SiebeRozendal

• SiebeRozendal 15 Jun 2019 15:16 UTC

  1 point

  on: What new EA pro­ject or org would you like to see cre­ated in the next 3 years?

  Anti-ex­tinc­tion website

  A web­site to de­ter rogue in­di­vi­d­u­als from de­vel­op­ing ideas to ‘de­stroy the world’ by show­ing that it’s just a very bad idea given their val­ues (e.g. A failed at­tempt will only make the world worse). The web­site would speci­fi­cally tar­get peo­ple vis­it­ing pro-ex­tinc­tion web­sites or search­ing the in­ter­net for things like ‘how to de­stroy the world’.

  This pro­ject would re­quire a very del­i­cate ap­proach and ideally would at­tract very lit­tle at­ten­tion from peo­ple un­in­ter­ested in de­stroy­ing the world. We don’t want to give any­one bad ideas. The text on it would re­quire care­ful phras­ing. The pro­ject prob­a­bly would need buy-in from knowl­edge­able stake­hold­ers (FHI, CSER, FLI, EAF, per­haps even the CIA?), a se­cure web­site, and very care­ful tar­get­ing.

• SiebeRozendal 15 Jun 2019 15:10 UTC

  3 points

  on: What new EA pro­ject or org would you like to see cre­ated in the next 3 years?

  Cause pri­ori­ti­za­tion tool

  Cause pri­ori­ti­za­tion is com­plex, and go­ing through it takes time. The rele­vant con­sid­er­a­tions are of­ten un­struc­tured and re­quire a lot of work to dis­cover. This leads so sub-op­ti­mal cause pri­ori­ti­za­tion. This can be done much bet­ter.

  An in­ter­ac­tive cause pri­ori­ti­za­tion tool could help peo­ple go through rele­vant con­sid­er­a­tions. I am imag­in­ing some­thing like the GPP’s flowchart, but where are asked only one ques­tion at a time.

  Fea­tures could in­clude:

  • Abil­ity to save progress and re­turn later

  • In­sight into how other users an­swered the question

  • Abil­ity to link to rele­vant arguments

  It should ideally be com­mu­nity-driven (the com­mu­nity can sug­gest con­sid­er­a­tions, and the com­mu­nity can vote on the im­por­tance of a par­tic­u­lar con­sid­er­a­tion).

• SiebeRozendal 15 Jun 2019 15:02 UTC

  1 point

  on: What new EA pro­ject or org would you like to see cre­ated in the next 3 years?

  Guided med­i­ta­tion record­ings for EAs

  I think med­i­ta­tion is a very in­ter­est­ing medium for learn­ing that we aren’t us­ing much yet. Guided med­i­ta­tion record­ings are highly scal­able and can be used to deal with spe­cific challenges that (effec­tive) al­tru­ists en­counter, like these by Holly Mor­gan, or to teach el­e­ments from the Re­plac­ing Guilt Series.

  I would be very in­ter­ested in see­ing some­one do a few of these to test the idea. Plau­si­bly, this only be­comes effec­tive once the EA has grown sub­stan­tially in size.

  Two other top­ics suit­ably for a guided med­i­ta­tion record­ing:

  • EA events (re­treats, con­fer­ences) can be very ex­cit­ing and very tax­ing, and the med­i­ta­tion could help to calm peo­ple down, re­fresh, pro­cess what’s go­ing on, set in­ten­tion or re­flect on one’s in­ten­tion, pri­ori­tize, etc.

  • De­mo­ti­vated mo­ments: some­times EA feels like run­ning against a wall—the effort just seems a com­plete waste, and it’d be nice to have some med­i­ta­tion to turn to. It could help peo­ple to deal healthily with set­backs by try­ing to learn from them, and then ac­cept­ing what­ever is.

• SiebeRozendal 15 Jun 2019 14:53 UTC

  1 point

  on: What new EA pro­ject or org would you like to see cre­ated in the next 3 years?

  EA e-learn­ing com­mu­nity:

  Prob­lem:

  EA’s are fo­cused a lot on learn­ing, but are do­ing this in­effi­ciently: they study alone, with­out a syl­labus, of­ten with­out peers, and with­out get­ting feed­back and eval­u­a­tion on their learn­ing out­comes.

