Denise_Melchin

• Denise_Melchin 13 Feb 2020 22:03 UTC

  10 points

  on: My per­sonal cruxes for work­ing on AI safety

  This was great, thank you. I’ve been ask­ing peo­ple about their rea­sons to work on AI safety as op­posed to other world im­prov­ing things, as­sum­ing they want to max­i­mize the world im­prov­ing things they do. Won­der­ful when peo­ple write it up with­out me hav­ing to ask!

  One thing this post/​your talk would have benefited from to make things clearer (or well, at least for me) is if you gave more de­tail on the ques­tion of how you define ‘AGI’, since all the cruxes de­pend on it.

  Thank you for defin­ing AGI as some­thing that can do reg­u­larly smart hu­man things and then ask­ing the very im­por­tant ques­tion how ex­pen­sive that AGI is. But what are those reg­u­larly smart hu­man things? What frac­tion of them would be nec­es­sary (though that de­pends a lot on how you define ‘task’)?

  I still feel very con­fused about a lot of things. My im­pres­sion is that AI is much bet­ter than hu­mans at quite a few nar­row tasks though this de­pends on the defi­ni­tion. If AI was sud­denly much bet­ter than hu­mans at half of all the tasks hu­man can do, but sucked at the rest, then that wouldn’t count as ar­tifi­cial ‘gen­eral’ in­tel­li­gence un­der your defi­ni­tion(?) but it’s un­clear to me whether that would be any less trans­for­ma­tive though this de­pends a lot on the cost again. Now that I think about it, I don’t think I un­der­stand how your defi­ni­tion of AGI is differ­ent to the re­sults of whole-brain em­u­la­tion, apart from the fact that they used differ­ent ways to get there. I’m also not clear on whether you use the same defi­ni­tion as other peo­ple, whether those usu­ally use the same one and how much all the other cruxes de­pend on how ex­actly you define AGI.

• Denise_Melchin 12 Feb 2020 21:44 UTC

  5 points

  in reply to: Robert_Wiblin's comment on: Who should give sperm/​eggs?

  I’m fairly sur­prised by this re­sponse, this doesn’t match what I have read. The Hu­man Fer­til­i­sa­tion and Em­bry­ol­ogy Author­ity im­poses a limit for sperm and egg donors to donate to a max­i­mum of ten fam­i­lies in the UK, al­though there is no limit on how many chil­dren might be born to these ten fam­i­lies (I’m strug­gling to link, but google ‘HFEA ten fam­ily limit’). But re­al­is­ti­cally, they won’t all want to have three chil­dren.

  I’m cu­ri­ous whether you have a source for the claim that 99% of prospec­tive sperm donors in the UK get re­jected? I’m much less con­fi­dent about this, but this doesn’t line up with my im­pres­sion. I also didn’t have the im­pres­sion they were par­tic­u­larly picky about egg donors, un­like in the US.

  But yes, it’s true for sperm and egg donors al­ike that in the UK they can be con­tacted once the offspring turns 18.

• Denise_Melchin 9 Feb 2020 19:27 UTC

  5 points

  in reply to: Khorton's comment on: Who should give sperm/​eggs?

  There are also mul­ti­ple med­i­cal and ge­netic ap­point­ments re­quired in ad­vance. I am cur­rently un­der­go­ing the pro­cess to be­come an egg donor in the UK (though there is a good chance that I will be re­jected) and the pro­cess is quite in­volved. To some ex­tent, this is also true for sperm donors.

• Denise_Melchin 9 Feb 2020 19:25 UTC

  5 points

  on: Who should give sperm/​eggs?

  Ad­ding to what Khor­ton said, it de­pends a lot on what your bar for do­ing good that you con­sider worth do­ing is and what you con­sider ‘do­ing good’ to be.

  In the UK, there is an egg and sperm donor short­age, so there is some chance you will cause chil­dren to ex­ist that wouldn’t have ex­isted oth­er­wise (in­stead of just ‘re­plac­ing’ chil­dren).

• Denise_Melchin 26 Jan 2020 18:18 UTC

  2 points

  in reply to: Milan_Griffes's comment on: Do­ing good is as good as it ever was

  No, I haven’t. Given the amount of up­votes Phil’s com­ment re­ceived (from which I con­clude a de­cent frac­tion of peo­ple do find ar­gu­ments in this space de­mo­ti­vat­ing which is im­por­tant to know) I will prob­a­bly read up on it again. But I very rarely write top-level posts and the prob­a­bil­ity of this in­ves­ti­ga­tion turn­ing into one is neg­ligible.

