Max­i­miz­ing is usu­ally a bad idea.

I looked up CRPS and kid­ney stones, and it looks like both of them have rel­a­tively mild symp­toms in most cases. Are you sure that this isn’t a case of con­flat­ing the pain of the most ex­treme cases and the prevalence of all cases?

You’re right about the 8% figure for chronic se­vere pain, though.

https://​​www.acsh.org/​​news/​​2018/​​09/​​14/​​who-hurt­ing-prevalence-chronic-pain-amer­ica-13407

I won­der if we suffi­ciently un­der­stand the psy­cholog­i­cal dy­nam­ics of chronic or ex­treme pain. The ex­is­tence of the bul­let ant glove rit­ual makes me won­der to what ex­tent the cul­tural con­text of pain in­fluences our re­mem­bered per­cep­tion of its qual­ity, in­ten­sity, and mean­ing.

It seems helpful to dis­t­in­guish be­tween mean­ingless, “I would’t wish this on my worst en­emy” pain, which prob­a­bly ac­counts for the vast ma­jor­ity of ex­treme pains, “a lit­tle pain is nec­es­sary to toughen you up” pain, and “this is a sa­cred ex­treme-pain rit­ual.”

Glad you liked it :) I added a sen­tence about in­di­rect effects in the first para­graph. I see your point about the ti­tle but I chose to leave it as it is be­cause:

1. I think that peo­ple don’t ex­pect cost-effec­tive­ness es­ti­mates to take all the in­di­rect effects into ac­count any­way. That said, I am afraid of it be­ing mis­in­ter­preted as an es­ti­mate of what an ad­di­tional donated dol­lar would achieve.

2. How­ever, any­one who would make im­por­tant de­ci­sions based on this es­ti­mate would hope­fully read more than just the ti­tle any­way.

3. I wanted to grab the at­ten­tion of some EAs who would not con­sider helping welfare re­forms oth­er­wise (and hence wouldn’t open an ar­ti­cle called “Cost-effec­tive­ness es­ti­mate of cor­po­rate an­i­mal welfare cam­paigns”).

4. I wanted the re­sult to be promi­nently fea­tured be­cause I’d rather the com­mon knowl­edge within EA would be “it af­fects 9-120 years per dol­lar but there are many com­pli­ca­tions and in­di­rect effects” rather than “it’s un­clear what the cost-effec­tive­ness is”. The former at least let’s peo­ple com­pare the scale of the effects with other in­ter­ven­tions.

5. I don’t want to change the ti­tle now be­cause it could con­fuse some fo­rum read­ers, and make it more difficult to find for read­ers who re­mem­ber the old ti­tle.

Why not just use Glass­door?

This post got me won­der­ing whether there should be Glass­door for EA.

I was asked to com­ment here. As you know, I did a data sci­ence in­tern­ship at Im­pos­si­ble Foods in late 2016. I’m mostly jot­ting down my own ex­pe­riences, along with some anonymized in­for­ma­tion from talk­ing to oth­ers.

NB: “Tech” be­low refers to jobs that are con­sid­ered main­stream tech in Sili­con Valley (soft­ware, data sci­ence, an­a­lyt­ics, etc), while “sci­ence” refers to the food sci­ence/​bio­chem­istry/​chem­istry work that is Im­pos­si­ble’s core product.

Pros:

• Highly mis­sion-driven. Many peo­ple were veg­e­tar­ian or ve­gan (all the food the com­pany served was ve­gan by de­fault), and peo­ple there seemed fairly ded­i­cated to the cause of re­plac­ing farmed an­i­mals with plants (less than I would ex­pect from a EA or AR non­profit)

• Diver­sity. The gen­der ra­tio in the main office was slightly more women than men, and there was a lot of rep­re­sen­ta­tion from differ­ent coun­tries that I usu­ally don’t see in Sili­con Valley (though this could just be be­cause biol­ogy/​bio­chem­istry draws from a differ­ent pop­u­la­tion than CS).

• Nice­ness. Peo­ple seemed re­ally nice to each other a lot, and there wasn’t a lot of the asshol­ish per­son­al­ities I some­times as­so­ci­ate with star­tups.

• In­ter­est­ing prob­lems. My sub­jec­tive sense is that tech there is usu­ally used to sup­port sci­en­tific pur­suits rather than eg, tech as a product or busi­ness de­vel­op­ment, and is more in­ter­est­ing in a broader sense than most big com­pany or startup work.

