AnnaSalamon

• AnnaSalamon 19 Dec 2017 20:59 UTC

  27 points

  on: CFAR 2017 Retrospective

  I con­tinue to think CFAR is among the best places to donate re: turn­ing money into ex­is­ten­tial risk re­duc­tion (in­clud­ing this year—ba­si­cally be­cause our good done seems al­most lin­ear in the num­ber of free-to-par­ti­ci­pants pro­grams we can run (be­cause those can tar­get high-im­pact AI stuff), and be­cause the num­ber of free-to-par­ti­ci­pants pro­grams we can run is more or less lin­ear in dona­tions within the range in which dona­tions might plau­si­bly take us). If any­one wants my take on how this works, or on our last year or our up­com­ing year or any­thing like that, I’d be glad to talk: anna at ra­tio­nal­ity dot org.

• AnnaSalamon 16 Dec 2017 20:59 UTC

  55 points

  in reply to: RyanCarey's comment on: In the pres­ence of dis­in­for­ma­tion, col­lec­tive episte­mol­ogy re­quires lo­cal modeling

  RyanCarey writes:

  If you are some­one of me­dian in­tel­li­gence who just want to carry out a usual trade like mak­ing shoes or some­thing, you can largely get by with re­cieved wis­dom.

  AFAICT, this only holds if you’re in a sta­ble so­ciopoli­ti­cal/​eco­nomic con­text—and, more speci­fi­cally still, the kind of sta­ble so­ciopoli­ti­cal en­vi­ron­ment that pro­vides rel­a­tively be­nign in­for­ma­tion-sources. Ex­am­ples of folks who didn’t fall into this cat­e­gory: (a) folks liv­ing in east­ern Europe in the late 1930′s (es­pe­cially if Jewish, but even if not; re­gard­less of how tra­di­tional their trade was); (b) folks liv­ing in the Soviet Union (re­quired nav­i­gat­ing a non-ex­plicit layer of re­cieved-from-un­der­ground knowl­edge); (c) folks liter­ally mak­ing shoes dur­ing time-pe­ri­ods in which shoe-mak­ing was dis­rupted by the in­dus­trial rev­olu­tion. It is to my mind an open ques­tion whether any sig­nifi­cant por­tion of the US/​Europe/​etc. will fall into the “can get by largely with re­ceived wis­dom” refer­ence class across the next 10 years. (They might. I just ac­tu­ally can’t tell.)

“Flinch­ing away from truth” is of­ten about *pro­tect­ing* the epistemology

Fur­ther dis­cus­sion of CFAR’s fo­cus on AI safety, and the good things folks wanted from “cause neu­tral­ity”

• AnnaSalamon 10 Dec 2016 21:36 UTC

  0 points

  in reply to: JonahSinick's comment on: CFAR’s new mis­sion state­ment (on our web­site)

  This is fair; I had in mind ba­sic high school /​ New­to­nian physics of ev­ery­day ob­jects. (E.g., “If I drop this penny off this build­ing, how long will it take to hit the ground?“, or, more mess­ily, “If I drive twice as fast, what im­pact would that have on the ki­netic en­ergy with which I would crash into a tree /​ what im­pact would that have on how badly de­formed my car and I would be if I crash into a tree?“).

• AnnaSalamon 10 Dec 2016 18:48 UTC

  5 points

  in reply to: Ben Pace's comment on: Fur­ther dis­cus­sion of CFAR’s fo­cus on AI safety, and the good things folks wanted from “cause neu­tral­ity”

  We would in­deed love to help those peo­ple train.

• AnnaSalamon 10 Dec 2016 18:44 UTC

  9 points

  in reply to: gjm's comment on: Fur­ther dis­cus­sion of CFAR’s fo­cus on AI safety, and the good things folks wanted from “cause neu­tral­ity”

  Yes. Or will se­ri­ously at­tempt this, at least. It seems re­quired for co­op­er­a­tion and good epistemic hy­giene.

• AnnaSalamon 10 Dec 2016 18:41 UTC

  1 point

  in reply to: jsteinhardt's comment on: CFAR’s new mis­sion state­ment (on our web­site)

  Thanks; good point; will add links.

• AnnaSalamon 10 Dec 2016 18:40 UTC

  1 point

  on: CFAR’s new mis­sion state­ment (on our web­site)

  In case there are folks fol­low­ing Dis­cus­sion but not Main: this mis­sion state­ment was re­leased along with:

  • CFAR’s new fo­cus, and AI Safety

  • Fur­ther dis­cus­sion of CFAR’s fo­cus on AI safety, and the good things folks wanted from “cause neu­tral­ity”

• AnnaSalamon 10 Dec 2016 18:32 UTC

  1 point

  in reply to: owencb's comment on: CFAR’s new fo­cus, and AI Safety

  Oh, sorry, the two new docs are posted and were in the new ETA:

  http://​less­wrong.com/​lw/​o9h/​fur­ther_dis­cus­sion_of_cfars_fo­cus_on_ai_safety/​ and http://​less­wrong.com/​r/​dis­cus­sion/​lw/​o9j/​cfars_new_mis­sion_state­ment_on_our_web­site/​

• AnnaSalamon 10 Dec 2016 8:31 UTC

  0 points

  on: [Er­ror com­mu­ni­cat­ing with LW2 server]

  Apolo­gies; the link is bro­ken and I’m not sure how to edit or delete it; real link is: http://​ra­tio­nal­ity.org/​about/​mission

• AnnaSalamon 10 Dec 2016 8:02 UTC

  3 points

  in reply to: owencb's comment on: CFAR’s new fo­cus, and AI Safety

  Thanks for the thoughts; I ap­pre­ci­ate it.

