Larks

• Larks 3 Apr 2020 17:39 UTC

  8 points

  on: Large EA House in Berkeley is look­ing for new housemates

  Per­haps you might want to share your quaran­tine strat­egy? I can imag­ine that be­ing im­por­tant to peo­ple right now.

• Larks 12 Mar 2020 2:01 UTC

  21 points

  in reply to: Milan_Griffes's comment on: What are the key on­go­ing de­bates in EA?

  I’m sure we can quib­ble about how the “Men­tal Health” should map to the “Psychedelics” cat­e­gory, though it seems clear that psychedelics are one of the most promis­ing de­vel­op­ments in men­tal health in the last few decades (break­through ther­apy des­ig­na­tion from the FDA and all that).

  If we as­sume half of the above con­sid­ered psychedelics to be in the men­tal health bucket …

  This does not seem like a quib­ble to me at all. It seems ‘clear’ to you but this is by no means the case for most peo­ple. I would hap­pily bet that well un­der half of those peo­ple were think­ing psychedelics when they said men­tal health.

• Larks 26 Feb 2020 19:20 UTC

  3 points

  on: Em­pa­thy and com­pas­sion to­ward other species de­crease with evolu­tion­ary di­ver­gence time

  Thanks for shar­ing, this was very in­ter­est­ing.

• Larks 16 Feb 2020 23:24 UTC

  3 points

  on: FHI Re­port: The Wind­fall Clause: Distribut­ing the Benefits of AI for the Com­mon Good

  B.2. “The Wind­fall Clause will shift in­vest­ment to com­pet­i­tive non-sig­na­tory firms.”
  
  The con­cern here is that, when mul­ti­ple firms are com­pet­ing for wind­fall prof­its, a firm bound by the Clause will be at a com­pet­i­tive dis­ad­van­tage be­cause un­bound firms could offer higher re­turns on new cap­i­tal. That is, in­vestors would pre­fer firms that are not sub­ject to a “tax” on their prof­its in the form of the Wind­fall Clause. This is es­pe­cially bad be­cause it could mean that more proso­cial firms (i.e., ones that have signed the Clause) would be at a dis­ad­van­tage to non-sig­na­tory firms, mak­ing a proso­cial “win­ner” of an AI de­vel­op­ment race less likely.238
  
  This is a valid con­cern which war­rants care­ful con­sid­er­a­tion. Our cur­rent best model for how to ad­dress this is that the Clause could com­mit (or at least al­low for the op­tion of) dis­tri­bu­tions of equity,* in­stead of cash. This could ei­ther take the form of stock op­tions or con­tin­gent con­vert­ible bonds. This avoids the con­cern iden­ti­fied by al­low­ing firms to, for ex­am­ple, is­sue new, preferred shares which would have su­pe­rior claim to wind­fall prof­its com­pared to donees. This sig­nifi­cantly diminishes the con­cern that the Clause would dilute the value of new shares is­sued in the com­pany and al­lows the bound firm to raise cap­i­tal un­en­cum­bered by debt owed un­der the Clause.† Notably, firm man­age­ment would still have fi­du­ciary du­ties to­wards stock-hold­ing wind­fall donees.

  I agree that the prob­lem (that in­vestors will pre­fer to in­vest in non-sig­na­to­ries, and hence it will re­duce the like­li­hood of pro-so­cial firms win­ning, if pro-so­cial firms are more likely to sign) does seem like a cred­ible is­sue. I found the de­scrip­tion of the pro­posed solu­tion rather con­fus­ing how­ever. Given that I worked as an equity an­a­lyst for five years, I would be sur­prised if many other read­ers could un­der­stand it!

  Here are my thoughts on a cou­ple of pos­si­ble ver­sions of what you might be get­ting at- apolo­gies if you ac­tu­ally in­tended some­thing else al­to­gether.
  
  1) The clause will al­low the com­pany to make the re­quired pay­ments in stock rather than cash.
  
