Dagon

• Dagon 19 May 2019 21:36 UTC

  −4 points

  in reply to: jessicata's comment on: Com­ment sec­tion from 05/​19/​2019

  Right. At least some ab­stract top­ics should be dis­cussed, and part of the dis­cus­sion is which, if any, speci­fics might be ex­em­plary of such ab­strac­tions. Other ab­stract top­ics should be avoided, if the rele­vant ex­am­ples are poli­ti­cally-charged and the ab­strac­tion doesn’t eas­ily en­com­pass other points of view.

  Choos­ing to dis­cuss ab­stracts pri­mar­ily which hap­pen to sup­port a spe­cific po­si­tion, with­out dis­clos­ing that tie, is not OK. It’s dis­cussing the spe­cific in the guise of the ab­stract. I can’t be sure that’s what Zack is do­ing, but that’s how it ap­pears from my out­sider view­point.

  
  
  

• Dagon 19 May 2019 15:24 UTC

  3 points

  in reply to: Zack_M_Davis's comment on: Com­ment sec­tion from 05/​19/​2019

  I find it un­pleas­ant that you always bring your hob­by­horse in, but in an “ab­stract” way that doesn’t al­low dis­cussing the ac­tual ob­ject level ques­tion.

  ...

  That’s un­der­stand­able, but I hope it’s also un­der­stand­able that I find it un­pleas­ant that our stan­dard Bayesian philos­o­phy-of-lan­guage some­how got poli­ti­cized (!?), such that my at­tempts to do cor­rect episte­mol­ogy are per­ceived as at­tack­ing peo­ple?!

  I note that this isn’t a de­nial of the ac­cu­sa­tion that you’re bring­ing up a hob­by­horse, dis­guised by ab­strac­tion. It sounds more like a defense of dis­cussing a poli­ti­cal spe­cific by means of ab­strac­tion. I’ve noted in at least some of your posts that I don’t find your ab­strac­tions very com­pel­ling with­out ex­am­ples, and I that I don’t much care for the ex­am­ples I can think of to reify your ab­strac­tions.

  It’s at times like this that I’m happy I’m not part of a “ra­tio­nal­ist com­mu­nity” that in­cludes repet­i­tive in­di­rec­tion of poli­ti­cal fights along with de­nial that that’s what they are. But I wish you’d keep it off less wrong.

  On the next level down, your in­sis­tence that words have con­sis­tent mean­ing and cat­e­gories are real and must be con­sis­tent across us­ages (in­clud­ing both con­text changes and in­ter­nal rea­son­ing vs ex­ter­nal com­mu­ni­ca­tion) seems a blind spot. I don’t know if it’s caused by the ex­am­ples you’re choos­ing (and not shar­ing), or if the re­verse is true.

• Dagon 17 May 2019 8:08 UTC

  2 points

  in reply to: Raemon's comment on: Boo votes, Yay NPS

  Yeah, I com­ment far more than I post, and I have a gen­eral idea that if I get no down­votes, it means I’m op­ti­miz­ing too much for ap­proval rather than ex­plo­ra­tion of non-ob­vi­ous ideas. The vote-delta fea­ture makes it a lit­tle eas­ier to see the down­votes (I’ve dis­abled the hid­ing of nega­tives), but mostly I have to look and figure out if the av­er­age votes-per-voter is par­tic­u­larly low.

• Dagon 17 May 2019 8:03 UTC

  0 points

  on: Why ex­actly is the song ‘Baby Shark’ so catchy?

  Sim­plic­ity, repet­i­tive­ness, and plenty of “doot”.

• Dagon 15 May 2019 21:43 UTC

  2 points

  on: Boo votes, Yay NPS

  I’m not sure it makes sense to put much effort into this kind of gam­ifi­ca­tion. Karma, vot­ing, re­ac­tions, etc. are all cost-free, and there­fore very weak sig­nals.

  It means differ­ent things to differ­ent peo­ple, and it’s so cheap that it’s hard to imag­ine that a change in text or mechanism will rad­i­cally change it’s use (though tweaks may be able to mod­er­ately re­frame things).

  As long as vot­ing has the gen­eral effect of in­creas­ing a sense of in­volve­ment, even for those who don’t post/​com­ment, it’s prob­a­bly worth keep­ing.

