Milan_Griffes

• Milan_Griffes 2 Jan 2019 2:44 UTC

  3 points

  in reply to: Habryka's comment on: An­nounc­ing Pri­or­i­tyWiki: A Cause Pri­ori­ti­za­tion Wiki

  I would love to see pri­or­ity.wiki suc­ceed, es­pe­cially given its high-qual­ity de­sign & the broad sweep of con­tent already on there.

  This seems like a dan­ger­ous pe­riod where the pro­ject is likely to be aban­doned (more likely than dur­ing the post-launch hon­ey­moon, at least).

  My cur­rent guess is that pri­or­ity.wiki falls into a state of di­suse & dis­re­pair over the next 6 to 18 months, if no one in­ter­venes. Does that match the views of oth­ers?

  Per­haps tighter in­te­gra­tion with the Fo­rum and/​or effec­tivealtru­ism.org would help in­vi­go­rate the wiki. I’ll try to think of oth­ers things that could help too.

• Milan_Griffes 2 Jan 2019 2:32 UTC

  3 points

  in reply to: Habryka's comment on: An­nounc­ing Pri­or­i­tyWiki: A Cause Pri­ori­ti­za­tion Wiki

  Heh, just came here to ask the same.

  Cu­ri­ous if you have data on num­ber of con­trib­u­tors, num­ber of ed­its, num­ber of unique vis­i­tors, etc.

• Milan_Griffes 2 Jan 2019 0:09 UTC

  3 points

  in reply to: Nicole_Ross's comment on: EA Grants ap­pli­ca­tions are now open

  Here’s a click­able ver­sion of the “no­tify me about next round of EA Grants” form.

• Milan_Griffes 31 Dec 2018 22:06 UTC

  6 points

  on: What’s the Use In Physics?

  The EA recom­men­da­tion I tend to see for physi­cists is to not do physics.

  I’ve been un­der­stand­ing this to mean that un­der the cur­rent in­sti­tu­tional paradigm, more physics re­search on the mar­gin prob­a­bly isn’t very helpful.

  Achiev­ing a fun­da­men­tal break­through seems ob­vi­ously great (though hard to do in the cur­rent paradigm), and mean­ingfully re­form­ing the cur­rent paradigm would prob­a­bly be very high-value as well (though tricky to do).

  Robin Han­son has more to say on this (a):

  Pre­vi­ously, physics foun­da­tions the­o­rists were dis­ci­plined by a strong norm of re­spect­ing the the­o­ries that best fit the data. But with less data, the­o­rists have turned to mainly judg­ing pro­posed the­o­ries via var­i­ous stan­dards of “beauty” which ad­vo­cates claim to have in­ferred from past pat­terns of suc­cess with data. Ex­cept that these stan­dards (and their in­fer­ences) are mostly in­for­mal, change over time, differ greatly be­tween in­di­vi­d­u­als and schools of thought, and tend to la­bel as “ugly” our ac­tual best the­o­ries so far.

  See also “What does any of this have to do with physics?” (a).

• Milan_Griffes 31 Dec 2018 19:34 UTC

  5 points

  on: Cause pro­file: men­tal health

  I tend to think per­son-af­fect­ing views are the least-bad of the options

  Whether or not one holds a per­son-af­fect­ing view seems like a big crux for pri­ori­tiz­ing men­tal health (es­pe­cially if the al­tru­ism cas­cade con­sid­er­a­tion doesn’t seem com­pel­ling).

  If it’s quick, could you say a bit more about why you hold a per­son-af­fect­ing view?

  (My guess is that many fo­rum read­ers fol­low Nick Beck­stead in think­ing that a to­tal­ist view makes more sense, so they won’t find men­tal health com­pel­ling to the ex­tent that it rests on a per­son-af­fect­ing view.)

• Milan_Griffes 31 Dec 2018 19:27 UTC

  2 points

  in reply to: Milan_Griffes's comment on: Cause pro­file: men­tal health

  Analo­gous “al­tru­ism cas­cade” ar­gu­ments could be made for in­ter­ven­tions like bed nets & GiveDirectly, though my in­tu­ition is that the cas­cade would be stronger for men­tal health in­ter­ven­tions as they more di­rectly at­tack un­hap­piness. (Re­lated GiveWell dis­cus­sion on flow-through effects here.)

  I don’t see analo­gous cas­cades com­ing out of x-risk pre­ven­tion or fun­da­men­tal philo­soph­i­cal re­search.

• Milan_Griffes 31 Dec 2018 19:23 UTC

  5 points

  on: Cause pro­file: men­tal health

  Thanks, Michael, for con­tin­u­ing to push for­ward on this im­por­tant front!