• Denise_Melchin 25 Jan 2020 19:58 UTC

  4 points

  in reply to: Denise_Melchin's comment on: Do­ing good is as good as it ever was

  Through think­ing about these com­ments, I did re­mem­ber an EA Fo­rum thread in which ii) and iii) were ar­gued about from 4 years ago: https://​​fo­rum.effec­tivealtru­ism.org/​​posts/​​ajPY6zxSFr3BbMsb5/​​are-givewell-top-char­i­ties-too-speculative

  It’s worth read­ing the com­ment sec­tion in full. Turns out my po­si­tion has been con­sis­tent for the past 4 years (though I should have re­mem­bered that thread!).

• Denise_Melchin 25 Jan 2020 19:31 UTC

  18 points

  in reply to: Aaron Gertler's comment on: Do­ing good is as good as it ever was

  I’ve been in­volved in the com­mu­nity since 2012 - the changes seem dras­tic to me, both based on in-per­son in­ter­ac­tions with dozens of peo­ple as well as changes in the on­line land­scape (e.g. the EA Fo­rum/​EA Face­book groups).

  But that is not in it­self sur­pris­ing. The EA com­mu­nity is on av­er­age older than when it started. Youth move­ments are known for be­com­ing less en­thu­si­as­tic and am­bi­tious over time, when it turns out that chang­ing the world is ac­tu­ally re­ally, re­ally hard.

  A bet­ter test is: how mo­ti­vated do EAs feel who are of a similar de­mo­graphic to long-term EAs years ago when EA started? I have the im­pres­sion they are much less mo­ti­vated. It used to be a com­mon oc­cur­rence in e.g. Face­book groups to see peo­ple writ­ing about how mo­ti­vat­ing they have found it to be around other EAs. This is much rarer than it used to be. I’ve met a few new-ish early 20s EAs and I don’t think I can even name a sin­gle one who is as en­thu­si­as­tic as the av­er­age EA was in 2013. I won­der whether the lack of new pro­jects be­ing started by young EAs is par­tially caused by this (though I am sure there are other causes).

  To be clear, I don’t think there has been as dras­tic a change since 2018, which is I think when you started par­ti­ci­pat­ing in the com­mu­nity.

Do­ing good is as good as it ever was

• Denise_Melchin 22 Jan 2020 22:08 UTC

  4 points

  in reply to: trammell's comment on: Do­ing good is as good as it ever was

  In prin­ci­ple you only need i) and iii), that’s true, but I think in prac­tice ii) is usu­ally also re­quired. Hu­mans are fairly scope in­sen­si­tive, and I doubt we’d see low com­mu­nity morale from or­di­nary do good­ing ac­tions be­ing less good by a fac­tor of two or three. As an ex­am­ple, his­tor­i­cally GiveWell es­ti­mates of how much sav­ing a life with AMF costs have differed by about this much—and it didn’t seem to have much of an im­pact on com­mu­nity morale. Not so now.

  Our crux seems to be that you as­sume clue­less­ness or ideas in the same space are a large fac­tor in pro­duc­ing low com­mu­nity morale for do­ing good. I must ad­mit that I was sur­prised by this re­sponse, I per­son­ally haven’t found these ar­gu­ments to be par­tic­u­larly per­sua­sive, and most peo­ple around me seem to feel similarly about such ar­gu­ments, if they are fa­mil­iar with them at all.

• Denise_Melchin 19 Jan 2020 9:54 UTC

  12 points

  in reply to: trammell's comment on: Do­ing good is as good as it ever was

  Yep, I agree that if i) you per­son­ally buy into the long-ter­mist the­sis, and ii) you ex­pect the long-term effects of or­di­nary do good­ing ac­tions to be big­ger than short-term effects, and iii) you ex­pect these long-term effects to be nega­tive, then it makes sense to be less en­thu­si­as­tic about your abil­ity to do good than be­fore.

  How­ever, I doubt most peo­ple who feel like I de­scribed in the post fall into this cat­e­gory. As you said, you were un­cer­tain about how com­mon this feel­ing is. Lots of peo­ple hear about the much big­ger im­pact you can have by fo­cussing on the far fu­ture. Sig­nifi­cantly fewer are well versed in the spe­cific de­tails and prac­ti­cal im­pli­ca­tions of long-ter­mism.