• Lots of op­por­tu­ni­ties to grow. Peo­ple who’re up for the challenge of­ten take on quite im­pres­sive challenges at low lev­els of se­nior­ity.

• Benefits. I didn’t use them much, but my im­pres­sion is that the com­pany seemed quite pro­gres­sive about things like va­ca­tion days and pa­ter­nity leave(?).

• Rea­son­able work-life bal­ance. This seemed true of the tech peo­ple I knew, how­ever the sci­en­tists seemed a lit­tle over­worked and the busi­ness de­vel­op­ment peo­ple seemed a lot over­worked. I don’t know how this com­pares to other star­tups.

• The CEO (Pat Brown) ap­peared highly com­pe­tent and clearly thought­ful. From my rel­a­tively brief in­ter­ac­tions with him, there’s a rea­son­able chance he would have been at home in Stan­ford EA if he was much younger. Eg, he talks about quan­ti­ta­tive cause pri­ori­ti­za­tion and had a short rant at one point about se­lec­tion bias in busi­ness ad­vice.

Cons:

• Low pay. I feel like there’s a large mis­sion/​salary trade­off that the com­pany makes be­cause it knows it could hire enough True Believ­ers. My in­tern pay was sub­stan­tially be­low mar­ket, and this seemed true of the other in­terns I talked to, as well as full-timers I talked to in broadly “tech” roles. I don’t know if this has changed by 2019. Another caveat is that I didn’t ask about equity, and Im­pos­si­ble’s val­u­a­tion ~quadru­pled in the last 3 years, so it’s quite pos­si­ble full-timers were ac­tu­ally well-com­pen­sated even if they didn’t per­ceive it that way at the time. A fi­nal caveat is that I’m com­par­ing with other for-profit com­pa­nies, and maybe a bet­ter point of com­par­i­son is (EA) non­prof­its or academia, and my guess is that Im­pos­si­ble pays bet­ter.

• Sub­par con­flict re­s­olu­tion. I was pretty shielded from the poli­tics as an in­tern, but I hear more bad sto­ries from oth­ers than I would ex­pect from a com­pany of its size (caveat: I have a very poor un­der­stand­ing of the ac­tual base rate of bad con­flicts at suc­cess­ful com­pa­nies). Pos­si­bly be­cause of the nice­ness? I feel like peo­ple leave on bad terms more than I would guess.

• Tech­ni­cal men­tor­ship. Be­cause tech is not the main product, you’ll get less se­nior men­tor­ship or guidance than a pri­mar­ily tech com­pany. (Ob­vi­ously, the op­po­site is true if you’re a food sci­en­tist or bio­chemist).

• In­cre­men­tal­ist work. Im­pos­si­ble always had a vi­sion of be­ing the even­tual re­place­ment of all an­i­mal-based prod­ucts, how­ever when I joined in 2016, it was very much at the tail end of ex­per­i­men­ta­tion and the be­gin­ning of be­ing laser-fo­cused on beef, which seems less in­tel­lec­tu­ally and al­tru­is­ti­cally in­ter­est­ing. My im­pres­sion is that this was much more true as of 2018, how­ever they seemed to have de­vel­oped pork and fish re­place­ments re­cently? [1]

Neu­trals:

• The com­pany seems fairly high-pres­tige in the pub­lic eye. It’s ex­tremely well-known for its size, and peo­ple are of­ten ex­cited to talk to me about the work there (in a way that I’ve never ex­pe­rienced be­fore or since). This seems good for ca­reer cap­i­tal, and well-be­ing, how­ever I want to cau­tion against see­ing this as a clear pos­i­tive. It’s easy to fall into pres­tige traps, and peo­ple should in­tro­spect about this be­fore they ap­ply. (Also lo­cal pres­tige mat­ters more than global pres­tige for most job pivots, so pub­lic opinion is a poor proxy for how much fu­ture em­ploy­ers care).

• En­vi­ron­men­tal­ism. Peo­ple at Im­pos­si­ble are much more likely to be en­vi­ron­men­tal­ists than an­i­mal welfare peo­ple. Per­son­ally I find Deep Ecol­ogy views to be philo­soph­i­cally un­ten­able, but ob­vi­ously other EAs have differ­ent philo­soph­i­cal views. I write this so peo­ple can make an in­formed de­ci­sion self-se­lect­ing in.

