Ben_West

• Ben_West 15 May 2019 16:07 UTC

  6 points

  in reply to: Jon_Behar's comment on: A Frame­work for Think­ing about the EA La­bor Market

  Cool. I think that is a helpful seg­men­ta­tion.
  
  But are you say­ing a) there are lots of peo­ple in co­hort 2 who are great can­di­dates so we should go af­ter them, or b) there are not very many peo­ple in co­hort 2, but be­cause they are from some un­der­rep­re­sented de­mo­graphic we should still go af­ter them?

• Ben_West 14 May 2019 15:34 UTC

  4 points

  in reply to: Jon_Behar's comment on: A Frame­work for Think­ing about the EA La­bor Market

  I think the EA com­mu­nity has got­ten caught up in the ob­ser­va­tion that “there are a lot of smart peo­ple will­ing to work for way be­low mar­ket” and lost track of the ques­tion of “what’s the right way to struc­ture EA com­pen­sa­tion to max­i­mize im­pact?”

  Not sure I fully un­der­stand this. You’re say­ing some­thing like: “it might be true that in­creas­ing wages will have only a small in­crease in the num­ber of can­di­dates, but those new can­di­dates are un­usu­ally im­pact­ful (be­cause they are from un­der­rep­re­sented groups) so it’s still worth do­ing?”

• Ben_West 13 May 2019 23:30 UTC

  8 points

  on: A Frame­work for Think­ing about the EA La­bor Market

  Thanks for writ­ing this Jon!

  “But I be­lieve many EAs are re­sis­tant to rais­ing salaries as a way to close tal­ent gaps be­cause they con­flate will­ing­ness to work for low pay with fit for a job”
  
  The claim I have heard most fre­quently is not this but rather that la­bor sup­ply is in­elas­tic be­low mar­ket rates. E.g. there are peo­ple who re­ally want to work for you (be­cause of your mis­sion or the pres­tige or what­ever), and to them be­ing paid 60% mar­ket rate is ba­si­cally the same as be­ing paid 90% mar­ket rate. So rais­ing your com­pen­sa­tion from 60% to 90% mar­ket rate won’t ac­tu­ally at­tract more can­di­dates.
  
  (I mod­ified your pic­ture here to show this – you can see that the quan­tity sup­plied Qs is very close to the equil­ibrium quan­tity sup­plied Qs*.)
  
  I don’t know if this is ac­tu­ally true (it seems sim­plis­tic, at the least), but it seems con­sis­tent with ev­ery­thing you’ve writ­ten above yet still doesn’t im­ply EA or­ga­ni­za­tions should fo­cus on rais­ing salaries.

• Ben_West 13 May 2019 22:57 UTC

  5 points

  on: Non-Profit In­surance Agency

  Thanks for con­sid­er­ing this! I be­lieve you are con­sid­er­ing what I would call “earn­ing to give” as an in­surance agent. 80 K has more info on earn­ing to give, if you have not already seen that: https://​​80000hours.org/​​ar­ti­cles/​​earn­ing-to-give/​​
  
  Build­ing on Gor­don’s an­swer: LLCs are pass-through en­tities, which means that you can deduct dona­tions up to half your rev­enue/​in­come. You could con­sider some­thing more elab­o­rate (e.g. a cor­po­ra­tion owned by a non­profit foun­da­tion, like New­man’s Own), but that’s prob­a­bly un­nec­es­sar­ily com­plex.

• Ben_West 4 May 2019 17:36 UTC

  9 points

  in reply to: AviNorowitz's comment on: What is the Im­pact of Beyond Meat?

  I didn’t re­al­ize they had dis­con­tinued the chicken prod­ucts. That’s too bad

[Question] What is the Im­pact of Beyond Meat?

• Ben_West 18 Apr 2019 18:25 UTC

  10 points

  on: Le­gal psychedelic re­treats launch­ing in Jamaica

  Weaker ev­i­dence shows that psychedelic ex­pe­riences pos­i­tively pre­dict liberal and anti-au­thor­i­tar­ian poli­ti­cal views, trait open­ness

  In­creas­ing open­ness does not seem uniformly good, e.g. SSC wrote a spec­u­la­tive blog post that psychedelic use may make one “open” to pseu­do­science, con­spir­acy the­o­ries etc. I’m cu­ri­ous if you have thoughts on this?