  While I have heard about peo­ple be­liev­ing ii) and iii), I haven’t seen ei­ther ar­gu­ment care­fully writ­ten up any­where. I’d as­sume this is true for lots of peo­ple. There has been a big push in the EA com­mu­nity to be­lieve i), this has not been true for ii) and iii) as far as I can tell.

• Denise_Melchin 8 Jan 2020 16:39 UTC

  51 points

  on: In praise of un­his­toric heroism

  Think­ing about this fur­ther, one con­cern I have with this post as well as Ollie’s com­ment is that I think peo­ple could un­duly un­der­rate the amount of good the av­er­age Westerner can ac­tu­ally do af­ter read­ing it.

  If you have a rea­son­ably high salary or donate more than 10% (and as­sum­ing dona­tions don’t be­come much less cost-effec­tive) to AMF or similarly effec­tive char­i­ties, you can save hun­dreds of lives over your life­time. Sav­ing one life via AMF is cur­rently es­ti­mated to cost around only £2,500. If you only earn the av­er­age grad­u­ate salary for­ever and only donate 10%, you can still save dozens of lives.

  For refer­ence, Oskar Sch­indler saved 1200 lives and is now fa­mous for it wor­ld­wide.

  My words at some­one’s funeral who saved dozens or even hun­dreds of lives would be a lot more lauda­tory than what was said about Dorothea.

• Denise_Melchin 8 Jan 2020 12:01 UTC

  6 points

  on: In praise of un­his­toric heroism

  Great post. I also think we could work more on the root cause of peo­ple feel­ing like this. Per­haps the mes­sage should be: “Do­ing good and hav­ing an im­pact is not about you. Do­ing good is for the world, its peo­ple and other liv­ing be­ings.”

• Denise_Melchin 6 Jan 2020 14:39 UTC

  34 points

  on: More info on EA Global admissions

  This might be a slightly silly sug­ges­tion and I’m not sure how best to im­ple­ment it, but I think it might be use­ful to re­mind po­ten­tial at­ten­dees that at­tend­ing EAG is not obli­ga­tory just be­cause you are part of the EA Com­mu­nity and/​or care a lot about do­ing good well. I heard from a few peo­ple who weren’t par­tic­u­larly ex­cited about at­tend­ing EAG, but still did it be­cause that’s ‘what you do as an EA’. It seems sad that these peo­ple take up spots from peo­ple who are ac­tu­ally keen on EAG it­self.

  It only oc­curred to me fairly late last year that at­tend­ing EAG is ac­tu­ally en­tirely op­tional. On a side note, ris­ing ticket prices did help me come to the re­al­i­sa­tion that I did not ac­tu­ally want to go (and there­fore didn’t take up a spot from some­one who was more keen on go­ing).

• Denise_Melchin 4 Jan 2020 23:19 UTC

  24 points

  in reply to: Habryka's comment on: More info on EA Global admissions

  I don’t feel like I get more value out of large con­fer­ences and I’d be cu­ri­ous about see­ing more data on this ques­tion. For me, hav­ing more peo­ple at a con­fer­ence makes it harder to phys­i­cally find the peo­ple I ac­tu­ally want to talk to. They make up a smaller frac­tion of at­ten­dees and are more spread out. I have also had the im­pres­sion that con­ver­sa­tions at large con­fer­ences are shorter. In com­bi­na­tion, I get much less value out of very large events com­pared to small or medium sized ones.

  The event size was one of the main rea­sons I de­cided not to at­tend EAG Lon­don this year for the first time. It is too big for me to get suffi­cient value out of it.

• Denise_Melchin 4 Jan 2020 19:56 UTC

  13 points

  in reply to: Stefan_Schubert's comment on: Thoughts on do­ing good through non-stan­dard EA ca­reer pathways

  5. also has a nega­tive im­pact on the peo­ple who are try­ing to de­cide be­tween differ­ent ca­reer op­tions and would ac­tu­ally be happy to hear con­struc­tive crit­i­cism. I of­ten feel like I can­not trust oth­ers to be hon­est in their feed­back if I’m de­cid­ing be­tween ca­reer op­tions be­cause they pre­fer to be ‘nice’.