On bal­ance, I don’t think I’m in­formed enough to judge whether work­ing at Im­pos­si­ble is bet­ter than a typ­i­cal reader’s al­ter­na­tives. My gut in­stinct is that if you have other al­tru­is­tic op­tions that can make full use of your skil­lsets (clean meat seems es­pe­cially ex­cit­ing), then it’s more im­pact­ful to do more early-stage work than be­ing at Im­pos­si­ble, but I’m very un­cer­tain about this opinion and it’s con­founded by a lot of de­tails on the ground.

Ad­di­tional Note 2019/​7/​20: Reread­ing this, I think peo­ple are usu­ally bi­ased against ap­ply­ing, and I think it’s still worth­while for peo­ple who con­sider farmed an­i­mal welfare their top (or close to top) cause area to ap­ply to Im­pos­si­ble.

[1] https://​​www.digi­taltrends.com/​​cool-tech/​​im­pos­si­ble-sausage-lit­tle-cae­sars/​​

I’m a big fan of this re­port and will prob­a­bly recom­mend it to in­ter­ested peo­ple as the best of the cost-effec­tive­ness mod­els I have seen on cor­po­rate welfare com­mit­ments.

I’m very glad for the”Ways this es­ti­mate could be mis­lead­ing” sec­tion. I think its very im­por­tant to make these wider con­sid­er­a­tions clear; they have not been so clear in pre­vi­ous cost-effec­tive­ness es­ti­mates. I also like that you make it clear how you think that these con­sid­er­a­tions weigh up with the pluses and minuses sys­tem.

It’s great that this in­for­ma­tion on var­i­ous un­cer­tain­ties is in­cluded and yet you are still able to provide a use­able es­ti­mate of cost-effec­tive­ness (that ex­cludes these in­di­rect effects). I would prob­a­bly lean to­wards mak­ing this re­sult less promi­nent in the write-up, e.g. not in­clud­ing it in the ti­tle. I do think that, de­spite your clar­ity on the un­cer­tain­ties, it is easy for read­ers to pick up and fo­cus on the fi­nal es­ti­mate and then dis­re­gard the rest of the post.

What does it mean to be “pro-sci­ence”? In other words, what might a po­ten­tial welfarist, max­i­miz­ing, im­par­tial, and non-nor­ma­tive move­ment that doesn’t meet this crite­rion look like?

I ask be­cause I don’t have a clear pic­ture of a defi­ni­tion that would be both in­for­ma­tive and un­con­tro­ver­sial. For in­stance, the main­stream sci­en­tific com­mu­nity was largely dis­mis­sive of SIAI/​MIRI for many years; would “proto-EAs” who sup­ported them at that time be con­sid­ered pro-sci­ence? I as­sume that ex­clud­ing MIRI does in­deed count as con­tro­ver­sial, but then I don’t have a clear pic­ture of what ac­tivi­ties/​causes be­ing “pro-sci­ence” would ex­clude.

edit: Why was this down­voted?

I looked into a dozen or so in­ter­ven­tions for pre­vent­ing firestorm given nu­clear war. I es­ti­mated that some of them could be cost effec­tive only look­ing at sav­ing lives of US cit­i­zens. How­ever, I aban­doned the pro­ject once I re­al­ized that some of the ideas were around in the Cold War, and they were still not im­ple­mented, so it was very un­likely we would im­ple­ment them now.

If one side be­lieves they are be­ing at­tacked and launches weapons, there would not be a pe­riod of ten­sion be­fore the at­tack. Even if there is a pe­riod of ten­sion like the Cuban mis­sile crisis, I don’t be­lieve that caused peo­ple to evac­u­ate cities, though of course the at­tack could not have oc­curred as quickly as it could now.

I’m not sure if your pro­posed diet has the re­quired es­sen­tial oils. Th­ese tend to go ran­cid, though they might still be safe to eat.

(b), per­haps with a dash of (a) too

Very minor point—in the table in Ap­pendix A, a value of “Likely small (3)” is shown, where I imag­ine that should be “Likely small (1)”. (This is for “The Us, France, and the UK all keep their nu­clear weapons…”) I imag­ine fix­ing this would then also change the over­all “score” from −1.5 to −0.5, be­cause the mul­ti­pli­ca­tion for that row would change from −0.5 * 3 to −0.5 * 1.