  Solu­tion:

  There are many e-learn­ing tools, much more than just MOOC videos, that can be used, al­though I’m not fa­mil­iar with them. A learn­ing com­mu­nity could be cre­ated within an ex­ist­ing e-learn­ing plat­form.

  Peo­ple in EA are smart; a lot of value can be gained by fa­cil­i­tat­ing peo­ple help each other through care­fully de­signed learn­ing pro­grams. Q&A’s with EA re­searchers whose work is on the syl­labus could be ar­ranged to min­i­mize time needed for teach­ing. The e-learn­ing course/​com­mu­nity could also cre­ate struc­ture and ac­countabil­ity to get things done, as well as as­sist stu­dents in sig­nal­ling their EA-rele­vant knowl­edge and skills. For ex­am­ple, they could show they passed a course on ‘cre­at­ing good Fermi es­ti­mates’.

• SiebeRozendal 15 Jun 2019 14:42 UTC

  1 point

  on: What new EA pro­ject or org would you like to see cre­ated in the next 3 years?

  EA Hackathon

  A mul­ti­ple-day event in which peo­ple (mostly EA’s) work to solve a set of prob­lems. Th­ese prob­lems can be in any cause area, as long as a solu­tion could plau­si­bly be cost-effec­tive.

  Or­ga­ni­za­tions could send in prob­lem cases, or par­ti­ci­pants could iden­tify their own. In fact, this thread might already give a nice list of pos­si­ble prob­lems/​solu­tions to work on!

  The event can be made more com­plex by adding in rounds, which filter the good solu­tions from mediocre and bad solu­tions. In fur­ther rounds, promis­ing teams/​in­di­vi­d­u­als could be matched with a coach. The best few ideas get fund­ing to take the next step in solv­ing the prob­lem (e.g. cre­at­ing a min­i­mum vi­able product).

• SiebeRozendal 5 Jun 2019 15:13 UTC

  1 point

  in reply to: Luisa_Rodriguez's comment on: Can­di­date Scor­ing Sys­tem, Fifth Release

  Thanks for the an­swer. I look for­ward to see­ing your write-up and how the mod­els evolve!

  One thing that seems miss­ing in the cur­rent model is how smoke maps to famine (does the lo­ca­tion of the smoke mat­ter?), but per­haps as­sum­ing a lin­ear re­la­tion­ship be­tween amount of smoke and amount of ca­su­alties from famine is a good ap­prox­i­ma­tion.

• SiebeRozendal 5 Jun 2019 11:09 UTC

  4 points

  on: Can­di­date Scor­ing Sys­tem, Fifth Release

  Also on the topic of nu­clear war: I found it very sur­pris­ing to see that, ac­cord­ing to the re­port, the ex­pected deaths of a US-Rus­sia ex­change is lower (3 billion) than US-China or (5.5 billion) or Rus­sia-China (7 billion). This im­plies that the ex­is­ten­tial risk from nu­clear war is higher from a Rus­sia-China ex­change than from a US-Rus­sia ex­change. How should we trans­late these ex­pected deaths to ex­is­ten­tial risk?

  I found the Guessti­mate mod­els in­ter­est­ing, and figured out that this is largely be­cause a nu­clear bomb­ing of China is ex­pected to pro­duce the most smoke. But there’s a fac­tor in TSG1 (“to­tal smoke gen­er­ated by nu­clear weapons deto­nated in China (Tg)”) that I don’t un­der­stand, but that fac­tor drives much of the differ­ences in to­tal smoke gen­er­ated. In the US-China model it’s 1.58, and the Rus­sia-China model it’s 2.04, while a Rus­sian bomb­ing of the US has only a fac­tor of 0.664. Could you ex­plain what this fac­tor is and where it’s de­rived from?

  Also: the Rus­sia-China model has some bro­ken links due to an ex­tra “]” in MNO2.

