter tax than an unim­proved land value tax

  I ap­pre­ci­ate first stabs at im­prove­ments in gov­er­nance; up­voted. Would I be cor­rect in in­fer­ring you are think­ing about this largely through a com­puter sci­ence lens, rather than an eco­nomic one?

  What do you think about deal­ing with price dis­con­ti­nu­ities? I have in mind things like: ma­jor hous­ing de­vel­op­ment in formerly ru­ral ar­eas due to ur­ban ex­pan­sion; the con­struc­tion of a new in­dus­trial fa­cil­ity; the dis­cov­ery of nat­u­ral re­sources be­neath the land; the ma­jor lo­cal in­dus­try col­laps­ing; crit­i­cal in­fras­truc­ture failures like the wa­ter sup­ply be­ing con­tam­i­nated.

  I’m con­fused about the re­la­tion­ship be­tween the bids on the prop­erty (the value) and the out­put of the pre­dic­tion en­g­ine (the pre­dicted value) as a con­se­quence of the above. It looks like the pre­dic­tion soft­ware is de­signed to ex­clude price dis­con­ti­nu­ities in its value es­ti­mate, which is how it pre­serves the in­cen­tive to im­prove the land; but this means we are sys­tem­at­i­cally pre­dict­ing wrong on pur­pose, which feels weird. The bids them­selves don’t ap­pear to have any im­por­tance.

• ryan_b 5 May 2020 19:34 UTC

  4 points

  in reply to: Hazard’s comment on: An Un­der­grad­u­ate Read­ing Of: Macro­scopic Pre­dic­tion by E.T. Jaynes

  I think this is cap­tured in Sec­tion 5, the Max­i­mum Cal­iber Prin­ci­ple:

  We are given macro­scopic in­for­ma­tion A which might con­sist of val­ues of sev­eral phys­i­cal quan­tities . . . such as dis­tri­bu­tion of stress, mag­ne­ti­za­tion, con­cen­tra­tion of var­i­ous chem­i­cal com­po­nents, etc. in var­i­ous space time re­gions. This defines a cal­iber . . . which mea­sures the num­ber of time de­pen­dent microstates con­sis­tent with the in­for­ma­tion A.

  So the idea is that you take the macro in­for­ma­tion A, use that to iden­tify the space of pos­si­ble microstates. For max­i­mum rigor you do this in­de­pen­dently for A and B, and if they do not share any microstates then B is im­pos­si­ble. When we make a pre­dic­tion about B, we choose the value of B that has the biggest over­lap with the pos­si­ble microstates of A.

  He talks a lit­tle bit more about the mo­ti­va­tion for do­ing this in the Con­clu­sion, here:

  We should cor­rect a pos­si­ble mis­con­cep­tion that the reader may have gained. Most re­cent dis­cus­sions of macrophe­nom­ena out­side of phys­i­cal chem­istry con­cen­trate en­tirely on the dy­nam­ics (micro­scopic equa­tions of mo­tion or an as­sumed dy­nam­i­cal model at a higher level, de­ter­minis­tic or stochas­tic) and ig­nore the en­tropy fac­tors of macrostates al­to­gether. In­deed, we ex­pect that such efforts will suc­ceed fairly well if the macrostates of in­ter­est do not differ greatly in en­tropy.

  Em­pha­sis mine. So the idea here is that if you don’t need to ac­count for the en­tropy of A, you will be able to tackle the prob­lem us­ing nor­mal meth­ods. If the nor­mal meth­ods fail, it’s a sign that we need to ac­count for the en­tropy of A, and there­fore to use this method.

  I can’t do the physics ex­am­ples ei­ther ex­cept in very sim­ple cases. I am com­forted by this line:

  Although the math­e­mat­i­cal de­tails needed to carry it out can be­come al­most in­finitely com­pli­cated...

• ryan_b 4 May 2020 18:45 UTC

  15 points

  on: In­sights from Eu­clid’s ‘Ele­ments’

  This is glo­ri­ous. On the flip side of the coin, I strug­gle with out­rage that we had copped to the prob­lem of pre­sent­ing in­for­ma­tion and ba­si­cally had it licked in the mid­dle of the 19th cen­tury, and then ap­par­ently sys­tem­at­i­cally purged such knowl­edge dur­ing the 20th. For ex­am­ple, there’s this in­ter­est­ing piece about Emma Willard, who drew gor­geous vi­su­als pro­vid­ing per­spec­tive to his­tory. She be­gan in ~1837. Good use of images seems only now to be un­der­go­ing a re­nais­sance, and that ow­ing to the availa­bil­ity of com­puter graph­ics more than any­thing else.

  What the devil hap­pened erst­while?

