Benito

• Benito 16 Apr 2019 9:15 UTC

  10 points

  on: Liter­a­ture Re­view: Distributed Teams

  This is awe­some, thanks.

  In case it’s of in­ter­est to any­one, I re­cently wrote down some short, ex­plicit mod­els of the costs of re­mote teams (I did not try to write the benefits). Here’s what I wrote:

  • Sub­stan­tially in­creases ac­ti­va­tion costs of col­lab­o­ra­tion, lead­ing to highly split fo­cus of staff

  • Sub­stan­tially in­creases costs of cre­at­ing com­mon knowl­edge (es­pe­cially in poli­ti­cal situ­a­tions)

  • Sub­stan­tially in­creases bar­ri­ers to build­ing trust (in-per­son in­ter­ac­tion is key for in­ter­per­sonal trust)

  • Sub­stan­tially de­creases com­mu­ni­ca­tion band­width—both rate and qual­ity of feed­back—in­creas­ing the cost of sub­tle, fine-grained and spe­cific pos­i­tive feed­back harder, and mak­ing strong nega­tive feed­back on bad de­ci­sions much eas­ier, lead­ing to risk-aver­sion.

  • Sub­stan­tially in­creases cost of trans­mit­ting po­ten­tially em­bar­rass­ing in­for­ma­tion, and in­cen­tivises cov­er­ing up of low pro­duc­tivity, as it’s very hard for a man­ager to see the day-to-day and week-to-week out­put.

• Benito 10 Apr 2019 23:29 UTC

  6 points

  on: Ex­cerpts from a larger dis­cus­sion about simulacra

  Let me check I’m fol­low­ing with some sim­ple claims:

  • If it’s com­mon knowl­edge that we’re in world 3, then we’re in world 4.

  • If it’s com­mon knowl­edge that we’re in world 2, then we’re in world 4.

  • The key value to be­ing in world n+1 is that you can out­play all the peo­ple in world n.

  • To move back from world 2 into world 1, one can pun­ish in­ac­cu­rate job titles

  • To move back from world 3 into world 2, one can pun­ish not treat­ing work­ers ac­cord­ing to their titles

  • You can’t move back from world 4 to world 3.

• Benito 8 Apr 2019 14:00 UTC

  2 points

  on: Long Term Fu­ture Fund: April 2019 grant de­ci­sions [link]

  Dupe.

• Benito 3 Apr 2019 22:32 UTC

  11 points

  on: What are the ad­van­tages and dis­ad­van­tages of know­ing your own IQ?

  Peo­ple seem to treat it in a fatal­is­tic way, like they’ve been told what their score will be at the end of the game, as op­posed to one of their base stats (like find­ing out how tall you are). I tested my­self on the big 5 lately and find­ing out I have a fairly ex­treme baseline on things like neu­roti­cism and in­tel­lect been sur­pris­ingly valuable for un­der­stand­ing my­self.

  I un­der­stand there are also sub­cat­e­gories of IQ, and I am in­ter­ested to know if there’s an IQ test I can take which gives me info on a va­ri­ety of the more ro­bust com­po­nents of IQ (what­ever they are, I don’t ac­tu­ally know). I could imag­ine this giv­ing me ad­vice of the type “In gen­eral try strongly to use ver­bal rea­son­ing over spa­cial rea­son­ing, and if you’re in a situ­a­tion where spa­cial rea­son­ing is nec­es­sary, make a con­scious plan to put in more de­liber­ate prac­tice than seems nec­es­sary for the me­dian similarly smart per­son around you who is learn­ing the same skill.” If I ex­pected to get 3+ big recom­men­da­tions like that I think I’d be quite ex­cited to pay for a test.

  I think that hav­ing IQ tie more closer to your de­ci­sions might help peo­ple un­der­stand it than if it’s just an ab­stract im­mutable num­ber that says you’re worse than these other peo­ple, and hav­ing it be mul­ti­fac­to­rial could be a way to help there?

  But un­less I can ac­tu­ally tie it to some de­ci­sions, I do ex­pect find­ing out my IQ to make me de­pressed on net. Per­haps I can just use it to figure out whether or not it’s on the table for me to do math at MIRI, though my sense is that philo­soph­i­cal so­phis­ti­ca­tion is much more the bot­tle­neck there.

• Benito 3 Apr 2019 20:23 UTC

  3 points

  on: LW Up­date 2019-04-02 – Front­page Rework

  I am es­pe­cially in­ter­ested to hear about any strong pos­i­tive/​nega­tives from mo­bile and tablet users.

• Benito 29 Mar 2019 22:28 UTC

  6 points

  in reply to: Raemon's comment on: Dependability

  Yeah, I think CFAR has been heavy tailed, and I would pre­dict that there are some in­di­vi­d­u­als for whom it has coun­ter­fac­tu­ally caused them to solve big prob­lems like this.

• Benito 29 Mar 2019 18:36 UTC

  7 points

  in reply to: ChristianKl's comment on: Please use real names, es­pe­cially for Align­ment Fo­rum?

  Peo­ple do as­sign a fair amount of sta­tus based on at­trac­tive­ness of faces, and I think it’s good on the mar­gin to not in­tro­duce that class of bias to the dis­cus­sion. My cur­rent guess is that the costs aren’t com­men­su­rate with the benefits of faster recog­ni­tion.

• Benito 28 Mar 2019 15:46 UTC

  5 points

  on: Hu­mans Who Are Not Con­cen­trat­ing Are Not Gen­eral Intelligences

  I feel like I learned some­thing very im­por­tant about my mind—you’re right, if I skim these low-level-pat­tern-matched para­graphs, they read as ba­si­cally fine to me. Has plau­si­bly quite im­por­tant im­pli­ca­tions for AI too. So I’ve cu­rated this post.