  Un­for­tu­nately this doesn’t re­ally make much differ­ence, be­cause it is very easy for com­pa­nies to al­ter this bal­ance them­selves. Con­sider that a com­pany which had to make a $1 billion cash pay­ment could fund this by is­su­ing $1 billion worth of stock; con­versely a com­pany which had to is­sue stock to the fund could neu­tral­ise the effect on their share count by pay­ing cash to buy back $1 billion worth of or­di­nary shares. This is the same rea­son why div­i­dends are es­sen­tially iden­ti­cal to share buy­backs.
  
  2) The clause will al­low sub­se­quent fi­nanc­ing to be raised that is se­nior to the wind­fall clause claim, and thus still at­trac­tive to in­vestors.
  
  ‘Se­nior’ does not mean ‘bet­ter’ - it sim­ply means that you have pri­or­ity in the event of bankruptcy. How­ever, the clause is already ju­nior to all other obli­ga­tions (be­cause a bankrupt firm will be mak­ing ~0% of GDP in profit and hence have no clause obli­ga­tions), so this doesn’t re­ally seem like it makes much differ­ence. The is­sue is dilu­tion in sce­nar­ios when the com­pany does well, which is when the most ju­nior claims (typ­i­cally com­mon equity, but in this case ac­tu­ally the clause) perform best.
  
  The fun­da­men­tal rea­son these two ap­proaches will not work is that the value of an in­vest­ment is de­ter­mined by the net pre­sent value of fu­ture cash­flows (and their prob­a­bil­ity dis­tri­bu­tion). Given that the clause is in­tended to have a fixed im­pact on these flows (as laid out in II.A.2), the im­pact on firm val­u­a­tion is also rather fixed, and there is rel­a­tively lit­tle that clever fi­nan­cial en­g­ineer­ing can do about it.
  
  3) The clause will have claim only to prof­its at­tributable to the ex­ist­ing shares at the time of the sign­ing on. Any sub­se­quent equity will have a claim on prof­its un­en­cum­bered by the clause. For ex­am­ple, if a com­pany with 80 shares signs on to the clause, then is­sues 10 more shares to the mar­ket, the max­i­mum % of prof­its that would be owed is 50%*80/​(80+10) = 44.4%
  
  This would in­deed avoid most of the prob­lems in at­tract­ing new cap­i­tal (save only the fear that a man­age­ment team will­ing to screw over their pre­vi­ous in­vestors will do so to you in the fu­ture, which is some­thing in­vestors think about a lot).
  
  How­ever, it would also largely un­der­mine the clause by be­ing easy to evade due to the fun­gi­bil­ity of cap­i­tal. Con­sider a new startup, founded by three guys in a base­ment, that signs the clause. Over the next few years they will raise many rounds of VC, even­tu­ally giv­ing up the ma­jor­ity of the com­pany, all ex­cluded from the clause. Ad­di­tion­ally, they pay them­selves and em­ploy­ees in stock or stock op­tions, which are also ex­empt from the clause. Even­tu­ally they IPO, hav­ing suc­cess­fully diluted the clause-af­fected shares to ~1%. In or­der to finish the job, they then is­sue some ad­di­tional new equity and use the pro­ceeds to buy back the origi­nal shares.

  
  

  One in­ter­est­ing point on the other side, how­ever, is the cu­ri­ous ten­dency for tech in­vestors to ig­nore dilu­tion. Many com­pa­nies will ex­clude stock-based-comp from their ad­justed earn­ings, and an­a­lysts/​in­vestors are of­ten will­ing to go along with this, say­ing “oh but it’s a non-cash ex­pense”. Fur­ther­more, SBC is ex­cluded from Free Cash Flow, which is the preferred met­ric for many tech in­vestors. So it is pos­si­ble that (for a while) in­vestors would sim­ply ig­nore it.
  

• Larks 15 Feb 2020 3:25 UTC

  19 points

  on: FHI Re­port: The Wind­fall Clause: Distribut­ing the Benefits of AI for the Com­mon Good

  Thanks very much for shar­ing this. It is nice to see some in­no­va­tive think­ing around AI gov­er­nance.

  I have a bunch of differ­ent thoughts, so I’ll break them over mul­ti­ple com­ments. This one mainly con­cerns the in­cen­tive effects.

  > C.2. “The Wind­fall Clause op­er­ates like a pro­gres­sive cor­po­rate in­come tax, and the ideal cor­po­rate in­come tax rate is 0%.”
  