• SiebeRozendal 5 Jun 2019 10:22 UTC

  1 point

  on: Can­di­date Scor­ing Sys­tem, Fifth Release

  About Nu­clear Non­pro­lifer­a­tion you write:

  Can­di­dates should be good at han­dling nu­clear pro­lifer­a­tion and threat­en­ing states such as North Korea [my em­pha­sis], Iran and Pak­istan. Th­ese states do not have the nu­clear ca­pac­ity to pose any ex­is­ten­tial risk, so they get less at­ten­tion here than the pri­mary pow­ers, but they com­pli­cate calcu­la­tions of in­ter­na­tional de­ter­rence and might in­di­rectly in­crease the risk of a ma­jor nu­clear war.

  What do you mean by ‘be­ing good at threat­en­ing’? It seems coun­ter­pro­duc­tive to threaten North Korea, as this could cause in­ter­na­tional con­flict, and back­fire by trig­ger­ing ac­tion from North Korea against South Korea and pos­si­bly Ja­pan. As ar­gued in this ar­ti­cle, it seems that the only op­tions are to per­suade China to in­crease eco­nomic pres­sure, or to ac­cept a nu­clear North Korea. North Korea has always been con­sis­tent and, given their goal, ra­tio­nal: they aim for nu­clear weapons for de­ter­rence and know that us­ing nukes will mean as­sured de­struc­tion.

• SiebeRozendal 4 Jun 2019 19:05 UTC

  1 point

  in reply to: HaukeHillebrandt's comment on: Crowd­fund­ing for Effec­tive Cli­mate Policy

  Ah, I as­sumed the lat­ter was a con­se­quence of the former be­cause they were in the same para­graph, my bad.

  How­ever, like Michael, I’m still a bit con­fused about the role ne­glect­ed­ness is play­ing in this anal­y­sis (and all other analy­ses). But don’t take that as crit­i­cism of your anal­y­sis. It of­ten seems that ne­glect­ed­ness and tractabil­ity (and scale) are used as in­de­pen­dent rea­sons to sup­port a par­tic­u­lar cause area or in­ter­ven­tion, rather than that they are used as a co­her­ent frame­work. It seems to me your ar­gu­ment would have been similarly strong if clean en­ergy R&D was not ne­glected—if you could just show that ad­di­tional spend­ing would have big benefits.

• SiebeRozendal 4 Jun 2019 11:18 UTC

  5 points

  in reply to: MichaelPlant's comment on: Crowd­fund­ing for Effec­tive Cli­mate Policy

  My un­der­stand­ing of these terms is roughly as fol­lows:

  Scale: gen­eral size of the prob­lem. Deter­mines the up­per bound of what can be achieved. Deter­min­ing the scale of a prob­lem is quite ar­bi­trary, be­cause how do you draw the bound­aries of ‘the prob­lem’ and when is it com­pletely solved?

  Tractabil­ity: de­ter­mines the av­er­age or global cost-effec­tive­ness if you don’t know where you are on the curve (i.e. if you don’t know how much of the prob­lem has been solved so far). Higher tractabil­ity means that the curve is steeper on av­er­age.

  Ne­glect­ed­ness: de­ter­mines the lo­ca­tion on the curve and gets you the marginal or lo­cal cost-effec­tive­ness. Be­cause we ex­pect the value curve to have diminish­ing re­turns, a good heuris­tic is ‘more ne­glected --> higher marginal value.’

  I think where EA’s go re­peat­edly wrong with the ap­pli­ca­tion of the model is that tractabil­ity and ne­glect­ed­ness get con­fused: tractabil­ity should re­fer to solv­ing the com­plete prob­lem. If it refers to marginal tractabil­ity, then it dou­ble counts the ne­glect­ed­ness con­sid­er­a­tion. The re­port seems to do this:

  Public clean en­ergy R&D is ne­glected: only $22 billion is spent per year globally. Many ad­vanced economies such as the US could unilat­er­ally in­crease this sub­stan­tially i.e. even with­out in­ter­na­tional co­or­di­na­tion—which makes this policy uniquely poli­ti­cally tractable.

  Here, ne­glect­ed­ness is taken as a rea­son for tractabil­ity, while it should be a rea­son for marginal cost-effec­tive­ness.