Thanks for this se­ries—it’s very in­ter­est­ing so far.

Whether a coun­terforce sec­ond strike by Rus­sia would ac­tu­ally cause fewer deaths than a first strike is con­di­tional on 1) the US strik­ing first, 2) Rus­sia choos­ing not to launch on warn­ing, and 3) Rus­sia be­ing sub­stan­tially un­der-pre­pared for a first strike. My best guess is that the prob­a­bil­ity of all three of these be­ing the case is fairly low. If we naively as­sume that the prob­a­bil­ity that the US strikes first is 50%, the prob­a­bil­ity that Rus­sia chooses not to launch on warn­ing is also 50%, and that the US coun­terforce strike de­stroyed the ‘cen­ter value’ of the range for the num­ber of nu­clear weapons that might be de­stroyed (870), or 79% of the num­ber of war­heads I ex­pect Rus­sia would use against the US dur­ing a coun­terforce _first _strike (1,100), I would ex­pect that about 5% fewer deaths would be caused by a Rus­sian sec­ond strike than by a Rus­sian first strike (0.50.50.21).

There’s a good chance I’m just mi­s­un­der­stand­ing this, but shouldn’t that be 19.75% fewer deaths in ex­pec­ta­tion? 0.5 * 0.5 * 0.79 (=0.1975), rather than 0.5 * 0.5 * 0.21 (=0.0525), be­cause the num­ber of weapons used (and thus the num­ber of deaths, if we stick with your as­sump­tion of lin­ear­ity) would go down by 79%, to 21%, rather than go­ing down by 21%. (Again, it’s very pos­si­ble I’m mi­s­un­der­stand­ing the maths here.)

Also, wouldn’t it ac­tu­ally be that a Rus­sian coun­terforce strike in gen­eral (not a Rus­sian coun­terforce sec­ond strike) would cause 19.75%* fewer deaths in ex­pec­ta­tion, given that the sec­ond strike may in­volve fewer nu­clear weapons? Put an­other way, would it ac­tu­ally be that a Rus­sian coun­terforce sec­ond strike would cause 39.5% (two times 19.75%) fewer deaths in ex­pec­ta­tion than a Rus­sian coun­terforce first strike? I ask this be­cause the first mul­ti­pli­ca­tion by 0.5, to rep­re­sent a 50% chance of the US strik­ing first, seems to ac­count for you tak­ing the half of the pos­si­ble wor­lds in which Rus­sia strikes sec­ond, and thus to not to be needed if you’re dis­cussing a Rus­sian sec­ond strike any­way. This seems rele­vant be­cause, if both that and the above are the case, then in the model of the to­tal num­ber of deaths ex­pected from both side’s weapons in an ex­change should be ad­justed to re­duce the deaths from Rus­sian weapons by 19.75%. (As op­posed to just re­duc­ing it by 9.875%, which you’d do if the 19.75% rep­re­sented the re­duc­tion in weapons used con­di­tional on Rus­sia strik­ing sec­ond, rather than in gen­eral.)

Apolo­gies if I haven’t ex­plained that very clearly or I’m mi­s­un­der­stand­ing your rea­son­ing.

*Or ~5%, if my calcu­la­tions to get 19.75% are mis­taken.

I don’t have any ad­vice but just wanted to say that I think it’s re­ally cool that you asked for ad­vice here.

I keep think­ing about this post.

How many EA would I like to see in the clergy? Not zero. I pretty much agree with the “port­fo­lio ap­proach” to ca­reers—I think there are likely un­ex­pected benefits of hav­ing EAs in a bunch of differ­ent ca­reers, even ones that don’t jump out as hav­ing ob­vi­ously EA ap­pli­ca­tions, and it’s good to have some spread.

But who would I like to see in those not-ob­vi­ously-EA ca­reers? 8 years ago, no­body (in­clud­ing me) thought we needed so­cial work­ers in EA. But I stayed with so­cial work be­cause it was ba­si­cally the only thing I could imag­ine do­ing. It fit with so many of my in­ter­ests and skills. Later it turned out that there was a use­ful way to ap­ply those skills to EA. But if 8 years ago some­one had said, “We should get some so­cial work­ers in EA,” I wouldn’t have wanted ran­dom peo­ple say­ing, “Well, I don’t feel all that in­ter­ested in so­cial work, but if that’s the thing we need I guess I’ll do it.” I don’t think that would have been good ei­ther for them or for EA.