  > Some com­men­ta­tors ar­gue that the ideal cor­po­rate tax rate is 0%. One com­mon ar­gu­ment for this is that cor­po­rate in­come tax is not as pro­gres­sive as its pro­po­nents think be­cause cor­po­rate in­come is ul­ti­mately des­tined for share­hold­ers, some of whom are wealthy, but many of whom are not. Bet­ter, then, to tax those wealthy share­hold­ers more di­rectly and let cor­po­rate prof­its flow less im­peded to poorer ones. Ad­di­tion­ally, cur­rent cor­po­rate taxes ap­pear to bur­den both share­hold­ers and, to a lesser ex­tent, work­ers.”

  I think this is a bit of a straw­man. While it is true that many peo­ple don’t un­der­stand tax in­ci­dence and falsely as­sume the bur­den falls en­tirely on share­hold­ers rather than work­ers and con­sumers, the main ar­gu­ment for the op­ti­mal­ity of a 0% cor­po­rate tax rate is Cham­ley-Judd (see for ex­am­ple here) and re­lated re­sults. (There are some in­for­mal de­scrip­tions of the re­sult here and here.) The ar­gu­ment is about dis­in­cen­tives to in­vest re­duc­ing long-run growth and thereby mak­ing ev­ery­one poorer, not a short-term dis­tri­bu­tional effect. (The stan­dard counter-ar­gu­ment to Cham­ley Judd, as far as I know, is to effec­tively ap­ply lots of tem­po­ral dis­count­ing, but this is not available to longter­mist EAs).
  
  This is sort of cov­ered in B.1., but I do not think the re­sponses are very per­sua­sive. The main re­sponse is rather glib:

  Fur­ther, by cap­ping firm obli­ga­tions at 50% of marginal prof­its, the Clause leaves room for in­no­va­tion to be in­vested in even at in­cred­ibly high profit lev­els.231

  There are a lot de­sir­able in­vest­ments which would be ren­dered un­eco­nomic. The fact that some in­vest­ment will con­tinue at a re­duced level does not mean that miss­ing out on the other for­gone pro­jects is not a great cost! For ex­am­ple, a 20% pre-tax re­turn on in­vest­ment for a mod­er­ately risky pro­ject is highly at­trac­tive—but af­ter ~25% cor­po­rate taxes and ~50% wind­fall clause, this is a mere 5% re­turn* - al­most definitely be­low their cost of cap­i­tal, and hence so­ciety will prob­a­bly miss out on the benefits. Ci­ta­tion 231, which seems like it should be do­ing most of the work here, in­stead refer­ences a pass­ing com­ment in a pop-sci book about in­di­vi­d­ual taxes:

  There’s also an ar­gu­ment that a big part of the very high earn­ings of many ‘su­per­stars’ are also rents. Th­ese ques­tions turn on whether most pro­fes­sional ath­letes, CEOs, me­dia per­son­al­ities, or rock stars are gen­uinely mo­ti­vated by the ab­solute level of their com­pen­sa­tion verses the rel­a­tive com­pen­sa­tion, their fame, or their in­trin­sic love of their work.

  But cor­po­ra­tions are much less mo­ti­vated by fame and love of their work than in­di­vi­d­u­als, so this does not seem very rele­vant, and fur­ther­more it does not ad­dress the in­ter-tem­po­ral is­sue which is the main ob­jec­tion to cor­po­ra­tion taxes.
  
  I also think the sub-re­sponses are un­satis­fy­ing. You men­tion that the clause will be vol­un­tary:

  > Firstly, we ex­pect firms to agree to the Clause only if it is largely in their self-interest

  But this does not mean it won’t re­duce in­cen­tives to in­no­vate. Firms can ra­tio­nally take ac­tions that re­duce their fu­ture in­no­va­tion (e.g. sel­l­ing off an in­no­va­tive but risky di­vi­sion for a good price). A firm might vol­un­tar­ily sign up now, when the ex­pected cost is low, but then see their in­cen­tives dra­mat­i­cally cur­tailed later, when the cost is large. Fur­ther­more, firms can vol­un­tar­ily but ir­ra­tionally re­duce their in­cen­tives to in­no­vate—for ex­am­ple a CEO might sign up for the clause be­cause he per­son­ally got a lot of pos­i­tive press for do­ing so, even at the cost of the firm.

