Lanrian

• Lanrian 9 Jul 2020 8:21 UTC

  5 points

  on: The New Front­page De­sign & Open­ing Tag Creation!

  Here <INSERT LINK TO TAG GUIDELINES> are some rough tag-guidelines about what makes a good tag.

  Is there no guideline yet? Or is there sup­posed to be a link here?

• Lanrian 8 Jul 2020 19:13 UTC

  1 point

  in reply to: Lanrian’s comment on: (an­swered: yes) Has any­one writ­ten up a con­sid­er­a­tion of Downs’s “Para­dox of Vot­ing” from the per­spec­tive of MIRI-ish de­ci­sion the­o­ries (UDT, FDT, or even just EDT)?

  As an aside, for re­ally large pop­u­la­tions, it would prob­a­bly be so­cially op­ti­mal to only have a small frac­tion of the pop­u­la­tion vot­ing (at least if we ig­nore things like le­gi­t­i­macy, feel­ing of par­ti­ci­pa­tion, etc). As long as that frac­tion is ran­domly sam­pled, you could get good statis­ti­cal guaran­tees that the out­come of the elec­tion would be the same as if ev­ery­one voted. South Korea did a pretty cool ex­per­i­ment where they ex­posed a rep­re­sen­ta­tive sam­ple of 500 peo­ple to pro- and anti-nu­clear ex­perts, and then let them de­cide how much nu­clear power the coun­try should have.

  I don’t think this is why CDTs re­fuses to vote, though.

• Lanrian 8 Jul 2020 18:56 UTC

  11 points

  in reply to: Daniel Kokotajlo’s comment on: (an­swered: yes) Has any­one writ­ten up a con­sid­er­a­tion of Downs’s “Para­dox of Vot­ing” from the per­spec­tive of MIRI-ish de­ci­sion the­o­ries (UDT, FDT, or even just EDT)?

  It was es­pe­cially cool in that it said that even al­tru­ist CDTers can’t ac­count for the ra­tio­nal­ity of vot­ing in suffi­ciently large elec­tions.

  That’s pretty sur­pris­ing. I checked out the page, and he un­for­tu­nately doesn’t mo­ti­vate what kind of model he’s us­ing, so it’s hard to ver­ify. From the book:

  If the im­por­tance of the elec­tion is pre­sumed pro­por­tionate to the size of the elec­torate, then for large enough elec­tions, ex­pected-util­ity calcu­la­tions can­not jus­tify the effort of vot­ing by ap­peal to the small but heav­ily weighted pos­si­bil­ity that your vote will be a tiebreaker. The odds of that out­come de­crease faster than lin­early with the num­ber of vot­ers, so the ex­pected value of your vote as a tiebreaker ap­proaches zero— even tak­ing ac­count of the value to ev­ery­one com­bined, not just your­self. Given enough vot­ers, then, the causal value (even to ev­ery­one) of your vote is over­shad­owed by the in­con­ve­nience to you of go­ing out to vote.”

  In an elec­tion with two choices, in a model where ev­ery­body has 50% chance of vot­ing for ei­ther side, I don’t think the claim is true. Maybe he’s as­sum­ing that the out­comes of elec­tions be­come eas­ier to pre­dict as they grow larger, be­cause in­di­vi­d­ual vari­abil­ity be­comes less im­por­tant? If ev­ery­one has a 51% prob­a­bil­ity of vot­ing for a cer­tain side, the elec­tion would be pretty much guaran­teed for an ar­bi­trar­ily large pop­u­la­tion, in which case a CDTer wouldn’t have any rea­son to vote (even if there was a coal­i­tion of CDTers who could swing the elec­tion). I’m not sure if it’s true that elec­tions in larger coun­tries are more pre­dictable, though.

• Lanrian 4 May 2020 22:10 UTC

  3 points

  in reply to: jonathanclark’s comment on: Could some­one please start a bright home light­ing com­pany?

  Check out our product, it’s a far-cry from the $400 you men­tioned but many times less ex­pen­sive and thin­ner than coelux.

  How ex­pen­sive is it? Maybe there’s some rea­son that you don’t want an ex­act price in the com­par­i­son, but can you give a rough range?

• Lanrian 3 May 2020 17:23 UTC

  LW: 1 AF: 1

  AF

  in reply to: Donald Hobson’s comment on: How does iter­ated am­plifi­ca­tion ex­ceed hu­man abil­ities?