  The SNT-model is also much more helpful for fund­ing than for ca­reer choice, be­cause ne­glect­ed­ness has lin­ear im­pli­ca­tions for the value of ex­tra fund­ing, but more com­plex im­pli­ca­tions for an ex­tra per­son do­ing work. Some skills are not very use­ful in the early stage of a cause area or prob­lem, but be­come valuable only later. In gen­eral, I think per­sonal fit is very im­por­tant, and the SNT-model does not ac­count for it.

  I ac­tu­ally hope to write a longer post ex­plain­ing all this in more de­tail, in­clud­ing some nice vi­sual ex­pla­na­tion I have of the SNT-model.

  
  

• SiebeRozendal 4 Jun 2019 11:00 UTC

  4 points

  in reply to: jackmalde's comment on: Not get­ting car­ried away with re­duc­ing ex­tinc­tion risk?

  I think EA’s be­lieve that this is definitely pos­si­ble, most likely by the cre­ation of an al­igned su­per­in­tel­li­gence. That could re­duce x-risk to in­finites­si­mal lev­els, if there are no other in­tel­li­gent ac­tors that we could en­counter. I think the gen­eral strat­egy could be sum­ma­rized as ‘re­duce ex­tinc­tion risk as much as pos­si­ble un­til we can safely build and de­ploy an al­igned su­per­in­tel­li­gence, then let the su­per­in­tel­li­gence (dis)solve all other prob­lems’.

  After the cre­ation of an al­igned su­per­in­tel­li­gence, so­ciety’s re­sources could fo­cus on other prob­lems. How­ever, I think some peo­ple also think there are no other prob­lems any­more once there is an al­igned su­per­in­tel­li­gence: with su­per­in­tel­li­gence all the other prob­lems like an­i­mal suffer­ing are triv­ial to solve.

  But most peo­ple—in­clud­ing my­self—seem to not have given very much thought to what other prob­lems might still ex­ist in an era of su­per­in­tel­li­gence.

  If you be­lieve a strong ver­sion su­per­in­tel­li­gence is im­pos­si­ble this com­pli­cates the whole pic­ture, but you’d at least have to in­clude the con­sid­er­a­tion that in the fu­ture it is likely we have sub­stan­tially higher (in­di­vi­d­ual and/​or col­lec­tive) in­tel­li­gence.

• SiebeRozendal 4 Jun 2019 10:45 UTC

  5 points

  in reply to: Moses's comment on: Not get­ting car­ried away with re­duc­ing ex­tinc­tion risk?

  If hu­man­ity wipes it­self out, those wild an­i­mals are go­ing to con­tinue suffer­ing for­ever.

  Not for­ever. Only un­til the planet be­comes too hot to sup­port com­plex life (<1 billion years from now). Giv­ing that the uni­verse can sup­port life 1-100 trillion years, this is a rel­a­tively short amount of suffer­ing com­pared to what could be.

  And also only on our planet! Which is much less re­stricted than the suffer­ing that can spread if hu­man­ity re­mains al­ive. (Although, as I write in my own an­swer, I don’t think hu­man­ity would spread wild an­i­mals be­yond the so­lar sys­tem.)

• SiebeRozendal 4 Jun 2019 10:42 UTC

  1 point

  on: Not get­ting car­ried away with re­duc­ing ex­tinc­tion risk?

  Hey Jack, I think this is a great ques­tion and I ded­i­cate a por­tion of my MA philos­o­phy the­sis to this. Here are some gen­eral points:

  • It is likely that the ex­pected moral value of the fu­ture is dom­i­nated by fu­tures in which there is op­ti­miza­tion for moral (dis)value. Since we would ex­pect it to be much more likely there will be op­ti­miza­tion for value than for dis­value, the ex­pected moral value of the fu­ture seems pos­i­tive (un­less you ad­here to strict/​lex­i­cal nega­tive util­i­tar­i­anism). This claim de­pends on the differ­ence be­tween pos­si­ble wor­lds that are op­ti­mized for (dis)value vs. states that are sub­ject to other pres­sures, like com­pe­ti­tion: this differ­ence is not ob­vi­ously large (see Price of Anar­chy).