  Ad­di­tion­ally, by pub­li­cis­ing this idea you are chang­ing the land­scape—a firm which might have seen no rea­son to sign up might now feel pres­sured to do so af­ter a pub­lic cam­paign, even though their sub­mis­sion is ‘vol­un­tary’.
  
  The re­port then goes on to dis­cuss ex­ter­nal­ities:

  > Se­condly, un­bri­dled in­cen­tives to in­no­vate are not nec­es­sar­ily always good, par­tic­u­larly when many of the po­ten­tial down­sides of that in­no­va­tion are ex­ter­nal­ized in the form of pub­lic harms. The Wind­fall Clause at­tempts to in­ter­nal­ize some of these ex­ter­nal­ities to the sig­na­tory, which hope­fully con­tributes to steer­ing in­no­va­tion in­cen­tives in ways that min­i­mize these nega­tive ex­ter­nal­ities and com­pen­sate their bear­ers.

  Here you ap­prov­ingly cite Seb’s pa­per, but I do not think it sup­ports your point at all. Firms have both pos­i­tive and nega­tive ex­ter­nal­ities, and caus­ing them to in­ter­nal­ise them re­quires tai­lored solu­tions—e.g. a car­bon tax. ‘Be­ing very prof­itable’ is not a nega­tive ex­ter­nal­ity, so a tax on prof­its is not an effec­tive way of min­imis­ing nega­tive ex­ter­nal­ities. Similarly, the Mal­i­cious Use pa­per is mainly about spe­cific bad use cases, rather than size qua size be­ing un­de­sir­able. More­over, size has lit­tle to do with Seb’s ar­gu­ment, which is about es­ti­mat­ing the costs of spe­cific re­search pro­pos­als when ap­ply­ing for grants.

  Fi­nally, one must con­sider that un­der wind­fall sce­nar­ios the gains from in­no­va­tion are already sub­stan­tial, sug­gest­ing that globally it is more im­por­tant to fo­cus on dis­tri­bu­tion of gains than in­cen­tiviz­ing ad­di­tional in­no­va­tion.

  I strongly dis­agree with this non-se­quitur. The fact that we have achieved some level of ma­te­rial suc­cess now doesn’t mean that the fu­ture op­por­tu­nity isn’t very large. Again, Cham­ley-Judd is the clas­sic re­sult in the space, sug­gest­ing that it is never ap­pro­pri­ate to tax in­vest­ment for dis­tri­bu­tional pur­poses—if the lat­ter must be done, it should be done with in­di­vi­d­ual-level con­sump­tion/​in­come tax­a­tion. This should be es­pe­cially clear to EAs who are aware of the as­tro­nom­i­cal waste of po­ten­tially for­go­ing or de­lay­ing growth.
  
  Else­where in the doc­u­ment you do hint at an­other re­sponse—namely that by adopt­ing the clause, com­pa­nies will help avoid fu­ture tax­a­tion (though I am scep­ti­cal):

  > A Wind­fall Clause could build good­will among the pub­lic, damp­en­ing harm­ful pub­lic an­tag­o­nism for a small (ex­pected) cost. Govern­ments may be less likely to ex­ces­sively tax or ex­pro­pri­ate firms com­mit­ted to pro­vid­ing a pub­lic good through the Wind­fall Clause.

  and

  > How­ever, from a pub­lic and em­ployee re­la­tions per­spec­tive, the Clause may be more ap­peal­ing than tax­a­tion be­cause the Clause is a co­op­er­a­tive, proac­tive, and su­pereroga­tory ac­tion. So, to the ex­tent that the Wind­fall Clause merely re­places tax­a­tion, the Wind­fall Clause con­fers rep­u­ta­tional benefits onto the sig­na­tory at no ad­di­tional cos

  How­ever, it seems that the doc­u­ment equiv­o­cates on whether or not the clause is to re­duce taxes, as el­se­where in the doc­u­ment you deny this:

  > the Wind­fall Clause is not in­tended to be a sub­sti­tute for tax­a­tion schemes. We also note that, as a pri­vate con­tract, the Wind­fall Clause can­not su­per­sede tax­a­tion. Thus, if a state wants to tax the wind­fall, the Clause is not in­tended to stop it. In­deed, tax­a­tion efforts that broadly al­ign with the goals and de­sign prin­ci­ples of the Wind­fall Clause are highly desirable

  
  

  \* for clar­ity of ex­po­si­tion I am as­sum­ing the dona­tion is not tax de­ductible, but the point is not dra­mat­i­cally al­tered if it is.