  If you picked the me­dian hu­man by math­e­mat­i­cal abil­ity, and put them in this setup, I would be rather sur­prised if they pro­duced a valid proof of Fer­mats last the­o­rem.

  I would too. IDA/​HCH doesn’t have to work with the me­dian hu­man, though. It’s ok to pick an ex­cel­lent hu­man, who has been trained for be­ing in that situ­a­tion. Paul has ar­gued that it wouldn’t be that sur­pris­ing if some hu­mans could be ar­bi­trar­ily com­pe­tent in an HCH-setup, even if some couldn’t.

• Lanrian 9 Apr 2020 8:31 UTC

  4 points

  in reply to: AllAmericanBreakfast’s comment on: The Sand­wich Argument

  The re­search of Philip Tet­lock shows that fore­cast­ers achieve bet­ter Brier scores when they ex­ag­ger­ate their con­fi­dence.

  They showed that it’s good to ex­trem­ise the pre­dic­tions of teams, when com­bin­ing pre­dic­tions that agree with each other, but I don’t think that in­di­vi­d­ual fore­cast­ers were sys­tem­at­i­cally un­der­con­fi­dent.

• Lanrian 3 Apr 2020 8:59 UTC

  4 points

  in reply to: Daniel Kokotajlo’s comment on: Im­pli­ca­tions of the Dooms­day Ar­gu­ment for x-risk reduction

  If an­ces­tor simu­la­tions are one of the main uses of cos­mic re­sources, we prob­a­bly will go ex­tinct soon (some­what de­pend­ing on how you define ex­tinc­tion), be­cause we’re prob­a­bly in an an­ces­tor simu­la­tion that will be turned off. If the simu­la­tors were to keep us al­ive for billions of years, it would be pretty un­likely that we didn’t find our­selves liv­ing in those billions of years, by the same logic as the dooms­day ar­gu­ment.

• Lanrian 31 Mar 2020 22:17 UTC

  3 points

  in reply to: Scott Alexander’s comment on: April Coron­avirus Open Thread

  Link to pa­per, the rele­vant figure is on page 12.

• Lanrian 26 Mar 2020 17:14 UTC

  1 point

  in reply to: ignoranceprior’s comment on: Coron­avirus Open Thread

  That ar­ti­cle is based on a twit­ter thread that is based on this ar­ti­cle that is based on the par­li­a­men­tary hear­ing that Wei Dai linked. The twit­ter thread dis­torted the ar­ti­cle a lot, and seems to be mostly spec­u­la­tion.

• Lanrian 25 Mar 2020 20:23 UTC

  3 points

  in reply to: Douglas_Knight’s comment on: Author­i­ties and Amateurs

  The gaus­sian as­sump­tion makes no differ­ence.

  As I said, I do agree that the piece’s qual­i­ta­tive con­clu­sion was cor­rect. How­ever, the gaus­sian as­sump­tion does make a large quan­ti­ta­tive differ­ence. Com­par­ing it to the ex­treme: If we always have the max­i­mum num­ber of peo­ple in ICUs, con­tin­u­ously, the time un­til herd-im­mu­nity would be 4.9 years, which is a fac­tor 3 less than what the nor­mal as­sump­tion gives you. Although it is still clearly too much, it’s only one or two ad­di­tional fac­tors of 3 away from be­ing rea­son­able. This ex­treme isn’t even that un­re­al­is­tic; some­thing like it could plau­si­bly be achieved if the gov­ern­ment con­tin­u­ously al­ter­nated be­tween more or less lock-down, keep­ing the av­er­age R close to 1.

  To be clear, I think that it’s good that the post was writ­ten, but I think it would have been sub­stan­tially bet­ter if it had used a con­stant num­ber of in­fected peo­ple. If you’re go­ing to use un­re­al­is­tic mod­els (which, in many cases, you should!) it’s good prac­tice to use the most con­ser­va­tive model pos­si­ble, to en­sure that your con­clu­sion holds no mat­ter what, and to let your reader see that it holds no mat­ter what. In ad­di­tion, it would have been sim­pler (you just have to mul­ti­ply/​di­vide 3 num­bers), and it would have looked less like re­al­is­tic-and-au­thor­i­ta­tive-math to the av­er­age reader, which would have bet­ter com­mu­ni­cated its real epistemic sta­tus.