  • There seems sub­stan­tial con­ver­gence be­tween im­prov­ing the qual­ity of the long-term fu­ture and re­duc­ing ex­tinc­tion risk. Things that can bring hu­man­ity to ex­tinc­tion (su­per­in­tel­li­gence, virus, nu­clear win­ter, ex­treme cli­mate change) can also very bad for the long-term fu­ture of hu­man­ity if they do not lead to ex­tinc­tion. There­fore, re­duc­ing ex­tinc­tion risk also has very pos­i­tive effects on the qual­ity of the long-term fu­ture. Po­ten­tial suffer­ing risk from mis­al­igned AI is one can­di­date. In ad­di­tion, I think global catas­tro­phes, if they don’t lead to ex­tinc­tion, cre­ate a nega­tive tra­jec­tory change in ex­pec­ta­tion. Either be­cause civ­i­liza­tional col­lapse puts us on a worse tra­jec­tory, but the most likely out­come is what I call gen­eral global dis­rup­tion: civ­i­liza­tion doesn’t quite col­lapse, but think are shaken up a lot. From my the­sis:

  Should we ex­pect global dis­rup­tion to be (in ex­pec­ta­tion) good or bad for the value of the fu­ture? This is spec­u­la­tive, but Beck­stead lays out some rea­sons to ex­pect that global dis­rup­tion will put hu­man­ity most cer­tainly on a worse tra­jec­tory: it may re­verse so­cial progress, limit the abil­ity to ad­e­quately reg­u­late the de­vel­op­ment of dan­ger­ous tech­nolo­gies, open an op­por­tu­nity for au­thor­i­tar­ian regimes to take hold, or in­crease in­ter-state con­flict (Beck­stead, 2015). We can also ap­proach the is­sue ab­stractly: dis­rup­tion can be seen as in­ject­ing more noise into a pre­vi­ously more sta­ble global sys­tem, in­creas­ing the prob­a­bil­ity that the world set­tles into a differ­ent semi-sta­ble con­figu­ra­tion. If there are many more un­de­sir­able con­figu­ra­tions of the world than de­sir­able ones, in­creas­ing ran­dom­ness is more likely to lead to an un­de­sir­able state of the world. I am con­vinced that, un­less we are cur­rently in a par­tic­u­larly bad state of the world, global dis­rup­tion would have a very nega­tive effect (in ex­pec­ta­tion) on the value of the long-term fu­ture.

  • I find it un­likely that we would ex­port wild-an­i­mal suffer­ing be­yond our so­lar sys­tem. It takes a lot of time to move to differ­ent so­lar sys­tems, and I don’t think fu­ture civ­i­liza­tions will re­quire a lot of wilder­ness: it’s a very in­effi­cient use of re­sources. So I be­lieve the amount of suffer­ing is rel­a­tively small from that source. How­ever, I think some com­pet­i­tive dy­nam­ics be­tween digi­tal be­ings could cre­ate as­tro­nom­i­cal amounts of suffer­ing, and this could come about if we fo­cus only on re­duc­ing ex­tinc­tion risk.

  • Whether you want to fo­cus on the qual­ity of the fu­ture also de­pends on your moral views. Some peo­ple weigh pre­vent­ing fu­ture suffer­ing much more heav­ily than en­abling the cre­ation of fu­ture hap­piness. For them, part of the value of re­duc­ing ex­tinc­tion risk is taken away, and they will have stronger rea­sons to fo­cus on the qual­ity of the fu­ture.

  • I found the post by Brauner & Grosse-Holz and the post by Beck­stead most helpful. I know that Haydn Belfield (CSER) is cur­rently work­ing on a longer ar­ti­cle about the long-term sig­nifi­cance of re­duc­ing Global Catas­trophic Risks.