• Ben_West 19 Mar 2019 18:12 UTC

  7 points

  on: Con­cept: EA Donor List. To en­able EAs that are start­ing new pro­jects to find seed donors, es­pe­cially for peo­ple that aren’t well connected

  Thanks for think­ing of this! My ex­pe­rience is that, in both for-profit and non­profit spaces, the limit­ing con­straint is not knowl­edge that fund­able pro­jects ex­ist. Rather, it’s the lack of due dili­gence on the pro­jects (and peo­ple who can do that sort of DD).
  
  In for-profit an­gel in­vest­ing, usu­ally one in­vestor will take the “lead”, mean­ing that they do a full ex­am­i­na­tion of the startup: speak with cus­tomers, au­dit the fi­nan­cials, do back­ground checks on the founders, etc. Other in­vestors will in­vest con­di­tional on the lead sign­ing off. Cer­tain groups will usu­ally pre­fer to lead or not; some of them will make in­vest­ments into hiring lawyers, ac­coun­tants etc. to help them do this due dili­gence, whereas oth­ers will pre­fer to just defer to other lead in­vestors.
  
  I’m not aware of any en­tity similar to a lead in­vestor in the EA com­mu­nity. Peo­ple some­times sug­gest just fol­low­ing on with OpenPhil (i.e. only donat­ing to or­ga­ni­za­tions which OpenPhil grants to) – this doesn’t seem un­rea­son­able, but it does mean that many or­ga­ni­za­tions will be left un­funded.

• Ben_West 14 Mar 2019 0:09 UTC

  2 points

  in reply to: David_Moss's comment on: Iden­ti­fy­ing Ta­lent with­out Cre­den­tial­ing In EA

  In­di­vi­d­u­als at com­pa­nies are bad at hiring for ex­pected performance

  Fair – an im­plicit as­sump­tion of my post is that mar­kets are effi­cient. If you don’t think so, then what I had to say is prob­a­bly not very rele­vant.

• Ben_West 13 Mar 2019 17:46 UTC

  4 points

  in reply to: Peter_Hurford's comment on: Iden­ti­fy­ing Ta­lent with­out Cre­den­tial­ing In EA

  At ba­si­cally ev­ery com­pany in the world, some­one who comes in with a 10-year-long track record of suc­cess is go­ing to be way more likely to be hired than some­one fresh out of col­lege, even if they are equally skil­led. It would be pretty sur­pris­ing to me if you are able to out­perform the hiring prac­tices of all these com­pa­nies—do you have a sense of why you think you can? Many orgs use work-sam­ple tests, so that seems too sim­ple of an ex­pla­na­tion.

• Ben_West 12 Mar 2019 0:33 UTC

  5 points

  in reply to: G Gordon Worley III's comment on: Iden­ti­fy­ing Ta­lent with­out Cre­den­tial­ing In EA

  Thanks for the feed­back! I meant “cre­den­tials” to in­clude things like work ex­pe­rience, and per­haps should have used more ex­am­ples like that to be clear.

Iden­ti­fy­ing Ta­lent with­out Cre­den­tial­ing In EA

• Ben_West 26 Dec 2018 21:16 UTC

  3 points

  on: An­i­mal-Welfare Eco­nomic Re­search Questions

  You might be in­ter­ested in The Role of Eco­nomics in Achiev­ing Welfare Gains for An­i­mals. Here is the ab­stract:

  The de­mand for an­i­mal prod­ucts and ser­vices is a pow­er­ful eco­nomic force in so­ciety, and multi­billion-dol­lar in­dus­tries are or­ga­nized around this de­mand. Th­ese in­dus­tries of­ten face in­creased costs by im­prov­ing an­i­mal welfare and are quick to use eco­nomic ar­gu­ments against pro­posed welfare re­forms (see side­bar on page 169). Th­ese ar­gu­ments, while of­ten specious, can in­fluence con­sumers, vot­ers, and policy mak­ers. Ci­ti­zens are less likely to sup­port an­i­mal welfare re­forms they’ve been told will dou­ble their shop­ping bill or im­pov­er­ish fam­ily farm­ers.