• Lanrian 25 Mar 2020 11:38 UTC

  6 points

  on: Author­i­ties and Amateurs

  No­talge­braist posted a rea­son­able cri­tique of Flat­ten­ing The Curve Is a Deadly Delu­sion, ex­plain­ing how it’s in­cor­rect to as­sume that flat­ten­ings won’t re­duce the to­tal num­ber of cases, and that it doesn’t make sense to as­sume a nor­mal dis­tri­bu­tion. I think that the the piece’s main point was cor­rect, though: that the curve would need to get crazy flat for hos­pi­tals to not be over­loaded.

  3 days later, the Im­pe­rial Col­lege study made the same point in a much bet­ter and more rigor­ous way (among lots of other good points), but it was less widely shared on so­cial me­dia.

  I’m not sure what the con­clu­sion here is. Non-ex­perts will some­times make false as­sump­tions and get the de­tails wrong, but they’re still ca­pa­ble of mak­ing good points that only re­quire you to mul­ti­ply num­bers to­gether, and will do so a few days faster and in a way that’s more memet­i­cally fit than pa­pers from ex­perts?

• Lanrian 23 Mar 2020 14:19 UTC

  4 points

  in reply to: remizidae’s comment on: Against Dog Ownership

  This made me cu­ri­ous about where most peo­ple get their dogs from. Ap­par­ently, some­thing like 34% are pur­chased from breed­ers and 23% are ob­tained from shelters, ac­cord­ing to https://​​www.as­pca.org/​​an­i­mal-home­less­ness/​​shelter-in­take-and-sur­ren­der/​​pet-statis­tics (though their num­bers don’t add up to 100%, and I’m not sure why). Get­ting them from friends/​​rel­a­tives is also pretty com­mon, at 20%.

• Lanrian 23 Mar 2020 10:48 UTC

  2 points

  on: Are vet­er­ans more self-dis­ci­plined than non-vet­er­ans?

  Is lack of dis­ci­pline a big prob­lem at an av­er­age work­place? I would ex­pect most offices to provide suffi­ciently good so­cial in­cen­tives that most peo­ple spend most of their time work­ing, in which case any pro­duc­tivity-boon from in­creased dis­ci­pline would be swamped by other (se­lec­tion) effects from mil­i­tary train­ing.

  In­creased dis­ci­pline could trans­late to no­tably higher pro­duc­tivity for un­der­grads, PhD stu­dents, or peo­ple who work from home, though. My im­pres­sion is that such peo­ple strug­gle much more with pro­cras­ti­na­tion.

  Re­lat­edly, China seems to be do­ing their best to teach dis­ci­pline in schools, so you could look at whether that seems to be work­ing. This is ob­vi­ously mixed in with lots of on­go­ing so­cial in­cen­tives, though. Rele­vant ssc: https://​​slat­estar­codex.com/​​2020/​​01/​​22/​​book-re­view-re­view-lit­tle-sol­diers/​​

• Lanrian 19 Mar 2020 9:49 UTC

  1 point

  in reply to: romeostevensit’s comment on: What are good ways of con­vinc­ing some­one to re­think an im­pos­si­ble dream?

  Why would you need to? Couldn’t you just con­vince them that other peo­ple won’t like their idea? In the sec­ond ex­am­ple above, there seems to be am­ple ev­i­dence that other peo­ple aren’t in­ter­ested.

• Lanrian 16 Mar 2020 21:12 UTC

  3 points

  in reply to: Scott Alexander’s comment on: Coron­avirus Open Thread

  Isn’t this ex­actly what “flat­ten the curve” is about? Be­cause a lot of peo­ple are talk­ing about that as a solu­tion, in­clud­ing some gov­ern­ments.

  The main prob­lem is that the curve needs to get re­ally flat for hos­pi­tals to have time with ev­ery­one. Depend­ing on how over­whelmed you want your hos­pi­tals to be, you could be in lock-down for sev­eral years. Some calcu­la­tions in this ar­ti­cle.

• Lanrian 14 Mar 2020 10:38 UTC

  2 points

  in reply to: Unnamed’s comment on: A Sig­nifi­cant Por­tion of COVID-19 Trans­mis­sion Is Presymptomatic

  In­so­far as the virus mostly spreads through presymp­tomatic trans­mis­sion in some coun­tries, that’s al­most cer­tainly be­cause the peo­ple with symp­toms are all iso­lated. Symp­tomatic peo­ple definitely spread the dis­ease.