• Dagon 9 May 2019 22:24 UTC

  14 points

  in reply to: Benquo's comment on: Towards op­ti­mal play as Villager in a mixed game

  Sorry, I was mak­ing a dis­tinc­tion be­tween a lone in­di­vi­d­ual mak­ing a claim (“I’m the king, listen to me” or “I have so­cial skills, listen to me”) vs enough OTHER peo­ple mak­ing the claim (“he’s the king, listen to him”) to give ev­i­dence that it’s already ac­cepted. The first is use­less, the sec­ond is pow­er­ful enough to ob­vi­ate the first.

• Dagon 9 May 2019 19:10 UTC

  7 points

  in reply to: Benquo's comment on: Towards op­ti­mal play as Villager in a mixed game

  The co­or­di­na­tion case is not di­rectly com­pa­rable to the di­rect claim of au­thor­ity. Get­ting many peo­ple to perform the cer­e­monies and to pub­li­cly pro­claim that one is king is di­rect ev­i­dence of deep in­fluence over at least those peo­ple. Claiming to be king is un­nec­es­sary if there is already such ev­i­dence, and in­effec­tive if there is not.

  Mul­ti­ple peo­ple within a group say­ing it is not a trans­for­ma­tion of the claim, it’s di­rect ev­i­dence for the claim.

• Dagon 9 May 2019 19:00 UTC

  3 points

  in reply to: John_Maxwell_IV's comment on: How To Use Bureaucracies

  Agreed—one is not ex­pected to read all that’s been writ­ten on a topic. How­ever, one should ac­knowl­edge al­ter­nate mod­els when they’re pointed out, and your sum­maries and con­clu­sions will carry a lot more weight if you can ex­plain why you pre­fer those over the other ideas.

• Dagon 9 May 2019 18:06 UTC

  6 points

  on: Open Thread May 2019

  [meta] is vari­able and strong vot­ing hav­ing the de­sired effect? I use it pretty rarely, and of­ten go back and re­duce it to a reg­u­lar up­vote if there are a num­ber of other vot­ers that I don’t want to out­weigh. I’ve no­ticed on my own com­ments, though, that some­times a sin­gle voter will move a score by 5-10 points, which is nuts for a com­ment that only has 4 votes and a to­tal of 12 to start with.

  Can I opt to just see num­ber of up- and down-votes rather than the hard-to-parse weighted to­tals, at least for my own posts and com­ments (the ones I want a sig­nal for)?

• Dagon 8 May 2019 20:20 UTC

  2 points

  in reply to: Benquo's comment on: Towards op­ti­mal play as Villager in a mixed game

  I’m very torn. I fully agree with Rae­mon’s con­cerns, and might even go fur­ther: com­pe­tent peo­ple are rare, and fully goal-al­igned peo­ple are nonex­is­tent. Look­ing at ac­counts from pre­vi­ous times is an ex­is­tence proof of differ­ent equil­ibria, but does not im­ply that they’re available to­day.

  And if you look closer, those pre­vi­ous equil­ibria were miss­ing some fea­tures that we hold dear to­day, such as fairly long pe­ri­ods at the be­gin­ning and end of life where eco­nomic pro­duc­tion isn’t a driv­ing need, some amount of re­spect for peo­ple very differ­ent from our­selves, and a knowl­edge that the cur­rent equil­ibrium isn’t per­ma­nent.

  The part I’m torn on is that I deeply sup­port ex­per­i­ment­ing and think­ing on these top­ics, and I very much hope that my pre­dic­tions are in­cor­rect. This is a case where in­vest­ing men­tal en­ergy on a low-prob­a­bil­ity high-pay­off topic seems jus­tified.

• Dagon 7 May 2019 19:16 UTC

  3 points

  in reply to: Raemon's comment on: Hier­ar­chy and wings

  Espe­cially for non-short posts, I of­ten read and di­gest the post be­fore look­ing at com­ments. Hav­ing a tag or other visi­ble in­di­ca­tor that the mods have con­sid­ered the ques­tion and I shouldn’t be quite as dis­tracted as I seem to be by the po­ten­tial dis­rup­tion of a poli­ti­cal topic would help me.

  I don’t look at front­page much (my book­mark is al­lPosts), so I don’t dis­t­in­guish much be­tween pro­moted and just nor­mal posts. Maybe that’s a prob­lem too—some in­di­ca­tion of en­dorse­ment that shows on the post when some­thing is pro­moted to front­page.

• Dagon 7 May 2019 18:28 UTC

  10 points

  on: Co­or­di­na­tion Sur­veys: why we should sur­vey to or­ga­nize re­spon­si­bil­ities, not just predictions

  Push-pol­ling (a sur­vey that pur­ports to col­lect in­for­ma­tion, but is ac­tu­ally meant to change the re­spon­dents’ be­hav­ior) is very clearly in the dark arts. It prob­a­bly has a place in the world, but it’s not about ra­tio­nal­ity, it’s about ma­nipu­la­tion.