• Larks 15 Feb 2020 2:35 UTC

  6 points

  on: Short-Term AI Align­ment as a Pri­or­ity Cause

  Many of these ad­van­tages (e.g. al­igned recom­menders push­ing peo­ple to­wards longter­mism, or an­i­mal rights) seem more like al­ign­ing recom­menders with our val­ues than any neu­tral ac­count of al­ign­ment. It seems than any ide­ol­ogy could similarly claim that al­igned recom­menders are im­por­tant for in­tro­duc­ing peo­ple to liber­tar­i­anism/​so­cial­ism/​con­ser­vatism/​fem­i­nism etc. In con­trast, this prob­a­bly isn’t the best in­ter­ests of the viewer—e.g. your av­er­age om­nivore prob­a­bly doesn’t want to be recom­mended videos about an­i­mal suffer­ing.

• Larks 13 Feb 2020 18:07 UTC

  2 points

  on: Do im­pact cer­tifi­cates help if you’re not sure your work is effec­tive?

  I’m con­cerned that split­ting the “vote” be­tween these two meth­ods will do harm to the com­mu­nity’s abil­ity to de­cide what types of work are good.

  Could you go into de­tail about why you think this would be bad? Typ­i­cally when you are un­cer­tain about some­thing it is good to have mul­ti­ple (semi-) in­de­pen­dent in­di­ca­tors, as you can get a more ac­cu­rate over­all im­pres­sion by com­bin­ing the two in some way.

• Larks 13 Feb 2020 17:54 UTC

  4 points

  in reply to: Habryka's comment on: Re­quest for Feed­back: Draft of a COI policy for the Long Term Fu­ture Fund

  Makes sense, thanks for clar­ify­ing!

• Larks 9 Feb 2020 23:38 UTC

  23 points

  on: Re­quest for Feed­back: Draft of a COI policy for the Long Term Fu­ture Fund

  Thanks for writ­ing this up for com­ment.

  Quick pos­si­ble over­sight—I didn’t see any dis­cus­sion of re­cusal be­cause the fund mem­ber is em­ployed or re­ceives funds from the po­ten­tial grantee? Sorry if I just mis­read! The clos­est I saw was this:

  A very close friend or part­ner of a fund mem­ber is em­ployed, re­ceiv­ing funds from, or has some kind of other di­rectly de­pen­dent re­la­tion­ship to the po­ten­tial grantee

  Right now I as­sume this would mainly ap­ply to you (CFAR) and pos­si­bly Alex.

  
  

  Separately, you men­tioned OpenPhil’s policy of (non-) dis­clo­sure as an ex­am­ple to em­u­late. I strongly dis­agree with this, for two rea­sons.

  Firstly, I think OpenPhil’s policy is bad. They en­acted this policy as part of their gen­eral move­ment to­wards se­crecy, but the ac­tual rea­sons they de­scribed for shar­ing less de­tails about their eval­u­a­tion (i.e. is­sues are too com­plex, don’t want to aid hos­tile ac­tors etc.) do not seem to be that rele­vant to not dis­clos­ing con­flicts. Cer­tainly, OpenPhil’s policy of non-dis­clo­sure makes me trust their work sig­nifi­cantly less now as I have to as­sume there is a sig­nifi­cant chance any given de­ci­sion was un­fairly bi­ased.