  It’d be in­ter­est­ing if the R of these pop­u­la­tions were <1 at the time the stud­ies were done, though. If so, presymp­tomatic trans­mis­sion might be in­suffi­cient to sus­tain ex­po­nen­tial growth, as long as all symp­tomatic trans­mis­sion is pre­vented.

• Lanrian 11 Mar 2020 21:09 UTC

  5 points

  in reply to: IL’s comment on: Coron­avirus Open Thread

  An­gela Merkel says that 60-70% of Ger­many is likely to be in­fected. That’s use­ful if peo­ple be­lieve that it won’t in­fect that many. Ex­am­ple source, though you can google for oth­ers.

  If they’re will­ing to be­lieve a red­di­tor’s sum­mary, this one says that WHO says that 20% of in­fected peo­ple needed hos­pi­tal treat­ment for weeks. (If they want pri­mary sources, maybe you could find those claims in the links /​ some­where else.)

  Put­ting to­gether 1 and 2 (and gen­er­al­is­ing from Ger­many to what­ever coun­try they’re in), they ought to be con­vinced that it’s pretty se­vere.

• Lanrian 7 Mar 2020 10:01 UTC

  11 points

  on: Cred­i­bil­ity of the CDC on SARS-CoV-2

  Our best guess is that the CDC is try­ing to con­serve masks for health care pro­fes­sion­als and oth­ers with the high­est need, in the face of a loom­ing mask short­age. That could eas­ily be the op­ti­mum mask al­lo­ca­tion. I can’t prove the lie wasn’t jus­tified for the greater good. But it is an­other ex­am­ple of the CDC plac­ing “get­ting the out­come it wants” over “tel­ling peo­ple the literal truth.”

  As far as I can tell, the CDC hasn’t ut­tered a literal lie about this. In the link, they only say “CDC does not recom­mend that peo­ple who are well wear a face­mask to pro­tect them­selves from res­pi­ra­tory dis­eases, in­clud­ing COVID-19”, which is a recom­men­da­tion, rather than a state­ment of effi­cacy. It could be mo­ti­vated by a de­sire to stop mask-hoard­ing, as you say, or by the be­lief that typ­i­cal us­age of masks (in­clud­ing reuse, fre­quently read­just­ing the mask and thereby touch­ing your face, etc) ac­tu­ally harms peo­ple more than it helps them.

  (It’s in­ter­est­ing that the link also says “The use of face­masks is also cru­cial for health work­ers and peo­ple who are tak­ing care of some­one in close set­tings (at home or in a health care fa­cil­ity).” This is (i) an ad­mis­sion that masks can pro­tect you when you’re close to some­one sick, and (ii) does provide an in­cen­tive for peo­ple to hoard masks, if they think they’re go­ing to be tak­ing care of some­one.)

  Re­gard­less, it’s fair to say that they’re plac­ing “get­ting the out­come it wants” at least over “tel­ling peo­ple the full truth”, and that this is a strike against the CDC’s trust­wor­thi­ness.

  Edit: This SSC says that the CDC has been ad­vis­ing the pub­lic against us­ing masks for a long time, so what­ever they’re say­ing, they’re prob­a­bly say­ing it for differ­ent rea­sons than to stop hoard­ing.

• Lanrian 6 Mar 2020 17:12 UTC

  2 points

  on: What does Solomonoff in­duc­tion say about brain du­pli­ca­tion/​con­scious­ness?

  This is highly re­lated to UDASSA. In the linked post, es­pe­cially Prob­lem #2 (about split­ting con­scious com­put­ers) and bits of Prob­lem #3 (e.g. “What hap­pens if we ap­ply UDASSA to a quan­tum uni­verse? For one, the ex­is­tence of an ob­server within the uni­verse doesn’t say any­thing about con­scious ex­pe­rience. We need to spec­ify an al­gorithm for ex­tract­ing a de­scrip­tion of that ob­server from a de­scrip­tion of the uni­verse”...)

• Lanrian 6 Mar 2020 11:32 UTC

  3 points

  in reply to: Bucky’s comment on: Is there a bet­ter way to define groups for COVID-19 im­pact?

  Some data on in­fants: https://​​ja­manet­work.com/​​jour­nals/​​jama/​​fullar­ti­cle/​​2761659