• ryan_b 29 Mar 2020 16:37 UTC

  4 points

  in reply to: John_Maxwell’s comment on: Com­pre­hen­sive COVID-19 Dis­in­fec­tion Pro­to­col for Pack­ages and Envelopes

  Well that sucks. Take care of your­self and stay sane dur­ing iso­la­tion!

• ryan_b 25 Mar 2020 17:32 UTC

  2 points

  in reply to: Douglas_Knight’s comment on: His­tory’s Biggest Nat­u­ral Experiment

  I feel like this is ev­i­dence for the nat­u­ral ex­per­i­ment in­ter­pre­ta­tion. This means we will get a steady stream of new find­ings as each mat­u­ra­tion win­dow ap­proaches, for decades to come.

• ryan_b 24 Mar 2020 2:12 UTC

  3 points

  in reply to: Matthew Barnett’s comment on: Are vet­er­ans more self-dis­ci­plined than non-vet­er­ans?

  To be more ex­act, if you have a group, then the group pro­vides so­cial in­cen­tives; but so­cial in­cen­tives do not im­ply a group. For ex­am­ple, if I were pub­li­cly hu­mil­i­ated in front of strangers, they might mock me if they saw me later in a restau­rant. This is a so­cial (dis)in­cen­tive, but the fact re­mains that we aren’t in a group.

  What qual­ifies peo­ple as a group in the sense that I in­tend is at least twofold: they have to share the same set of in­cen­tives; this fact has to be com­mon knowl­edge among them.

  I do agree that if per­son trains suc­cess­fully it would im­prove long-run dis­ci­pline, but do­ing mil­i­tary train­ing won’t mean­ingfully change the out­come from non-mil­i­tary train­ing be­cause the group con­text is what does the ex­tra work. If that is not the fo­cus, ie vet­er­ans are just an ex­am­ple dis­ci­plined pop­u­la­tion, then my com­ments are prob­a­bly not rele­vant to the true con­cern.

• ryan_b 23 Mar 2020 19:24 UTC

  12 points

  in reply to: Scott Alexander’s comment on: Can crimes be dis­cussed liter­ally?

  It seems to me pre­cisely the op­po­site: my read­ing is that Ben­quo is driv­ing ex­actly at how to talk about the prob­lem of sys­temic falsifi­ca­tion of in­for­ma­tion.

  If the post is non­cen­tral, what is the cen­tral thing in­stead?

• ryan_b 23 Mar 2020 16:21 UTC

  6 points

  on: Are vet­er­ans more self-dis­ci­plined than non-vet­er­ans?

  I am a vet­eran, and my in­side view sug­gests two things: one, the least dis­ci­plined mem­bers of the pop­u­la­tion are filtered out by the mil­i­tary (which is to say they are not ac­cepted or kicked out early); two, the mil­i­tary ex­pe­rience pushes vet­er­ans to­wards the ex­tremes.

  Rea­sons to con­sider that vet­er­ans would be more pro­duc­tive than av­er­age:

  • Ac­cli­mated to long and/​or stren­u­ous work pe­ri­ods.

  • Bet­ter ac­cess to ed­u­ca­tion through vet­er­ans pro­grams and ad­mis­sion boosts.

  • Direct boost to em­ploy­a­bil­ity in a va­ri­ety of in­dus­tries.

  Rea­sons to con­sider that vet­er­ans would be less pro­duc­tive than av­er­age:

  • Higher rates of homelessness

  • Higher rates of men­tal ill­ness and suicide

  • Higher rates of sub­stance abuse

  • Etc.

  My ex­pec­ta­tion is that the pro­duc­tivity ad­van­tage is high­est when vet­er­ans en­ter a civilian in­dus­try that matches mil­i­tary tasks closely, like com­pli­ance with reg­u­la­tions or un­com­fortable work en­vi­ron­ments. I also ex­pect that the vet­er­ans who fail to re-adapt to civilian life suffer an al­most com­plete col­lapse of pro­duc­tivity.

  Turn­ing to the ques­tion of dis­ci­pline, I think we will benefit from a lit­tle con­text. Dis­ci­pline in the mil­i­tary is very much a team phe­nomenon; Army train­ing is fo­cused over­whelm­ingly on es­tab­lish­ing and main­tain­ing a group iden­tity. Most of the things peo­ple as­so­ci­ate with mil­i­tary dis­ci­pline re­quire other peo­ple to make sense, like the chain of com­mand, pul­ling se­cu­rity, and how tasks are di­vided. Even the in­di­vi­d­ual things like phys­i­cal fit­ness or mem­o­riz­ing trivia, are thor­oughly steeped in the team en­vi­ron­ment be­cause they are mo­ti­vated by be­ing able to help your buddy out and are how sta­tus is sorted in the group.