• Denise_Melchin 11 Dec 2019 19:31 UTC

  8 points

  in reply to: Milan_Griffes's comment on: EA Meta Fund Novem­ber 2019 Pay­out Report

  Well, I’d as­sume this is be­cause the LTFF team has more time available than the Meta Fund team. Plau­si­bly largely driven by one vol­un­teer who is very happy to spend a lot of time on the LTFF.

• Denise_Melchin 11 Dec 2019 18:45 UTC

  14 points

  in reply to: Milan_Griffes's comment on: EA Meta Fund Novem­ber 2019 Pay­out Report

  I’m part of the Meta Fund com­mit­tee and was the per­son who de­cided against giv­ing feed­back in the afore­men­tioned cases.

  Un­for­tu­nately, giv­ing good feed­back is very difficult and some­thing the Meta Fund com­mit­tee cur­rently isn’t re­li­ably able to provide. I have pro­vided can­di­dates with feed­back when I felt I could give eas­ily un­der­stand­able prac­ti­cal sug­ges­tions that would ac­tu­ally lead to the pro­ject be­ing more likely to be funded in the fu­ture (or ex­plained why this was not likely to hap­pen) and I could do this with­out in­vest­ing more than a cou­ple of hours per ap­pli­cant.

  In prac­tice, this sadly means ap­pli­cants do not get pro­vided with feed­back very of­ten (I would need to check, but it might be in 20% of cases). I think giv­ing good feed­back is very valuable, but this is un­for­tu­nately cur­rently be­yond our re­sources.

• Denise_Melchin 10 Oct 2019 8:11 UTC

  3 points

  in reply to: Max_Daniel's comment on: My ex­pe­rience on a sum­mer re­search programme

  Thank you Max, strong up­voted. It sounds like you put a lot of thought into mak­ing the pro­gramme run bet­ter than in pre­vi­ous years and suc­ceeded.

• Denise_Melchin 25 Sep 2019 10:27 UTC

  19 points

  in reply to: SiebeRozendal's comment on: My ex­pe­rience on a sum­mer re­search programme

  I agree with this. I think the setup of the CEA Sum­mer Fel­low­ship pro­gramme is a bit con­cern­ing.

  Ad­ding to the points you men­tioned (lit­tle su­per­vi­sion, doesn’t provide good ev­i­dence for ca­reer paths out­side EA or­gani­sa­tions) it is also an un­paid pro­gramme that does not by de­fault re­sult in job offers be­ing made to the best perform­ers. *

  I’m wor­ried that stu­dents will think this pro­gramme will ad­vance their fu­ture ca­reer, while I doubt this is true in most cases. In­stead they might just pay high op­por­tu­nity costs.

  
  

  *At least this was true 1-2 years ago, I’m not en­tirely sure what the most re­cent iter­a­tion of the pro­gramme looks like.

• Denise_Melchin 15 Sep 2019 21:36 UTC

  16 points

  in reply to: saulius's comment on: [Link] What opinions do you hold that you would be re­luc­tant to ex­press in front of a group of effec­tive al­tru­ists? Anony­mous form.

  I would ex­pect that apart from con­tra­cep­tion global health in­ter­ven­tions to be most helpful in re­duc­ing deaths of un­born hu­mans. Mis­car­riages and stil­lbirths are a much big­ger deal than abor­tions, and in de­vel­op­ing coun­tries there is still a lot of room for health in­ter­ven­tions to help for lit­tle money.

  I would be sur­prised if other in­ter­ven­tions to re­duce un­born deaths were very cost-effec­tive, even if you have a wor­ld­view which val­ues em­bryos as much as new­borns.*

  I’d just be cu­ri­ous to see a writeup, es­pe­cially of the im­pact of con­tra­cep­tion ac­cess. Un­born hu­mans don’t fea­ture in tra­di­tional QALY-based effec­tive­ness analy­ses and I’d be in­ter­ested how the re­sults would change if they were in­cluded, even if at a dis­counted rate. I am not ex­pect­ing this to be a promis­ing area for most peo­ple in­ter­ested in effec­tive al­tru­ism.

  *An ex­cep­tion might be if you value pre-im­plan­ta­tion blas­to­cysts as much as born hu­mans, in which case your pri­or­ity could well be to ster­il­ize ev­ery­one. See also Toby Ord’s pa­per The Scourge.