• Dagon 7 May 2019 16:30 UTC

  4 points

  on: Hier­ar­chy and wings

  Meta: this topic is im­por­tant, and gen­eral enough to fit within my views of what LessWrong can han­dle, but it’s get­ting close to the edge, es­pe­cially with the bias to­ward eval­u­at­ing right-wing as “or­ga­nized to ex­pro­pri­ate” left-wing as “in­clu­sive lef­tovers, who tend to ex­pro­pri­ate as a side-effect”.

  I would like there to be a way to clearly la­bel it as “poli­tics-ad­ja­cent” or some­thing to ac­knowl­edge that we don’t ac­tu­ally want to move the win­dow of what’s ap­pro­pri­ate for LW.

• Dagon 7 May 2019 16:16 UTC

  9 points

  on: Towards op­ti­mal play as Villager in a mixed game

  Also, one of the keys to win­ning as a villager is to rec­og­nize that there ARE were­wolves, and they will in­terfere in two ways: they’ll gum up your com­mu­ni­ca­tion and trust, and they will kill you if you are the biggest threat to them.

  That said, I’m not sure the game metaphor ap­plies well to real in­ter­ac­tions. In re­al­ity, ev­ery­one is si­mul­ta­neously a villager and a were­wolf (a value pro­ducer and a com­peti­tor for re­sources/​at­ten­tion), and we shift be­tween the roles pretty fluidly on differ­ent top­ics in differ­ent situ­a­tions. And there are very real differ­ences in un­der­ly­ing per­son­al­ity and ca­pa­bil­ity di­men­sions, so there’s always the un­known dis­tinc­tion be­tween “con­fused villager” and “were­wolf”. Com­bined with the fact that at least some scarcity is not ar­tifi­cial (and even ar­tifi­cial scarcity is a con­straint that can last for many years), I have to ad­mit I’m not su­per hope­ful of a solu­tion.

  And now that that’s said, I look for­ward to hear­ing about the ex­per­i­ments—find­ing ways to co­op­er­ate in the face of scarcity and dis­tinct in­di­vi­d­ual wants is the sin­gle most difficult prob­lem in the fu­ture of in­tel­li­gence and flour­ish­ing that I know of. I’ll gladly par­ti­ci­pate if it’ll help, but I highly recom­mend you start in small groups with more than just in­ter­net con­tact among them. (note: I’m not sure how strongly I recom­mend this—the scal­ing prob­lem is real, and what works for a small group with only a few dozen po­ten­tial coal­i­tions may be near-ir­rele­vant to larger struc­tures with trillions of edges in the graph).

• Dagon 7 May 2019 15:47 UTC

  4 points

  in reply to: Andaro's comment on: Should Effec­tive Altru­ism be at war with North Korea?

  “win­ning” or “los­ing” a war, out­side of to­tal an­nihila­tion, are just steps to­ward the fu­ture vi­sion of galax­ies teem­ing with in­tel­li­gent life. It seems very un­likely, but isn’t im­pos­si­ble, that sim­ply con­ced­ing is ac­tu­ally the best path for­ward for the long view.

• Dagon 7 May 2019 5:29 UTC

  3 points

  in reply to: Benquo's comment on: Blame games

  Hmm, maybe I don’t un­der­stand the lev­els. I’d as­sumed that higher lev­els in­clude lower ones, rather than deny­ing their ex­is­tence. It’s ab­solutely pos­si­ble, in my world, to rec­og­nize that there is a re­al­ity, and to still weigh the so­cial ap­pear­ance against it.

• Dagon 7 May 2019 5:27 UTC

  2 points

  in reply to: Andaro's comment on: Should Effec­tive Altru­ism be at war with North Korea?

  It’s a ques­tion of timeframes—if you ac­tu­ally know your util­ity func­tion and be­lieve it ap­plies to the end of the uni­verse, there’s very lit­tle com­pro­mise available. You’re go­ing to act in what­ever ways benefit the far fu­ture, and any­thing that makes that less likely you will (and must) de­stroy, or make pow­er­less.

  If your util­ity func­tion only looks out a few dozen or a few hun­dred years, it’s not very pow­er­ful, and you prob­a­bly don’t know (or don’t have) an ac­tual ideal of fu­ture util­ity. In this case, you’re likely to seek changes to it, be­cause you don’t ac­tu­ally give up much.