  In con­clu­sion, I think re­duc­ing ex­tinc­tion risk is a very pos­i­tive in terms of ex­pected value, even if one ex­pects the fu­ture to be nega­tive! How­ever, de­pend­ing on differ­ent pa­ram­e­ters, there might be bet­ter op­tions than fo­cus­ing on ex­tinc­tion risk. Can­di­dates in­volve par­tic­u­lar parts of moral cir­cle ex­pan­sion and suffer­ing risks from AI.

  I can send you the cur­rent draft of my the­sis in case you’re in­ter­ested, and will post it on­line once I have finished it.

• SiebeRozendal 4 Jun 2019 9:52 UTC

  13 points

  on: In­gre­di­ents for cre­at­ing dis­rup­tive re­search teams

  Re­search teams seem more likely to re­al­ize their full dis­rup­tive po­ten­tial if the re­searchers do not have to do any­thing but re­search and have easy ac­cess to all the re­sources they need.

  Note that Richard Ham­ming dis­agreed with this. He makes the point that re­stricted re­sources force cre­ativity, and the best break­throughs come through this cre­ativity. I think he has got a point, but I’m not sure what they all-things-con­sid­ered con­clu­sion should be. I think the ‘con­straints breed cre­ativity’ ap­plies more to the tools peo­ple work with, and other con­straints like teach­ing, ad­minis­tra­tive tasks, and grant ap­pli­ca­tions mostly waste time.

  Here’s an ex­cerpt form his fa­mous talk on re­search (for in­di­vi­d­ual re­searchers, not teams):

  What most peo­ple think are the best work­ing con­di­tions, are not. Very clearly they are not be­cause peo­ple are of­ten most pro­duc­tive when work­ing con­di­tions are bad. One of the bet­ter times of the Cam­bridge Phys­i­cal Lab­o­ra­to­ries was when they had prac­ti­cally shacks—they did some of the best physics ever.

  I give you a story from my own pri­vate life. Early on it be­came ev­i­dent to me that Bell Lab­o­ra­to­ries was not go­ing to give me the con­ven­tional acre of pro­gram­ming peo­ple to pro­gram com­put­ing ma­chines in ab­solute bi­nary. It was clear they weren’t go­ing to. But that was the way ev­ery­body did it. I could go to the West Coast and get a job with the air­plane com­pa­nies with­out any trou­ble, but the ex­cit­ing peo­ple were at Bell Labs and the fel­lows out there in the air­plane com­pa­nies were not. I thought for a long while about, ``Did I want to go or not?″ and I won­dered how I could get the best of two pos­si­ble wor­lds. I fi­nally said to my­self, ``Ham­ming, you think the ma­chines can do prac­ti­cally ev­ery­thing. Why can’t you make them write pro­grams?″ What ap­peared at first to me as a defect forced me into au­to­matic pro­gram­ming very early. What ap­pears to be a fault, of­ten, by a change of view­point, turns out to be one of the great­est as­sets you can have. But you are not likely to think that when you first look the thing and say, ``Gee, I’m never go­ing to get enough pro­gram­mers, so how can I ever do any great pro­gram­ming?″

  And there are many other sto­ries of the same kind; Grace Hop­per has similar ones. I think that if you look care­fully you will see that of­ten the great sci­en­tists, by turn­ing the prob­lem around a bit, changed a defect to an as­set. For ex­am­ple, many sci­en­tists when they found they couldn’t do a prob­lem fi­nally be­gan to study why not. They then turned it around the other way and said, ``But of course, this is what it is″ and got an im­por­tant re­sult. So ideal work­ing con­di­tions are very strange. The ones you want aren’t always the best ones for you.

  In the same talk he also crit­i­cizes the In­sti­tute for Ad­vanced Study, claiming that they only pro­duced good out­put be­cause they re­cruited the very best, but that coun­ter­fac­tu­ally these re­searchers would have pro­duced bet­ter out­put.

• SiebeRozendal 4 Jun 2019 9:02 UTC

  2 points

  on: EA Fo­rum Prize: Win­ners for April 2019

  I’d like to see ev­ery ma­jor re­search ar­ti­cle in EA find its way to the Fo­rum, ideally with sum­maries and au­tho­rial re­sponses. This re­port [on farmed fish welfare] is an ex­cel­lent ex­am­ple of how to bring in­ter­est­ing work to a broad au­di­ence.

