Ruby

• Ruby 26 May 2020 0:18 UTC

  2 points

  in reply to: wizzwizz4’s comment on: Tag In­dex (Beta)

  Oops, that’s a mis­take. Fixed now. Thanks.

• Ruby 12 May 2020 0:44 UTC

  23 points

  on: Liter­a­ture Re­view For Aca­demic Out­siders: What, How, and Why

  Cu­rated (with mul­ti­ple en­dorse­ments from the mod team). As noted in my pre­vi­ous com­ment, this post in­cludes lots of links and refer­ences to fur­ther re­sources, but it also mo­ti­vates the need for lit re­views well. It’s not just a “how to guide”, but also a “why guide” as well. It’s a timely post too.
  
  Go back a few years, and luke­prog was the cham­pion/​sym­bol of schol­ar­ship on LessWrong. Un­for­tu­nately for us, he’s not able to con­tribute to LessWrong as much any­more; which it makes great that oth­ers are tak­ing up the ban­ner and re­mind­ing us of the need to build on ex­ist­ing knowl­edge (and helping peo­ple know how to do so).
  
  I say this post is timely, that’s be­cause mak­ing LessWrong more schol­arly con­tinues to be a ma­jor fo­cus of my work on the LessWrong team. Schol­ar­ship/​Lit Re­views are ac­tu­ally a ma­jor goal of the new Tag­ging/​Wiki sys­tem, whose larger goal still is in­creas­ing LessWrong’s in­tel­lec­tual out­put. The hope is to make it much eas­ier for writ­ers on LessWrong to dis­cover and build upon LessWrong’s decade of pre­vi­ous work. “Shoulder of Gi­ants”, etc.

  Ob­vi­ously, the over­whelming su­per­ma­jor­ity of the world’s knowl­edge isn’t in LessWrong’s posts (though the very best in­sights might be), and our thinkers ab­solutely need to the skills (and virtue) to mine the troves of knowl­edge out­side our shores. Hence the value in this post.
  
  [At the same time, I do think we shouldn’t let a re­quire­ment of lit re­view be­come too high a bar­rier to con­tribut­ing on LessWrong. There’s a lot of value in think­ing through things for your­self fresh, and some­times just get­ting ran­dom un­in­formed thoughts pub­lished stim­u­lates dis­cus­sion and pro­vides mo­ti­va­tion to then go for a thor­ough sur­vey of the liter­a­ture.]

  All in all, ku­dos.
  
  (And thanks for the recom­men­da­tion of In­tel­lec­tual Foun­da­tion of In­for­ma­tion Or­ga­ni­za­tion, that was a good one.)

• Ruby 11 May 2020 2:38 UTC

  4 points

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

  Wel­come!
  
  The dic­tio­nary defi­ni­tion of “per­suade” misses some of the con­no­ta­tions. Per­suad­ing some­one of­ten means “get them to agree with you” and not “jointly ar­rive at what’s true, which in­cludes the pos­si­bil­ity that oth­ers can point out your mis­takes and you change your mind.” Ex­plain­ing usu­ally means more some­thing like “ex­plain your rea­son­ing and facts, which might lead some­one to agree with if they think your rea­son­ing is good.”
  
  The key differ­ence might be some­thing like “per­suade” is writ­ten to get the reader to ac­cept what is writ­ten re­gard­less of whether it’s true, while “ex­plain” wants you to ac­cept the con­clu­sion only if it’s true. It’s the idea sym­met­ric/​asym­met­ric weapons in this post.
  
  Sorry if that’s still a bit un­clear, I hope it helps.

• Ruby 10 May 2020 17:29 UTC

  11 points

  on: Liter­a­ture Re­view For Aca­demic Out­siders: What, How, and Why

  Many thanks for writ­ing this. Great over­all and I re­ally like the large num­ber of links and refer­ences to other re­sources too (and would have said that even if it wasn’t ac­tu­ally the whole topic :P). I’m so pleased when LW gets an­other thing about how to study/​re­search. I gave this a strong tag rele­vance vote on the Schol­ar­ship & Learn­ing wik­i­tag.

• Ruby 7 May 2020 5:29 UTC

  7 points

  in reply to: Spiracular’s comment on: Cop­ing and Cultures

  I be­lieve that mil­i­tary stuff, in­clud­ing and maybe es­pe­cially cul­ture, is a long-term in­ter­est of LW user, Lion­hearted. You could mes­sage him, also look at his writ­ing on men­tal tough­ness within the Strate­gic Re­view se­ries.