  I be­lieve your friend’s state­ment:

  If you can just train your­self like you’re in the army, then you can be­come just as self dis­ci­plined as a soldier

  is wrong as a con­se­quence, be­cause you can never train your­self like you are in the Army. That fun­da­men­tally needs a group, en­tirely sep­a­rate from the ques­tion of so­cial in­cen­tives and en­vi­ron­ment. Out­side of the group con­text, dis­ci­pline doesn’t re­ally mean any­thing more than habit for­ma­tion.

• ryan_b 19 Mar 2020 17:28 UTC

  4 points

  in reply to: Davidmanheim’s comment on: LessWrong Coron­avirus Agenda

  If the same type of fa­cil­ity works for al­most ev­ery kind of vac­cine, do we think there would be in­ter­est in con­struct­ing the fa­cil­ities as a spec­u­la­tive ven­ture? Con­sider:

  1. The econ­omy is in chaos and may re­main so, which I ex­pect to pro­duce un­usu­ally af­ford­able ac­cess to de­sign firms, con­struc­tion crews, raw ma­te­ri­als, and land.

  2. There will be a strong in­cen­tive for reg­u­la­tors/​in­spec­tors to move with best speed, and the cur­rent ad­minis­tra­tion at least in the US has a track record of be­ing friendly to short­cuts.

  3. If the fa­cil­ities are already built, this al­lows a limit to the risk the com­pa­nies pro­duc­ing the vac­cines need to ab­sorb in or­der to in­crease sup­ply.

  4. We could squeeze out un­scrupu­lous op­por­tunists.

• ryan_b 12 Mar 2020 21:36 UTC

  2 points

  on: The Amer­i­can Dream Is Alive In China

  My model for this is that China is achiev­ing suc­cess largely by ig­nor­ing ex­ter­nal­ities. En­vi­ron­men­tal pol­lu­tion is a prime ex­am­ple, like in the case of their pre­vi­ous re­cy­cling policy and min­ing of rare earth min­er­als. It is ac­tu­ally against the law for the US to build as quickly or as cheaply as China, but this is rea­son­ably mo­ti­vated by try­ing to ac­count for things like pol­lu­tion and safety, and avoid­ing things like re­set­tling en­tire towns.

  Chi­nese suc­cess looks a lot like the WWII and post­war years in the US, and for much the same rea­sons.

• ryan_b 7 Mar 2020 17:40 UTC

  2 points

  in reply to: Daniel Kokotajlo’s comment on: Cortés, Pizarro, and Afonso as Prece­dents for Takeover

  be­cause why was it that the con­quis­ta­dors were able to ex­ploit the lo­cals and not the other way around?

  Have you con­sid­ered the pos­si­bil­ity that it was a case of mu­tual ex­ploita­tion? The Aztec al­lies of the con­quis­ta­dors weren’t there out of the good­ness of their hearts; they had found a new an­gle that would help them defeat Tenochti­t­lan. They lost the post-vic­tory power strug­gle, but it was always go­ing to be some­one.

• ryan_b 10 Feb 2020 19:41 UTC

  2 points

  in reply to: Wei_Dai’s comment on: Have epistemic con­di­tions always been this bad?

  You make good points here. Any ideas why those other shifts hap­pened and how can we help re­verse them or pre­vent them from hap­pen­ing el­se­where?

  Mostly it looks to me like a se­ries of un­re­lated changes built up over time, and the un­in­tended con­se­quences were mostly ad­verse.

  An ex­am­ple is the War on Cancer and the changes that came with it to fund­ing. It had long been the case that fund­ing was mostly handed out on a pro­ject-by-pro­ject ba­sis, but in or­der to get the fund­ing ded­i­cated to can­cer re­search it was nec­es­sary to ex­plain how can­cer re­search would benefit. The ob­vi­ous first-or­der im­pact is an in­crease in ad­minis­tra­tive over­head for get­ting the money.

  Alongside this sci­ence sort of pro­fes­sion­al­ized. I ex­pect that when the sense of how im­por­tant some­thing is per­me­ates, pro­fes­sion­al­iza­tion is viewed as a nat­u­ral con­se­quence, but it seems to have mis­fired here. Pro­fes­sion­al­iza­tion, like other forms of la­bor or­ga­ni­za­tion, isn’t about max­i­miz­ing any­thing but about en­sur­ing a min­i­mum. This means things like more met­rics, which is why our civ­i­liza­tion for­mally prefers a lot of crappy sci­en­tific pa­pers to a few good ones, and doesn’t want any kind of non-pa­per pre­sen­ta­tion of sci­en­tific progress at all. Science jobs be­come sub­ject to Good­hart­ing, be­cause peo­ple start think­ing that the right way to get more sci­ence is just to in­crease the num­ber of sci­en­tists, on ac­count of them all be­ing in­ter­change­able pro­fes­sion­als with a re­li­able min­i­mum out­put.