  That would cer­tainly be great!

• SiebeRozendal 4 Jun 2019 8:28 UTC

  1 point

  in reply to: aarongertler's comment on: Con­sid­er­ing peo­ple’s hid­den mo­tives in EA outreach

  Re­lat­edly, there are the con­cepts of ‘un­cer­tainty’ and ‘in­se­cu­rity’. I think there’s a risk that un­cer­tainty as per­ceived, and per­haps even ex­pe­rienced, as in­se­cu­rity. In­ter­est­ingly, both con­cepts are trans­lated into one and the same word in Dutch! (“onzek­er­heid”)

  How­ever, I think stat­ing epistemic un­cer­tainty in a very pre­cise and con­fi­dent way (e.g. “I be­lieve X, and I am 60% cer­tain my hy­poth­e­sis is cor­rect”) can show meta-con­fi­dence and strong epistemics. I would rather learn to be con­vince while still com­mu­ni­cat­ing un­cer­tain­ties, than learn­ing to hide my epistemic un­cer­tainty.

  Also, ex­perts in any do­main face this challenge, and use­ful les­sons could be drawn from liter­a­ture on it, such as this pa­per (I only read the ab­stract, it seems use­ful).

• SiebeRozendal 4 Jun 2019 8:13 UTC

  4 points

  in reply to: 80000_Hours's comment on: Is EA Grow­ing? EA Growth Met­rics for 2018

  How big do you ex­pect that frac­tion to be? (Or: what per­centage of those num­bers do you ex­pect to be ‘real listen­ers’?)

• SiebeRozendal 21 May 2019 12:08 UTC

  1 point

  on: EA Fo­rum: Foot­notes are live, and other updates

  Per­haps not the best place to ask this, but:

  When I was go­ing through my ac­count op­tions to edit my bio, I saw there’s an op­tion to au­to­mat­i­cally hide com­ments gen­er­ated by GPT-2. Why are we not au­to­mat­i­cally mark­ing these com­ments as spam or at least as bot-gen­er­ated?

• SiebeRozendal 20 May 2019 11:39 UTC

  1 point

  in reply to: lukeprog's comment on: Which sci­en­tific dis­cov­ery was most ahead of its time?

  provide es­pe­cially strong ev­i­dence w.r.t. the dis­cov­ery coun­ter­fac­tu­als.

  I was a bit con­fused here. Do you mean the re­dis­cov­ery pro­vides ev­i­dence that the idea was ahead of its time, es­pe­cially when the re­dis­cov­ery was much later, be­cause we have an ac­tual coun­ter­fac­tual?

• SiebeRozendal 20 May 2019 11:29 UTC

  1 point

  in reply to: robm_73@hotmail.com's comment on: Med­i­ta­tion and Effec­tive Altruism

  That is nice! Here are some other ideas for guided med­i­ta­tion:

  • For dur­ing EA events: they can be very ex­cit­ing and very tax­ing, and the med­i­ta­tion could help to calm peo­ple down, re­fresh, pro­cess what’s go­ing on, etc.

  • For de­mo­ti­vated mo­ments: some­times EA feels like run­ning against a wall—the effort just seems a com­plete waste, and it’d be nice to have some med­i­ta­tion to turn to. It could help peo­ple deal healthily with set­backs by try­ing to learn from them, and then ac­cept­ing what­ever is.

  • There’s a bunch of other types that might be in­ter­est­ing to ex­plore via med­i­ta­tion, such as ob­ses­sion, suffer­ing (of oth­ers and one­self), ego­ism, (dis)con­nec­tion to EA’s and non-EA’s, value drift, mul­ti­ple things from the re­plac­ing guilt se­ries and other EA self-care stuff, etc.

  In gen­eral, I think med­i­ta­tion is a very in­ter­est­ing medium for learn­ing that we aren’t us­ing much yet. I would be very in­ter­ested in see­ing some­one do a few of these to test the idea, and I’d be in­ter­ested in get­ting some­one fund­ing to in­crease the vol­ume if the pi­lots are suc­cess­ful.