  Se­condly, there are sig­nifi­cant differ­ences be­tween OpenPhil and you guys. In par­tic­u­lar, OpenPhil’s main job is ad­vis­ing a very small num­ber of in­di­vi­d­u­als, who (I pre­sume) have ac­cess to many pri­vate de­tails that they do not need to make pub­lic. Ad­di­tion­ally, those in­di­vi­d­u­als have a con­sid­er­able amount of in­fluence over OpenPhil. In the case of the LTFF, how­ever, donors are re­li­ant on pub­lic dis­clo­sure in or­der to be able to eval­u­ate the fund. It is like the differ­ence be­tween a pri­vate com­pany (who have al­most no pub­lic dis­clo­sure re­quire­ments) and a pub­lic one (who have a lot of dis­clo­sure re­quire­ments).

• Larks 9 Feb 2020 23:38 UTC

  16 points

  in reply to: Habryka's comment on: Re­quest for Feed­back: Draft of a COI policy for the Long Term Fu­ture Fund

  One of the things that I am most con­cerned about if you were to just move to­wards re­cusal, is you just end up in a situ­a­tion where by ne­ces­sity the other fund mem­bers just have to take the re­cused per­son’s word for the grant be­ing good (or you pass up on all the most valuable grant op­por­tu­ni­ties).

  It also seems that the re­cusal be­ing dis­cussed is quite weak. In my limited ex­pe­rience, re­cusal means to­tally va­cat­ing one­self from the de­ci­sion mak­ing pro­cess. For ex­am­ple, when a SCOTUS Judge re­cuses him­self, he doesn’t take any role in the de­bate or vot­ing. Similarly, when mod­er­at­ing the EA face­book group, gen­er­ally con­flicted mods won’t ar­gue for a po­si­tion (though this is just my de­scrip­tion of a norm and not an ex­plicit rule we’ve had). If fund mem­bers with large con­flicts of in­ter­est end up de facto mak­ing the de­ci­sion any­way then the COI policy doesn’t seem to have achieved any­thing.

  I would sug­gest in­stead that other fund man­agers re­search the ap­pli­ca­tion and make the de­ci­sion. This would help avoid an un­fair bias to­wards fund­ing peo­ple who are ‘in the com­mu­nity’.
  

• Larks 9 Feb 2020 22:11 UTC

  8 points

  in reply to: rmoehn's comment on: Re­quest for Feed­back: Draft of a COI policy for the Long Term Fu­ture Fund

  If I imag­ine my­self in the shoes of a more con­ser­va­tive po­ten­tial donor, who is check­ing the fine print as part of their due dili­gence, I would be put off by phrases like ‘meta­mour’, ‘con­sump­tion of drugs’ and the re­peated men­tion­ing of sex­ual re­la­tion­ships.

  The pur­pose of dis­clo­sure is to provide po­ten­tial donors with in­for­ma­tion they con­sider rele­vant to their de­ci­sion pro­cess. That some of donors will be per­suaded not to donate by the in­for­ma­tion is a fea­ture, not a bug.

• Larks 8 Feb 2020 5:30 UTC

  12 points

  in reply to: bmg's comment on: Con­cern­ing the Re­cent 2019-Novel Coron­avirus Outbreak

  The peo­ple mak­ing the bet aren’t, even pretty in­di­rectly, in a po­si­tion to in­fluence the man­age­ment of the tragedy or the ded­i­ca­tion of re­sources to it. It doesn’t ac­tu­ally mat­ter all that much, in other words, if one of them is over- or un­der-con­fi­dent about some as­pect of the tragedy.

  Do you think the bet would be less ob­jec­tion­able if Justin was able to in­crease the num­ber of deaths?

• Larks 1 Feb 2020 1:44 UTC

  10 points

  in reply to: RyanCarey's comment on: EA Fo­rum Prize: Win­ners for De­cem­ber 2019

  You’re wel­come! It was a tough de­ci­sion, as I did find it quite mo­ti­vat­ing last year, but figured it would cre­ate the ap­pear­ance of a con­flict of in­ter­est if it won this year.

• Larks 17 Jan 2020 1:57 UTC

  2 points

  in reply to: G Gordon Worley III's comment on: Space gov­er­nance is im­por­tant, tractable and neglected

  One ex­am­ple might be that early on, colonies are ex­tremely re­li­ant on the home world (as the ISS is to­day) so a lot of cen­tral con­trol is ex­erted by earth. Later on the large dis­tances in­volve make both aid and (many forms of) co­er­cion much more difficult, so much more de­cen­tral­i­sa­tion and in­de­pen­dence seem likely. From our van­tage point it seems un­likely if we can do much to in­fluence the lat­ter given we first have to go through the rad­i­cally differ­ent former.