• Ruby 5 May 2020 1:06 UTC

  5 points

  in reply to: Raemon’s comment on: [Site Meta] Fea­ture Up­date: More Tags! (ex­per­i­men­tal)

  Ra­tion­al­ist cul­ture and life ex­ten­sion might make sense. We have a Cry­on­ics tag already. If we can round up a few posts on ei­ther of those top­ics, would cre­ate these.

• Ruby 30 Apr 2020 16:07 UTC

  2 points

  in reply to: Yoav Ravid’s comment on: [Site Meta] Quick Guide to Tagging

  To re­move a tag, just down­vote (it might look like it’s gone to −2, which is fine, upon re­fresh it will be gone).
  
  Yeah, some of those definitely seem like good tag. I’ve had the idea for Co­or­di­na­tion/​Co­op­er­a­tion, Group Ra­tion­al­ity, and Com­mu­ni­ca­tion.
  
  The oth­ers I think we’d want to en­sure there isn’t too much over­lap with ex­ist­ing things. There’s a pro­gram­ming tag, does that do the thing for Soft­ware? And then cu­ri­ous about what you see­ing go­ing in tools vs the ex­ist­ing tech­niques (which might also cover “soft skills”)
  
  It’s good to see all these sug­ges­tions though. Even if we don’t make a tag be­cause of an ex­ist­ing one, soon we might send up “redi­rects” for terms to­wards things that are al­most the same, or at least the clos­est match.

• Ruby 26 Apr 2020 4:56 UTC

  4 points

  in reply to: ESRogs’s comment on: [Site Meta] Fea­ture Up­date: More Tags! (ex­per­i­men­tal)

  It’s rea­son­able to men­tion “there’s this com­ment which is rele­vant to this topic...”

• Ruby 26 Apr 2020 2:06 UTC

  4 points

  in reply to: Scott Garrabrant’s comment on: [Site Meta] Fea­ture Up­date: More Tags! (ex­per­i­men­tal)

  Th­ese are re­ally good.

  Embed­ded Agency is a clear win.

  Mechanism De­sign/​Align­ing In­cen­tives seems good too. Agree there are choices about the name, and I guess scope too. Do you mean it to be ma­te­rial about how to al­ign in­cen­tives but ex­clude re­lated stuff of ex­am­ples where in­cen­tives failed to be al­igned. Would Boe­ing 737 MAX MCAS as an agent cor­rigi­bil­ity failure be part of it?
  
  ”Re­source Bounded Epistemics” sounds like a cool cat­e­gory. So does “In­ter­dis­ci­plinary Analo­gies”, or should it be “In­ter­dis­ci­plinary Ap­pli­ca­tions”?
  
  Any­how, these are great. More are wel­come.
  
  Fake Frame­works, yeah, hmm. We might con­sider “only au­thors can ap­ply these tags”, I’m not sure. Those might make sense for gen­eral “epistemic state” tags.
  

• Ruby 26 Apr 2020 1:52 UTC

  2 points

  in reply to: ESRogs’s comment on: [Site Meta] Fea­ture Up­date: More Tags! (ex­per­i­men­tal)

  Th­ese are great! I’ll make these soon. Those posts definitely jus­tify do­ing so in my mind. Re: Wei Dai’s com­ment, I think it’s rea­son­able to men­tion in the tag de­scrip­tion text (and those will soon be ev­ery­one-ed­itable wiki en­tries and should in­clude ex­tra info rele­vant to the tag/​wiki/​con­cept, in­clud­ing “no­table com­ments”).

• Ruby 23 Apr 2020 21:54 UTC

  3 points

  in reply to: habryka’s comment on: [Site Meta] Fea­ture Up­date: More Tags! (ex­per­i­men­tal)

  Yeah, I agree with most of that.
  
  I do think there will be an ap­pre­cia­ble num­ber of tags (even if they’re a minor­ity) that are strictly sub­sets of, say, AI al­ign­ment. Like ev­ery­thing un­der Value Learn­ing or Embed­ded Agents, etc, and maybe it’s worth it to have that au­to­mat­i­cally up­date.
  
  I do feel tag de­scrip­tions link­ing to other tags is ex­tremely im­por­tant for the sys­tem to work and will help a lot here.

• Ruby 23 Apr 2020 21:41 UTC

  2 points

  in reply to: Thomas Kwa’s comment on: [Site Meta] Fea­ture Up­date: More Tags! (ex­per­i­men­tal)

  In­ter­est­ing. I agree we want more spe­cific tags too for that post. Though “Prob­lem-Solv­ing Tac­tics”, ac­tu­ally feels pretty broad too, though a good defi­ni­tion/​de­scrip­tion might help give it shape. I’ll think about it, not sure if you had one in mind.
  
  Another thing that helps is hav­ing other posts in mind too for the tag.