So my best guess as to which EAs should en­ter the clergy is the ones who can’t imag­ine do­ing any­thing else, or at least ones who feel pretty ex­cited about it. If that’s not you, you shouldn’t try to force your­self into that mold.

Great ques­tion!

Among the com­pa­nies I’ve spo­ken with, tech­ni­cal po­si­tions seem to be a huge bot­tle­neck, so I’d imag­ine those be­ing the high­est im­pact. In this in­dus­try, I’m con­sid­er­ing tech­ni­cal po­si­tions to gen­er­ally have ti­tles like food sci­en­tist, re­search/​pro­cess as­so­ci­ate, bio­pro­cess en­g­ineer, food tech­nol­o­gist/​de­vel­oper, etc. This bot­tle­neck seems es­pe­cially in­tense among clean meat com­pa­nies, so if you have a rele­vant back­ground, I’d strongly en­courage you to ap­ply.

Beyond that, it’s hard for me to guess ex­actly which po­si­tions are high­est im­pact.

I an­swer these ques­tions and go over the method­ol­ogy in de­tail in the video. A work­ing pa­per will be com­ing soon, but for now all of the de­tails are in the video.

In that case, it seems plau­si­ble that you (and your cowork­ers) will do more and bet­ter work if you’re not just as­ceti­cally grind­ing away for decades (and if they aren’t spend­ing time around some­one like that). Per­haps, a good next step is to shadow/​in­tern with/​talk to peo­ple cur­rently do­ing these jobs to learn what they look like day to day?

I su­per agree with the ti­tle, but I think the text ac­tu­ally re­ally un­der­sells it! Run­way not only in­creases your flex­i­bil­ity to not earn, but also re­duces your stress and re­moves all sorts of psy­cholog­i­cally difficult power dy­nam­ics that come with hav­ing a boss or oth­er­wise be­ing be­holden to ex­ter­nal fac­tors for your well be­ing (Yes, you may still have a boss or ex­ter­nal fac­tors, but now you won’t need their con­tinued ap­proval or suc­cess to pay bills, and that makes all the differ­ence). Also, fru­gal­ity en­ables you to re­ally splurge with­out wor­ry­ing when it re­ally counts. Ad­di­tion­ally, If you do not have any large and ex­pen­sive pos­ses­sions, tend to live in low cost apart­ments, and don’t have any de­pen­dents, you can move to what­ever lo­ca­tion it is most pro­duc­tive for you to be in with lit­tle to no over­head—whether that be across town or across the globe. Fru­gal­ity in an ur­ban con­text also forces close liv­ing situ­a­tions (house­mates) which can dra­mat­i­cally in­crease your so­cial net­work. Fur­ther, you end up build­ing scrappy skills and habits (e.g. ne­go­ti­at­ing apart­ments, meal plan­ning, knowl­edge of pub­lic ser­vices, bik­ing) which can re­ally come in handy even when you’re not be­ing fru­gal.

If you have the priv­ilege to be in cir­cum­stances where you are able to make money with­out spend­ing most of it, it’s good to take ad­van­tage of this if you can. Don’t feel bad about it if you can’t—it’s not always sim­ple or pos­si­ble for ev­ery­one. But if you feel like it would be pretty easy for you to be fru­gal and you’re choos­ing not to be­cause you think spend­ing a lot more makes you more pro­duc­tive, I strongly sug­gest re­con­sider.

Another point worth con­sid­er­ing is that if you are suffi­ciently fru­gal, and if “pro­duc­tivity” is truly your goal here, you can “in­crease your pro­duc­tivity” by tak­ing that money and hiring a sec­ond per­son to work on your pro­ject with you. Can all your time sav­ing ex­penses in­crease your pro­duc­tivity more than a whole sec­ond per­son? (I’m sure there are some cir­cum­stances for which the an­swer is yes, but I imag­ine that is rare.)

What tem­per­a­ture change does this as­sume? How do ex­cess deaths scale with differ­ent de­grees of warm­ing?