• Benito 23 Mar 2019 15:42 UTC

  6 points

  in reply to: Zvi's comment on: An­nounce­ment: AI al­ign­ment prize round 4 winners

  Re­minder to do this.

  (I will stop re­mind­ing you if you ask, but un­til then I am a fan of helping pub­lic com­mit­ment get acted on.)

• Benito 21 Mar 2019 23:32 UTC

  5 points

  in reply to: habryka's comment on: Has “poli­tics is the mind-kil­ler” been a mind-kil­ler?

  Yes, the new read­ing is “Poli­tics isn’t a good place to prac­tice ra­tio­nal­ity un­less all the dis­cus­sants are already ra­tio­nal”. Not that you shouldn’t en­gage in dis­cus­sion of poli­tics, just that you shouldn’t go to train ra­tio­nal­ity there (when not already well prac­ticed in other ar­eas).

• Benito 18 Mar 2019 15:08 UTC

  18 points

  on: “Other peo­ple are wrong” vs “I am right”

  I already told Buck that I loved this post. For this cu­ra­tion no­tice, let me be spe­cific about why.

  • Posts from peo­ple who think care­fully and se­ri­ously about difficult ques­tions writ­ing about some of the big ways they changed their mind over time are rare and valuable (other ex­am­ples: Holden, Eliezer, Kah­ne­man).

  • OP is un­usu­ally trans­par­ent, in a way that leads me to feel I can ac­tu­ally up­date on the data rather than hold­ing it in an in­ter­nal sand­box. In feel it has not been as ad­ver­sar­i­ally se­lected as most other writ­ings by some­one about them­selves, mak­ing it ex­tremely valuable data. (Where data is nor­mally cov­ered up, even small amounts of true data are of­ten very sur­pris­ing.)

  • I find the spe­cific up­date quite use­ful, in­clud­ing all of the ex­am­ples. It fits to­gether with Eliezer’s claim (at the end of sec­tion 5 here) that you can figure out which ex­perts are right/​wrong far more of­ten than you can come up with the cor­rect the­ory your­self.

• Benito 18 Mar 2019 14:15 UTC

  4 points

  on: Has “poli­tics is the mind-kil­ler” been a mind-kil­ler?

  Note: even with you mak­ing a point of it, it took me two reads to un­der­stand why my ini­tial read of “un­less all the dis­cus­sants are already ra­tio­nal” was wrong.

• Benito 14 Mar 2019 17:24 UTC

  12 points

  in reply to: Jan_Kulveit's comment on: Un­der­stand­ing in­for­ma­tion cascades

  Agreed. I re­al­ise the OP could be mis­read; I’ve up­dated the first para­graph with an ex­tra sen­tence men­tion­ing that sum­maris­ing and trans­lat­ing ex­ist­ing work/​liter­a­ture in re­lated do­mains is also re­ally helpful.

• Benito 14 Mar 2019 16:16 UTC

  7 points

  in reply to: Jan_Kulveit's comment on: Un­der­stand­ing in­for­ma­tion cascades

  Thanks for the poin­t­ers to net­work sci­ence Jan, I don’t know this liter­a­ture, and if it’s use­ful here then I’m glad you un­der­stand it well enough to guide us (and oth­ers) to key parts of it. I don’t see yet how to ap­ply it to think­ing quan­ti­ta­tively about sci­en­tific and fore­cast­ing com­mu­ni­ties.

  If you (or an­other LWer) thinks that the the­ory around uni­ver­sal­ity classes is ap­pli­ca­ble in think­ing about how to en­sure good info prop­a­ga­tion in e.g. a sci­en­tific com­mu­nity, and you’re right, then I (and Ja­cob and likely many oth­ers) would love to read a sum­mary, posted here as an an­swer. Might you ex­plain how un­der­stand­ing the linked pa­per on uni­ver­sal­ity classes has helped you think about info prop­a­ga­tion in fore­cast­ing com­mu­ni­ties /​ re­lated com­mu­ni­ties? Con­crete heuris­tics would be es­pe­cially interesting

  (Note that Ja­cob and I have not taken a math course in topol­ogy or graph the­ory and won’t be able to read an­swers that as­sume such, though we’ve both stud­ied for­mal fields of study and could likely pick it up quickly if it seemed prac­ti­cally use­ful.)

  In gen­eral we’re not look­ing for *novel* con­tri­bu­tions. To give an ex­treme ex­am­ple, if one per­son trans­lates an ex­ist­ing the­o­ret­i­cal liter­a­ture into a fully fleshed out the­ory of info-cas­cades for sci­en­tific and fore­cast­ing com­mu­ni­ties, we’ll give them the en­tire prize pot.

[Question] For­mal­is­ing con­tin­u­ous info cas­cades? [Info-cas­cade se­ries]

[Question] How large is the harm from info-cas­cades? [Info-cas­cade se­ries]

[Question] How can we re­spond to info-cas­cades? [Info-cas­cade se­ries]

[Question] Distri­bu­tion of info-cas­cades across fields? [Info-cas­cade se­ries]

[Question] Un­der­stand­ing in­for­ma­tion cascades

• Benito 13 Mar 2019 9:29 UTC

  10 points

  AF

  on: Thoughts on Hu­man Models

  This is a very care­fully rea­soned and de­tailed post, which lays out a clear frame­work for think­ing about ap­proaches to al­ign­ment, and I’m es­pe­cially ex­cited be­cause it points to one quad­rant—en­g­ineer­ing-fo­cused re­search with­out hu­man mod­els—as highly ne­glected. For these three rea­sons I’ve cu­rated the post.