  An­i­mal welfare ad­vo­cates can­not re­spond to these eco­nomic ar­gu­ments with moral rhetoric alone. In­stead, non-gov­ern­men­tal ob­servers (NGOs) must challenge the eco­nomic as­sump­tions, calcu­la­tions, and con­clu­sions of an­i­mal in­dus­tries and pro­duce re­li­able eco­nomic ar­gu­ments of their own. To do so they should un­der­stand some ba­sic eco­nomic prin­ci­ples, which we re­view be­low, and, when pos­si­ble, en­list the help of economists.

  Fear­ing, J., & Ma­theny, G. (2007). The role of eco­nomics in achiev­ing welfare gains for an­i­mals. In D.J. Salem & A.N. Rowan (Eds.), The state of the an­i­mals 2007 (pp. 159-173). Wash­ing­ton, DC: Hu­mane So­ciety Press.

• Ben_West 20 Dec 2018 1:21 UTC

  34 points

  on: 2018 AI Align­ment Liter­a­ture Re­view and Char­ity Comparison

  Thanks Ben! Great as always. One quib­ble:

  As such, in gen­eral I do not give full cre­dence to char­i­ties say­ing they need more fund­ing be­cause they want more than a year of run­way in the bank. A year’s worth of re­serves should provide plenty of time to raise more fund­ing.

  1. 12 months of run­way means that an or­ga­ni­za­tion with an an­nual fundrais­ing drive will be near bankruptcy once per year, ev­ery year. That seems bad.

  2. I agree that most or­ga­ni­za­tions should be able to raise money in less than 12 months if they are 100% fo­cused on rais­ing money, but not hav­ing to worry about fundrais­ing seems pretty valuable to me. An 18 month run­way means, for ex­am­ple, that if your an­nual fundrais­ing drive goes poorly you still have 6+ months to find some solu­tion.

• Ben_West 11 Dec 2018 22:48 UTC

  5 points

  on: Why I’m fo­cus­ing on in­ver­te­brate sentience

  When I bring this up with EAs who are fo­cused on AI safety, many of them sug­gest that we only need to get AI safety right and then the AI can solve the ques­tion of what con­scious­ness is. This seems like a plau­si­ble re­sponse to me. How­ever, there are some pos­si­ble fu­ture sce­nar­ios where this might not be true. If we have to di­rectly spec­ify our val­ues to a su­per­in­tel­li­gent AI, rather than it learn­ing the value more in­di­rectly, we might have to spec­ify a defi­ni­tion of con­scious­ness for it. It might also be good to have a failsafe mechanism that would pre­vent an AI from switch­ing off be­fore im­ple­ment­ing any sce­nario that in­volved a lot of suffer­ing, and to do this we might have to roughly un­der­stand in ad­vance which be­ings are and are not con­scious.

  It seems like there is some asym­me­try here as is com­mon with ex­tinc­tion risk ar­gu­ments: if we think that we will, even­tu­ally, figure out what con­scious­ness is then, as long as we don’t go ex­tinct, we will even­tu­ally cre­ate pos­i­tive AGI. Whereas, if we fo­cus on con­scious­ness and then AGI kills ev­ery­one, we never get to a pos­i­tive out­come.

  I think the origi­nal ar­gu­ment works if our val­ues get “locked in” once we cre­ate AGI, which is not an un­rea­son­able thing to as­sume, but also doesn’t seem guaran­teed. Am I think­ing through this cor­rectly?

• Ben_West 29 Nov 2018 22:34 UTC

  5 points

  on: From hu­mans in Canada to bat­tery caged chick­ens in the United States, which an­i­mals have the hard­est lives: results

  This is re­ally cool! One thing which stuck out to me: you list that there are the same num­ber of bugs as there are fac­tory farmed fish. Is that re­ally cor­rect? I would have thought that there would be many more bugs than fish.