  The uni­ver­sity en­vi­ron­ment also got leaned on as a lever for progress; the stu­dent loan pro­grams all grew over this same pe­riod, which seems to have driven a long pe­riod of com­pe­ti­tion for head­count. This shifted uni­ver­si­ties’ pri­ori­ties from ex­e­cut­ing their nom­i­nal mis­sion to­wards sig­nal­ling de­sir­a­bil­ity among stu­dents/​par­ents/​etc. I am cer­tain at least part of that came at the ex­pense of fac­ulty, even if only by in­creas­ing the ad­minis­tra­tive bur­den still fur­ther by yet more met­rics.

  On the fix­ing side, I am ac­tu­ally pretty op­ti­mistic. A few sim­ple things would prob­a­bly help a lot, two ex­am­ples be­ing fund­ing and or­ga­ni­za­tion. Ex­am­ple: Bell Labs and Xerox PARC have been dis­cussed here a lot. Both cases de­vi­ated sig­nifi­cantly from the stan­dard uni­ver­sity/​gov­ern­ment sys­tem of fund­ing in­di­vi­d­ual pro­jects case by case. Un­der the pro­ject/​grant sys­tem be­ing a sci­en­tist re­duces to be­ing able to suc­cess­fully get fund­ing for a se­ries of pro­jects over time. At Bell and at PARC, they rather made long-term in­vest­ments on a per­son-by-per­son ba­sis. I think this has wide-rang­ing effects, but not least among them is that there wasn’t a lot of ad­minis­tra­tive over­head to a given in­ves­ti­ga­tion; rather they could all be picked up, put down, or adapted as needed. Another effect, maybe in­ten­tional but seem­ingly hap­pen­stance, is that they built a com­mu­nity of re­searchers in the col­lo­quial sense. This is pretty differ­ent from the for­mal em­ployee re­la­tion­ships that dom­i­nate now. Around 7 years ago I listened to a re­cruit­ing pitch from San­dia Na­tional Lab­o­ra­to­ries for en­g­ineer­ing stu­dents, and asked how com­mu­ni­ca­tion was be­tween differ­ent groups in the lab. The rep­re­sen­ta­tive said that she knew of a case where two labs right across the hall from each other were in­ves­ti­gat­ing the same thing for over a year be­fore they re­al­ized it, be­cause no­body talks.

  This sug­gests to me that a uni­ver­sity that was strug­gling fi­nan­cially, or maybe just needed to take a gam­ble on mov­ing up in the world, could cheaply im­ple­ment what ap­pears to be a su­pe­rior re­search-pro­duc­ing ap­para­tus, just by shift­ing their meth­ods of fund­ing and track­ing re­sults.

• ryan_b 6 Feb 2020 17:56 UTC

  5 points

  on: ryan_b’s Shortform

  Are math proofs use­ful at all for writ­ing bet­ter al­gorithms? I saw on Red­dit re­cently that they proved Batch­e­lor’s Law in 3D, the core idea of which seems to be us­ing stochas­tic as­sump­tions to prove it can­not be vi­o­lated. The Quanta ar­ti­cle does not seem to con­tain a link to the pa­per, which is weird.

  Batch­e­lor’s Law is the ex­per­i­men­tally-ob­served fact that tur­bu­lence oc­curs at a spe­cific ra­tio across scales, which is to say when you zoom in on a small chunk of the tur­bu­lence it looks re­mark­ably like all of the tur­bu­lence, and so on. Some­thing some­thing frac­tals some­thing.

  Look­ing up the re­la­tion­ship be­tween proofs and al­gorithms mostly goes to proofs about spe­cific al­gorithms, and some­times us­ing al­gorithms as a form of proof; but what I am af­ter is whether a pure-math proof like the above one can be mined for use­ful in­for­ma­tion about how to build an al­gorithm in the first place. I have read el­se­where that al­gorith­mic effi­ciency is about prob­lem in­for­ma­tion, and this makes in­tu­itive sense to me; but what kind of in­for­ma­tion am I re­ally get­ting out of math­e­mat­i­cal proofs, as­sum­ing I can un­der­stand them?

  I don’t sup­pose there’s a list some­where that hand­ily matches tricks for prov­ing things in math­e­mat­ics to tricks for con­struct­ing al­gorithms in com­puter sci­ence?

ryan_b’s Shortform

• ryan_b 5 Feb 2020 17:44 UTC

  2 points

  in reply to: DanielFilan’s comment on: Have epistemic con­di­tions always been this bad?

  Oops! Fixed.