• Dagon 6 May 2019 19:11 UTC

  6 points

  in reply to: Benquo's comment on: Blame games

  I fear you’re as­sum­ing a con­sis­tency of ex­pec­ta­tions that leads to­ward a vastly over­sim­ple model of in­cen­tives. Groups value some kinds of di­ver­sity, and some in­stances of de­viant com­pe­tence (un­usual things that work out), while pun­ish­ing some that on many di­men­sions seem very similar to the things they re­ward.

  De­tails mat­ter, and path-de­pen­den­cies abound (where you can’t get from here to there with­out a vi­able in­ter­me­di­ate which may not be op­ti­mal for any­thing).

  I want to be uniquely de­sir­able to my mate and to my em­ployer, to avoid com­pet­ing on a level field with oth­ers, and to avoid the ad­mis­sion that I’m just one of 7.5 billion liv­ing hu­mans, and so can’t ac­tu­ally be all that spe­cial. I si­mul­ta­neously want to sig­nal that I’m pre­dictably pro­duc­tive on stan­dard di­men­sions, and a pos­i­tive out­lier on at least one. Counter-sig­nal­ing is an im­por­tant part of this. Show­ing that I de­vi­ate from norms on some (harm­less) di­men­sions is an in­di­ca­tor that I’m com­pe­tent enough in more im­por­tant di­men­sions that I don’t have to care about con­for­mity.

  You have to dance in sync for some parts of some songs, and show your unique strengths (which in­cludes show­ing weak­nesses and neu­tral dis­tinc­tions as part of counter-sig­nal­ing) dur­ing other parts of the dance. even line dances and highly-pre­scrip­tive vic­to­rian dances have seg­ments where some amount of freestyle perfor­mance is benefi­cial.

• Dagon 6 May 2019 18:40 UTC

  2 points

  on: Should Effec­tive Altru­ism be at war with North Korea?

  “stronger AI offers weaker AI part of its util­ity func­tion in ex­change for con­ced­ing in­stead of fight­ing” is the ob­vi­ous way for AGIs to re­solve con­flicts, in­so­far as trust can be es­tab­lished. (This method of re­solv­ing dis­putes is also prob­a­bly part of why an­i­mals have sex.)

  Wow, this seems like a huge leap. It seems like an in­ter­est­ing thought ex­per­i­ment (es­pe­cially if the weaker ALSO changes util­ity func­tion, so the AIs are now perfectly al­igned). But it kind of ig­nores what is mak­ing the de­ci­sion.
  If a util­ity func­tion says it’s best to change the util­ity func­tion, it was re­ally a meta-func­tion all along.

  Re­mem­ber that in re­al­ity, all games are re­peated games. How many com­pro­mises will you have to make over the com­ing eons? If you’re will­ing to change your util­ity func­tion for the sake of con­flict avoidance (or re­source gains), doesn’t that mean it’s not re­ally your util­ity func­tion?

  Hav­ing a util­ity func­tion that in­cludes avoid­ing con­flict is definitely in line with co­op­er­at­ing with very differ­ent be­ings, at least un­til you can cheaply erad­i­cate/​ab­sorb them. But no util­ity func­tion can be will­ing to change it­self vol­un­tar­ily.

  It also seems like there are less risky and cheaper op­tions, like iso­la­tion­ism and de­struc­tion. There’s plenty of fu­ture left for re­cov­ery and growth af­ter near (but not ac­tual) ex­tinc­tion, but once you give up your goals, there’s no go­ing back.

  Note that this en­tire dis­cus­sion is pred­i­cated on there ac­tu­ally be­ing some con­sis­tent the­ory or func­tion caus­ing this un­com­fortable situ­a­tion. It may well be that mon­key brains are in con­trol of far more power than they are evolved to think about, and we have to ac­cept that dom­i­nance dis­plays are go­ing to hap­pen, and just try to sur­vive them.

• Dagon 6 May 2019 18:02 UTC

  3 points

  on: Blame games

  All equil­ibria are a bal­ance of at least two forces. You’re lay­ing out a case for the forces push­ing to­ward pri­vacy/​ly­ing/​hypocrisy side of so­cial in­ter­ac­tions. What are the forces op­pos­ing them? Why is any­one try­ing to tell the truth in the first place?

  Are there any ad­van­tages to play­ing the level 1 game, or does this all boil down to “hu­man im­plies poli­ti­cal, get good at it”?