• Ben_West 21 Nov 2018 19:05 UTC

  2 points

  in reply to: KevinWatkinson's comment on: Effec­tive Altru­ism Mak­ing Waves

  The em­pha­sis for me has been a race to make short term gains whilst medium to longer term pro­jects have been marginal­ised or just not considered

  ACE re­cently did an anal­y­sis of how re­sources are al­lo­cated in the farmed an­i­mal move­ment. You can see from figure 7 that ACE fund­ing goes more to­wards build­ing al­li­ances and ca­pac­ity (the “long-term” parts of their on­tol­ogy) than in the move­ment more gen­er­ally.

  (ACE ar­gues that the amount is still too small. But it seems weird to crit­i­cize EAA for that, since ACE is do­ing bet­ter than the rest of the move­ment, and seems to be plan­ning to do even more.)

• Ben_West 7 Nov 2018 22:36 UTC

  0 points

  in reply to: Peter_Hurford's comment on: Sur­vey of EA org lead­ers about what skills and ex­pe­rience they most need, their staff/​dona­tions trade-offs, prob­lem pri­ori­ti­sa­tion, and more.

  I per­son­ally feel much more fund­ing con­strained /​ man­age­ment ca­pac­ity con­strained /​ team cul­ture “don’t grow too quickly” con­strained than I feel “I need more tal­ented ap­pli­cants” con­strained.

  I feel like part of the defi­ni­tion of “tal­ented ap­pli­cant” is that they don’t stretch your man­age­ment ca­pac­ity, don’t mess up your cul­ture, etc. For ex­am­ple, if there was some­one who had vol­un­teered at Re­think for a while, you had a lot of trust in them, they knew your pro­jects in­ti­mately and could hit the ground run­ning etc., my guess is that you would value that per­son much more highly than some­one who had “gen­eral” com­pe­tency.

  And the next level up would be can­di­dates who not only don’t stretch your man­age­ment ca­pac­ity or cul­ture but ac­tu­ally add to it.

  My ex­pe­rience is that there are lots of peo­ple who are good at re­search or pro­gram­ming or what­ever but fewer who have those skills and can add value to the or­ga­ni­za­tion with­out sub­tract­ing from other limited re­sources.

• Ben_West 22 Feb 2018 18:49 UTC

  2 points

  in reply to: Gregory_Lewis's comment on: Why I pri­ori­tize moral cir­cle ex­pan­sion over ar­tifi­cial in­tel­li­gence alignment

  AI de­sign­ers, even if speciesist them­selves, might nonethe­less provide the right ap­para­tus for value learn­ing such that re­sult­ing AI will not prop­a­gate the moral mis­takes of its creators

  This is some­thing I also strug­gle with in un­der­stand­ing the post. it seems like we need:

  1. AI cre­ators can be con­vinced to ex­pand their moral circle

  2. De­spite (1), they do not wish to be con­vinced to ex­pand their moral circle

  3. The AI fol­lows this sec­ond de­sire to not be con­vinced to ex­pand their moral circle

  I imag­ine this hap­pen­ing with cer­tain re­li­gious things; e.g. I could imag­ine some­one say­ing “I wish to think the Bible is true even if I could be con­vinced that the Bible is false”.

  But it seems rel­a­tively im­plau­si­ble with re­gards to MCE?

  Par­tic­u­larly given that AI safety talks a lot about things like CEV, it is un­clear to me whether there is re­ally a strong trade-off be­tween MCE and AIA.

  (Note: Jacy and I dis­cussed this via email and didn’t re­ally come to a con­sen­sus, so there’s a good chance I am just mi­s­un­der­stand­ing his ar­gu­ment.)

• Ben_West 10 Jan 2018 15:10 UTC

  0 points

  in reply to: DavidMoss's comment on: EA Sur­vey 2017 Series: Cause Area Preferences

  Thanks. I was hop­ing that there would be ag­gre­gate re­sults so I don’t have to re­peat the anal­y­sis. It looks like maybe that in­for­ma­tion ex­ists el­se­where in that folder though? https://​​github.com/​​pe­ter­hur­ford/​​ea-data/​​tree/​​mas­ter/​​data/​​2017