• Ruby 22 Apr 2020 22:59 UTC

  2 points

  in reply to: ChristianKl’s comment on: [Site Meta] Fea­ture Up­date: More Tags! (ex­per­i­men­tal)

  I think I un­der­stand the mo­ti­va­tion be­hind that. They’re too easy to cre­ate and end up ap­ply­ing to too many differ­ent things? Does that seem right?
  
  A challenge which stark in my mind is how to avoid cre­at­ing too many heav­ily over­lap­ping tags, which seems easy to do with higher-level tags.

• Ruby 22 Apr 2020 19:09 UTC

  3 points

  in reply to: leggi’s comment on: [Site Meta] Fea­ture Up­date: More Tags! (ex­per­i­men­tal)

  To un­tag a post, just down­vote its tag rele­vance. (Either in the hover-over or on the tag page).
  
  Yeah, agree with need a bet­ter solu­tion for show­ing cur­rently available tags. In the mean­time, you can look at www.less­wrong.com/​​tags or www.less­wrong.com/​​tags/​​all
  
  A heuris­tic the team has dis­cussed is that tags should have 3 good post by at least two differ­ent au­thors. I do want some kind of wellbe­ing cat­e­gory, and a sep­a­rate health one make sense too. Anatomy, if it isn’t a topic dis­cussed by oth­ers, may or may not make sense. I’m not sure. If it’s to help peo­ple find your other writ­ing (the main goal of tag­ging), you could cre­ate a se­quence or two to link them.

• Ruby 22 Apr 2020 19:05 UTC

  3 points

  in reply to: ChristianKl’s comment on: [Site Meta] Fea­ture Up­date: More Tags! (ex­per­i­men­tal)

  I’m in­clined to treat COVID-19 posts as ex­cep­tion and not tag them with any­thing ex­cept Coron­avirus, un­less they’re also ap­pli­ca­ble more broadly and time­lessly.

• Ruby 22 Apr 2020 19:05 UTC

  6 points

  in reply to: namespace’s comment on: [Site Meta] Fea­ture Up­date: More Tags! (ex­per­i­men­tal)

  Nice, will definitely look at these.

[Site Meta] Fea­ture Up­date: More Tags! (Ex­per­i­men­tal)

• Ruby 11 Apr 2020 1:06 UTC

  6 points

  on: What’s the up­per bound of how long COVID is con­ta­gious?

  Th­ese pa­pers on viral load prob­a­bly help in­form the an­swer. It was flagged to me that Ct might not have straight­for­ward in­ter­pre­ta­tion, but I haven’t looked into it. So post­ing these as re­sources.
  
  https://​​www.the­lancet.com/​​jour­nals/​​lan­inf/​​ar­ti­cle/​​PIIS1473-3099(20)30113-4/​​ful­l­text?fb­clid=IwAR3crOZx­hVP1eVPMcO_wu­jJBxHFAjp2fj4_jNj30ld_nVcKTqtcT1IjXozI
  
  https://​​www.nejm.org/​​doi/​​full/​​10.1056/​​NEJMc2001737

• Ruby 10 Apr 2020 23:47 UTC

  2 points

  on: What is the safe in-per­son dis­tance for COVID-19?

  This came up on my Face­book feed. I have only glanced at a briefly, but is prob­a­bly of in­ter­est here:
  
  Bel­gian-Dutch Study: Why in times of COVID-19 you should not walk/​run/​bike close to each other.

• Ruby 7 Apr 2020 19:24 UTC

  2 points

  on: Core Tag Ex­am­ples [tem­po­rary]

  AI Align­ment Tag Examples

  1. Embed­ded Agency (full-text ver­sion)

  2. Good­hart Taxonomoy

  3. An Un­trol­lable Math­e­mat­i­cian Illustrated

  4. AlphaGo Zero and the Foom Debate

  5. The un­ex­pected difficulty of com­par­ing AlphaS­tar to humans

  6. Out­perform­ing the hu­man Atari benchmark

  7. How does OpenAI’s lan­guage model af­fect our AI timeline es­ti­mates?

  8. Jeff Hawk­ins on neu­ro­mor­phic AGI within 20 years

  9. What failure looks like

  10. Soft take­off can still lead to de­ci­sive strat­egy advantage

  11. My cur­rent frame­work for think­ing about AGI timelines

  Ex­am­ples of AI Align­ment Com­mu­nity Posts (tagged both AI Align­ment and Com­mu­nity)

  1. 2019 AI Align­ment Liter­a­ture Re­view and Char­ity Comparison

  2. What I’ll be do­ing at MIRI

  3. Offer of col­lab­o­ra­tion and/​or men­tor­ship

  4. Where are peo­ple think­ing and talk­ing about global co­or­di­na­tion for AI safety?