  I agree with you that fram­ing is im­por­tant; I just deleted the old ETA. (For any­one in­ter­ested, it used to read:

  ETA: Hav­ing talked just now to peo­ple at our open house, I would like to clar­ify: Even though our aim is ex­plic­itly AI Safety...
  CFAR does still need an art of ra­tio­nal­ity, and a com­mu­nity of ra­tio­nal­ity geeks that sup­port that. We will still be in­vest­ing at least some in that com­mu­nity. We will also still be run­ning some “ex­plore” work­shops of differ­ent sorts aiming at patch­ing gaps in the art (fund­ing per­mit­ting), not all of which will be de­liber­ately and ex­plic­itly backchained form AI Safety (al­though some will). Play is gen­er­a­tive of a full ra­tio­nal­ity art. (In ad­di­tion to some­times tar­get­ing things more nar­rowly at par­tic­u­lar high-im­pact groups, and oth­er­wise more di­rectly backchain­ing.) (More in sub­se­quent posts.)

  I’m cu­ri­ous where our two new docs leave you; I think they make clearer that we will still be do­ing some ra­tio­nal­ity qua ra­tio­nal­ity.

  Will com­ment later re: sep­a­rate or­ga­ni­za­tions; I agree this is an in­ter­est­ing idea; my guess is that there isn’t enough money and staff fire­power to run a good stan­dalone ra­tio­nal­ity or­ga­ni­za­tion in CFAR’s stead, and also that CFAR re­tains quite an in­ter­est in a stan­dalone ra­tio­nal­ity com­mu­nity and should there­fore sup­port it… but I’m definitely in­ter­ested in thoughts on this.

  Ju­lia will be launch­ing a small spinoff or­ga­ni­za­tion called Con­ver­gence, fa­cil­i­tat­ing dou­ble crux con­ver­sa­tions be­tween EAs and EA-ad­ja­cent peo­ple in, e.g., tech and academia. It’ll be un­der the aus­pices of CFAR for now but will not have opinions on AI. I’m not sure if that hits any of what you’re af­ter.

• AnnaSalamon 8 Dec 2016 9:23 UTC

  1 point

  in reply to: calebwithers's comment on: CFAR’s new fo­cus, and AI Safety

  I think we’ll have it all posted by Dec 18 or so, if you want to wait and see. My per­sonal im­pres­sion is that MIRI and CFAR are both very good buys this year and that best would be for each to re­ceive some dona­tion (col­lec­tively, not from each per­son); I ex­pect the care for MIRI to be some­what more straight-for­ward, though.

  I’d be happy to Skype/​email with you or any­one re: the likely effects of donat­ing to CFAR, es­pe­cially af­ter we get our posts up.

• AnnaSalamon 4 Dec 2016 20:33 UTC

  1 point

  in reply to: ChristianKl's comment on: CFAR’s new fo­cus, and AI Safety

  No, it was just a pun. I be­lieve try­ing to im­prove pre­dic­tive ac­cu­racy is bet­ter than try­ing to pro­mote view X (for ba­si­cally any value of “X”); which I was hop­ing the pun of “Brier Boost­ing” off “Sig­nal Boost­ing” would point to; but not Bri­ers Score as such.

• AnnaSalamon 4 Dec 2016 20:31 UTC

  6 points

  in reply to: oge's comment on: CFAR’s new fo­cus, and AI Safety

  Com­ing up. Work­ing on a blog post about it; will prob­a­bly have it up in ~4 days.

• AnnaSalamon 3 Dec 2016 20:53 UTC

  0 points

  in reply to: Vaniver's comment on: CFAR’s new fo­cus, and AI Safety

  Thanks!

CFAR’s new fo­cus, and AI Safety

• AnnaSalamon 3 Dec 2016 6:44 UTC

  6 points

  in reply to: orthonormal's comment on: CFAR’s new fo­cus, and AI Safety

  We’ll be con­tin­u­ing the work­shops, at least for now, with less di­rect fo­cus, but with prob­a­bly a similar amount of net de­vel­op­ment time go­ing into them even if the em­pha­sis is on more tar­geted pro­grams. This is partly be­cause we value the ex­is­tence of an in­de­pen­dent ra­tio­nal­ity com­mu­nity (varied folks do­ing varied things adds to the art and in­creases its in­tegrity), and partly be­cause we’re still de­pen­dent on the work­shop rev­enue for part of our op­er­at­ing bud­get.

  Re: oth­ers tak­ing up the man­tel: we are work­ing to boot­strap an in­struc­tor train­ing; have long been en­courag­ing our men­tors and alumni to run their own thin­gies; and are glad to help oth­ers do so. Also Kaj So­tala seems to be de­vel­op­ing some in­ter­est­ing train­ing thin­gies de­signed to be shared.

• AnnaSalamon 3 Dec 2016 4:56 UTC

  0 points

  in reply to: Academian's comment on: CFAR’s new fo­cus, and AI Safety

  Got any that con­trast with “rais­ing aware­ness” or “out­reach”?

• AnnaSalamon 3 Dec 2016 2:50 UTC

  5 points

  in reply to: Vladimir_Nesov's comment on: Fact Posts: How and Why

  You can ba­si­cally be­lieve the con­tents of an in­tro chem­istry text­book; you have to be much more care­ful with the con­tents of an in­tro psy­chol­ogy or so­ciol­ogy text­book.