• Larks 15 Jan 2020 4:43 UTC

  2 points

  on: Long-term in­vest­ment fund at Founders Pledge

  This com­ment is less well cited than my usual, but per­haps one of the ideas might spark some use­ful thought:

  To learn more about e.g. risks of value drift, risks of ex­pro­pri­a­tion and le­gal struc­tures, we’re look­ing into case stud­ies of similar funds, both ac­tive now and in the past.

  Un­for­tu­nately I can­not find the refer­ence, but I seem to re­call the Ford Foun­da­tion hav­ing suffered such se­vere value drift—far from what Henry Ford in­tended—that the fam­ily ba­si­cally di­s­owned it.

  Creative ideas for op­ti­mal gov­er­nance of the fund

  I have heard (though have not ver­ified, so con­sider this very spec­u­la­tive) that there is a le­gal struc­ture left over from the cru­sades (for pro­tect­ing your cas­tle while you were away in the holy land) that might be use­ful for cry­on­ics—per­haps it might be of use here.

• Larks 14 Jan 2020 23:15 UTC

  4 points

  in reply to: Khorton's comment on: Khor­ton’s Shortform

  Un­for­tu­nately, that means that those of us with our rep­u­ta­tions on the line are the ones who have the most skin in the game to keep peo­ple from do­ing stupid unilat­er­al­ist things that make ev­ery­one in the com­mu­nity look bad.

  Surely if some­one doesn’t iden­tify as an EA, their ac­tions in­cur less rep­u­ta­tional risk for the move­ment?

• Larks 14 Jan 2020 23:09 UTC

  5 points

  on: Dataset of Trillion Dol­lar figures

  In­ter­est­ing idea!

  I no­tice that many of the largest num­bers are deriva­tive no­tion­als. It is im­por­tant to note that this is a to­tally ir­rele­vant num­ber; deriva­tive no­tion­als are es­sen­tially ar­bi­trary up to a scalar mul­ti­ple.

  As an ex­am­ple, sup­pose you and I want to make a bet about overnight in­ter­est rates on the first day of 2021 - speci­fi­cally we agree that I will pay you $1 for ev­ery 1% the Fed Funds overnight rate is above 2%, and you will pay me $1 for ev­ery 1% it is be­low 2%, capped at $2 ei­ther way. The way we would for­mal­ise this as a con­tract would be:

  • An in­ter­est rate swap with 2% rate, one day tenor and $36,500 no­tional.

  • A re­ceiver swap­tion with a 4% strike, one day tenor and $36,500 no­tional.

  • A payer swap­tion with a 0% strike, one day tenor and $36,500 no­tional.

  In to­tal this is over $100,000 worth of no­tional… for a $2 bet! What mat­ters is the eco­nomic ex­po­sure of the deriva­tives, but this can be hard for non-spe­cial­ists to calcu­late, so peo­ple of­ten sub­sti­tute the eas­ier but ir­rele­vant ques­tion of gross no­tional. Un­for­tu­nately this can in­clude reg­u­la­tions, which has caused a va­ri­ety of prob­lems in the mar­ket.

  
  

  Separately, you list ‘global debt in the non-fi­nan­cial sec­tor’ as $1,521 trillion, but the source pro­vided sug­gests it is $152 trillion. I sus­pect your scrap­ing tool may have mis­taken a foot­note for an or­der of mag­ni­tude.

• Larks 6 Jan 2020 20:34 UTC

  2 points

  in reply to: Linch's comment on: The Cen­ter for Elec­tion Science Year End EA Appeal

  He op­poses the Fed, but does he want in­ter­est rates set by poli­ti­ci­ans? My un­der­stand­ing is he wanted a re­turn to the gold stan­dard, where there is less need to di­rectly con­trol the money sup­ply—the only ques­tion is whether you in­sist on 100% back­ing or ac­cept a lower ra­tio, but once that is set growth in the money sup­ply is de­ter­mined by the vol­ume of phys­i­cal gold.