  Another rea­son that men­tal health in­ter­ven­tions seem promis­ing: im­prov­ing some­one’s men­tal health may in­crease their abil­ity to help to oth­ers. This could cre­ate large in­di­rect im­pacts, where peo­ple who’s men­tal health im­proves are able to bring their newly un­locked en­ergy to bear on their own al­tru­is­tic efforts (which im­pact other peo­ple, who per­haps also go on to help oth­ers in a cas­cade of al­tru­ism).

  This ar­gu­ment as­sumes that peo­ple tend to grow (at least some­what) more al­tru­is­tic as they be­come hap­pier. This has been true in my ex­pe­rience & ob­ser­va­tions; I’m not sure how far it gen­er­al­izes.

• Milan_Griffes 31 Dec 2018 19:02 UTC

  1 point

  in reply to: Milan_Griffes's comment on: 2018 AI Align­ment Liter­a­ture Re­view and Char­ity Comparison

  I sup­pose this could be rol­led into the Google Brain sec­tion, as it looks like most Distill con­trib­u­tors have a Google af­fili­a­tion.

• Milan_Griffes 31 Dec 2018 18:53 UTC

  2 points

  in reply to: Agrippa's comment on: EA Ho­tel Fundraiser 1: the story

  (A) is a great point.

• Milan_Griffes 31 Dec 2018 18:49 UTC

  5 points

  on: 2018 AI Align­ment Liter­a­ture Re­view and Char­ity Comparison

  Cu­ri­ous why Distill wasn’t in­cluded.

  Their stuff on in­ter­pretabil­ity seems like it has im­pli­ca­tions for al­ign­ment, and their work seems high-qual­ity in gen­eral.

• Milan_Griffes 31 Dec 2018 18:45 UTC

  2 points

  in reply to: rohinmshah's comment on: 2018 AI Align­ment Liter­a­ture Re­view and Char­ity Comparison

  Thanks, I was also cu­ri­ous about how you sourced the newslet­ter :-)

  Why do you think Twit­ter is de­grad­ing?

• Milan_Griffes 29 Dec 2018 20:33 UTC

  1 point

  on: 2018 AI Align­ment Liter­a­ture Re­view and Char­ity Comparison

  Thanks, I found this very helpful!

  What pro­cess do you use to stay on top of the new liter­a­ture as it comes out?

  I have a rough model of what to do to track or­ga­ni­za­tional out­put: sign up for newslet­ters & RSS feeds, check their web­sites oc­ca­sion­ally, ask them if I’ve missed any­thing near the end of the year.

  I have no idea what to do to track the work com­ing out of academia (i.e. the stuff in your “Other Re­search” sec­tion) - arxiv seems like a morass to nav­i­gate. How do you stay on top of that?

• Milan_Griffes 16 Dec 2018 22:43 UTC

  3 points

  on: The ex­pected value of ex­tinc­tion risk re­duc­tion is positive

  Cu­ri­ous how you’re think­ing about efforts that are in­tended to re­duce x-risk but in­stead end up in­creas­ing it.

  e.g. pub­lic-fac­ing aerosol in­jec­tion re­search:

  Given this strate­gic land­scape, the effects of call­ing at­ten­tion to strato­spheric aerosol in­jec­tion as a cause are un­clear. It’s pos­si­ble that fur­ther pub­lic-fac­ing work on the in­ter­ven­tion re­sults in in­ter­na­tional agree­ments gov­ern­ing the use of the tech­nol­ogy. This would most likely be a re­duc­tion in ex­is­ten­tial risk along this vec­tor.

  How­ever, it’s also pos­si­ble that fur­ther pub­lic-fac­ing work on aerosol in­jec­tion makes the tech­nol­ogy more dis­cov­er­able, re­veal­ing the tech­nol­ogy to de­ci­sion-mak­ers who were pre­vi­ously ig­no­rant of its promise. Some of these de­ci­sion-mak­ers might be in­clined to pur­sue re­search pro­grams aimed at de­vel­op­ing a strato­spheric aerosol in­jec­tion ca­pa­bil­ity, which would most likely in­crease ex­is­ten­tial risk along this vec­tor.

• Milan_Griffes 16 Dec 2018 18:40 UTC

  1 point

  on: EA Fo­rum Prize: Win­ners for Novem­ber 2018

  ...fol­lowed by a runoff vote to re­solve ties.

  Is the runoff vote also ap­proval vot­ing?

• Milan_Griffes 16 Dec 2018 18:27 UTC

  4 points

  on: EA Fo­rum Prize: Win­ners for Novem­ber 2018

  A lit­tle strange that a post’s karma isn’t part of the prize eval­u­a­tion pro­cess.

  (I guess it is im­plic­itly, would be in­ter­est­ing to see an ex­plicit karma com­po­nent.)