  Similarly bit­coin peo­ple op­pose the Fed, but not be­cause they want poli­ti­ci­ans in charge—they have an­other ex­ter­nal rule (e.g. fixed max quan­tity) that re­duces demo­cratic dis­cre­tion.

• Larks 5 Jan 2020 2:25 UTC

  19 points

  on: The Cen­ter for Elec­tion Science Year End EA Appeal

  Thanks for writ­ing this; I find ap­proval vot­ing an in­ter­est­ing and in­tu­itively at­trac­tive sys­tem (and have used it in the past).

  I was sur­prised by the table you show about the im­pacts of ap­proval vot­ing on the demo­cratic can­di­date se­lec­tion. I gen­er­ally think of ap­proval vot­ing as sup­port­ing mod­er­ate can­di­dates, as you men­tion, but here it seems to be favour­ing the most ex­treme (War­ren, San­ders) over the more mod­er­ate (Bi­den) - though per­haps I have mis­in­ter­preted the chart.

  This makes me worry that in­creas­ing the ‘demo­crat­ic­ness’ of elec­tions might lead to worse out­comes. You sort of alude to this con­cern in the FAQ, but move over it pretty quickly. There are in fact many cases where it is I think rel­a­tively com­mon to think that re­duc­ing ‘demo­crat­ic­ness’ is a good move:

  • In­de­pen­dent cen­tral banks are much more com­pe­tent, cred­ible and tech­no­cratic than hav­ing elected poli­ti­ci­ans con­trol the money sup­ply. I’m not re­ally aware of any­one who thinks this move was a bad idea.

  • Elected judges are of­ten viewed as far more pop­ulist and gen­er­ally lower qual­ity than ap­pointed ones.

  • Elected util­ity com­mis­sions are less com­pe­tent than ap­pointed ones (though per­haps I am bi­ased on this is­sue).

  • Referen­dums and bal­lot ini­ti­a­tives are of­ten blamed as part of the cause for the de­cline in Cal­ifor­nian gov­er­nance.

  Elected vs Ap­pointed isn’t ex­actly the same as FPTP vs Ap­proval Vot­ing, but it seems like they have similar as­pects.

  Gar­rett Jones has a book on this. It’s only on pre-or­der at the mo­ment, but you can read Han­son’s com­ments here.

• Larks 5 Jan 2020 1:42 UTC

  6 points

  on: Ge­netic En­hance­ment as a Cause Area

  I thought this was a very in­ter­est­ing ar­ti­cle, but I would ques­tion how much coun­ter­fac­tual differ­ence we could ex­pect in­ter­ven­tion to make here. My de­fault ex­pec­ta­tion is that a lot of this is go­ing to hap­pen any­way due to de­mand from par­ents. My im­pres­sion is that IVF and PGS screen­ing both be­came very com­mon­place with lit­tle ex­plicit policy sup­port for ex­actly this rea­son. This doesn’t ap­ply to set­ting ex­ter­nal in­cen­tives, but does ap­ply to many of the spe­cific tech­nolo­gies you men­tion.

  I also thought this was a lit­tle strange:

  One way of alle­vi­at­ing the harm due to in­equal­ity is by ad­vo­cat­ing a tax on in­nate, un­earned qual­ities, such as fa­vor­able ge­net­ics and in­her­i­tance. I be­lieve that these poli­cies will be pop­u­lar once the tech­nol­ogy comes up on the hori­zon, and will likely play a large role in miti­gat­ing the worst risks of in­equal­ity.

  With ge­netic en­hance­ment, favourable ge­net­ics are (no longer) ran­dom—they are the re­sult of a de­liber­ate de­ci­sion that you are try­ing to en­courage. How many par­ents would want to curse their child with higher taxes? It seems rather strange that we should start tax­ing (e.g. dis­cour­ag­ing) this good thing pre­cisely at the mo­ment it be­comes pos­si­ble to pro­mote it!

  Fi­nally, you might en­joy this ar­ti­cle by Eliezer. One in­ter­est­ing point is that there is some­thing of a col­lec­tive ac­tion prob­lem, be­cause each mu­ta­tion is prob­a­bly bad for the in­di­vi­d­ual/​fam­ily with it but pro­vides use­ful in­for­ma­tion for ev­ery­one else.