• Milan_Griffes 16 Dec 2018 18:26 UTC

  5 points

  on: EA Fo­rum Prize: Win­ners for Novem­ber 2018

  I re­ally like the prize idea as a method of con­tent cu­ra­tion – I wasn’t plan­ning to read San­jay’s Cool Earth post (be­cause I didn’t un­der­stand what it was about & it didn’t seem rele­vant to my in­ter­ests at first brush), but now I will.

  Thanks all for mak­ing this hap­pen :-)

• Milan_Griffes 16 Dec 2018 17:26 UTC

  1 point

  in reply to: Robert_Wiblin's comment on: Giv­ing more won’t make you hap­pier

  For my­self, I would re­gard those gains to be suffi­ciently small that I would think it ir­ra­tional for an ego­ist to fo­cus much of their at­ten­tion on earn­ing more money at that point, rather than fos­ter­ing strong re­la­tion­ships, a sense of pur­pose, or im­prov­ing their self-talk.

  I agree with this.

  The main take­away I’m push­ing here is some­thing like:

  “After a cer­tain point, mak­ing more money has se­vere diminish­ing re­turns re: your hap­piness, as does donat­ing lots of money.

  So don’t lean on mak­ing lots of money to make you happy, and don’t lean on giv­ing away lots of money to make you happy.”

  There’s a temp­ta­tion to use “donate a lot of money to effec­tive causes” to scratch the “sense of pur­pose” itch, which I don’t think works very well (due to the diminish­ing re­turns).

• Milan_Griffes 16 Dec 2018 17:18 UTC

  4 points

  in reply to: Robert_Wiblin's comment on: Giv­ing more won’t make you hap­pier

  Whether this counts as “ex­tra in­come con­tin­u­ing to af­fect hap­piness quite a bit” or “ex­tra in­come not af­fect­ing hap­piness that much” I guess is for read­ers to judge.

  I no­tice I have some difficulty think­ing through the im­pli­ca­tions of a 0.5 bump in life satis­fac­tion on a 0.0 to 10.0 scale, es­pe­cially when the 0.5 in­crease is in ag­gre­gate across an en­tire life­time.

  On one view, 0.5 doesn’t seem like that much. “7.5 in­stead of 8.0? That’s a neg­ligible effect. Once you’re at 7.5 life-satis­fac­tion-wise, time to fo­cus on other things.”

  On an­other view, the 0.5 bump is quite a lot. If 10.0 on the scale is “most satis­fy­ing life pos­si­ble”, go­ing from 7.5 to 8.0 could be a big frickin’ deal. Also could be a big deal if the 0.5 bump cashes out to some­thing like “one less ter­rible day per month, for the rest of your life”.

  This con­sid­er­a­tion is prob­a­bly dom­i­nated by mea­sure­ment prob­lems though. When I sub­jec­tively as­sess my life satis­fac­tion, I have trou­ble dis­cern­ing the differ­ence be­tween a 7 and an 8 on a 0-10 scale (though I’m bench­mark­ing on 10 be­ing “best out of the ways my life has tended to go”, not “most satis­fy­ing life pos­si­ble”).

  I’ve started us­ing a 0-5 scale be­cause of this gran­u­lar­ity con­sid­er­a­tion. It’s much eas­ier for me to tell apart the differ­ence be­tween 3 and 4 on a 0-5 scale than it is to tell apart 7 and 8 on a 0-10 scale.

  This is all to say that a 0.5 bump on a 0.0-10.0 scale might not be sub­jec­tively de­tectable at all to most peo­ple. (Though a 0.5 in-ag­gre­gate effect could still cash out to large sub­jec­tive gains for many peo­ple.)

• Milan_Griffes 16 Dec 2018 16:37 UTC

  7 points

  on: New edi­tion of “Ra­tion­al­ity: From AI to Zom­bies”

  I think pa­per books are an un­der-ap­pre­ci­ated tech­nol­ogy these days; very ex­cited that pa­per ver­sions of Ra­tion­al­ity are com­ing out!

  Thanks for mak­ing this hap­pen!

• Milan_Griffes 11 Dec 2018 18:49 UTC

  1 point

  in reply to: Bluefalcon's comment on: Giv­ing more won’t make you hap­pier

  I think you’re ap­proach­ing an in­ter­est­ing point on the earn­ings curve. Prob­a­bly mak­ing $200k and giv­ing away $100k would be more satis­fy­ing than earn­ing $100k and giv­ing away $1,000 (wildly spec­u­lat­ing here).

  I’m not sure that mak­ing $100k and giv­ing away $50k of it would be more satis­fy­ing than earn­ing $50k and giv­ing away $1,000 though. After re­flect­ing on it for a minute, I think it prob­a­bly would be, but nowhere near 50x. And you’d be leav­ing a lot of other happy-mak­ing pos­si­bil­ities on the table if you gave away $50k out of $100k (which seems less true for giv­ing away $100k out of $200k).