Links

Ge­net­ics:

• Every­thing Is Her­i­ta­ble:

  • “Quan­ti­fy­ing the im­pact of rare and ul­tra­-rare cod­ing vari­a­tion across the phe­no­typic spec­trum”, Ganna et al 2017
  • “Dis­cov­ery Of The First Genome-Wide Sig­nifi­cant Risk Loci For ADHD”, De­mon­tis et al 2017
  • “Ge­netic con­tri­bu­tion to two fac­tors of neu­roti­cism is as­so­ci­ated with afflu­ence, bet­ter health, and longer life”, Hill et al 2017
  • “En­vi­ron­men­tal fac­tors dom­i­nate over host ge­net­ics in shap­ing hu­man gut mi­cro­biota com­po­si­tion”, Roth­schild et al 2017 (com­men­tary; “bio­me-ex­plain­abil­ity”—a rare use of vari­ance com­po­nents to check whether there is even a nee­dle in the haystack)

• Re­cent Evo­lu­tion:

  • “Wide­spread sig­na­tures of pos­i­tive se­lec­tion in com­mon risk al­le­les as­so­ci­ated to autism spec­trum dis­or­der”, Poli­manti & Gel­ern­ter 2017

  • dys­gen­ics:

    • dis­ease mu­ta­tion load: “The Ge­nomic Health Of An­cient Ho­minins”, Berens et al 2017 (me­dia; +35% per­centile in­crease in gen­eral dis­ease risks over past ~mil­len­ni­a?)
    • se­lec­tion against ed­u­ca­tion in the UK: “Sig­na­tures of neg­a­tive se­lec­tion in the ge­netic ar­chi­tec­ture of hu­man com­plex traits”, Zeng et al 2018
    • “De­tect­ing poly­genic adap­ta­tion in ad­mix­ture graphs”, Racimo et al 2017 (No sign of se­lec­tion is found for Eu­ro­peans, but if I un­der­stand the method right, it’s only look­ing at net se­lec­tion with time steps cor­re­spond­ing to pop­u­la­tions branch­ing. So given the Eu­ro­pean pat­tern of se­lec­tion for in­tel­li­gence at least up un­til agri­cul­ture and then heavy re­cent dys­gen­ics against education/intelligence, those might mostly can­cel com­pared to an East Asian pop­u­la­tion like Japan which started lat­er.)

• En­gi­neer­ing:

  • “The Moral Ques­tion That Stan­ford Asks Its Bio­engi­neer­ing Stu­dents: It’s about the ca­pac­ity to print a hu­man genome.”

AI:

• “Deep re­in­force­ment learn­ing from hu­man pref­er­ences”, Chris­tiano et al 2017 (blogs: 1, 2)
• “Re­lu­plex: An Effi­cient SMT Solver for Ver­i­fy­ing Deep Neural Net­works”, Katz et al 2017
• “A sim­ple neural net­work mod­ule for re­la­tional rea­son­ing”, San­toro et al 2017 (blog)
• “One Model To Learn Them All”, Kaiser et al 2017 (a sin­gle NN for mul­ti­-modal tasks: from im­age clas­si­fi­ca­tion to im­age cap­tion­ing to Eng­lish pars­ing to Eng­lish⟺Ger­man⟺French trans­la­tion)
• “At­ten­tion Is All You Need” for SOTA ma­chine trans­la­tion, Vaswani et al 2017
• “Deal or No Deal? End-to-End Learn­ing for Ne­go­ti­a­tion Di­a­logues”, Lewis et al 2017 (blog)
• “Learn­ing to Learn from Noisy Web Videos”, Ye­ung et al 2017

Statistics/meta-science/mathematics:

• “Is the stag­ger­ingly profitable busi­ness of sci­en­tific pub­lish­ing bad for sci­ence?”
• “Mea­sure­ment Er­ror, Re­gres­sion to the Mean, and Group Differ­ences”
• “What works in e-com­merce—a meta-analy­sis of 6700 on­line ex­per­i­ments”, Brown & Jones 2017 (In­for­ma­tive pri­ors: most A/B ex­per­i­ments fail and the suc­cess­ful ones have small effects on the or­der of a few % at most. Note the im­pli­ca­tions that most suc­cess­ful, in the sense of p < 0.05, A/B tests will grossly over­es­ti­mate the true effect size; de­tect­ing re­al­is­tic effects may re­quire large sam­ple sizes; and some cat­e­gories of tests may well not be worth run­ning at all. It would also be in­ter­est­ing to know how many of those nulls were based on cor­re­la­tional data…)

Politics/religion:

• “In De­fense of In­di­vid­u­al­ist Cul­ture”
• “The Dark Legacy of Car­los Cas­taneda”

Psychology/biology:

• “Does High Self­-es­teem Cause Bet­ter Per­for­mance, In­ter­per­sonal Suc­cess, Hap­pi­ness, or Health­ier Lifestyles?”, Baumeis­ter et al 2003; “The Man Who De­stroyed Amer­i­ca’s Ego: How a rebel psy­chol­o­gist chal­lenged one of the 20th cen­tu­ry’s biggest—and most dan­ger­ous—ideas”; “‘It was qua­si­-re­li­gious’: the great self­-es­teem con”
• “Lithium in Drink­ing Wa­ter and In­ci­dence of Sui­cide: A Na­tion­wide In­di­vid­u­al-Level Co­hort Study with 22 Years of Fol­low-Up”, Knud­sen et al 2017 (a ma­jor blow to the lithium hy­poth­e­sis)
• “Where do hy­pothe­ses come from?”, Das­gupta et al 2017
• “How To Fall To Your Death And Live To Tell The Tale”
• “My déjà vu is so ex­treme I can’t tell what’s real any more”

Tech­nol­o­gy:

• “Com­pu­ta­tional Com­plex­ity of Air Travel Plan­ning”, De Mar­cken 2003 / ITA Soft­ware (Air­line ticket search is not just NP-hard or worse, but un­de­cid­able.)
• “In­for­ma­tion Leak­age from Op­ti­cal Em­a­na­tions”, Loughry & Umphress 2002
• “Google-Wide Pro­fil­ing: A Con­tin­u­ous Pro­fil­ing In­fra­struc­ture for Data Cen­ters”, Ren et al 2010 (“…GWP pro­files re­vealed that the zlib li­brary ac­counted for nearly 5% of all CPU cy­cles con­sumed…”)

Eco­nom­ics:

• “Prop­erty Is An­other Name for Monopoly: Fa­cil­i­tat­ing Effi­cient Bar­gain­ing with Par­tial Com­mon Own­er­ship of Spec­trum, Cor­po­ra­tions, and Land”, Pos­ner & Weyl 2016
• “The Per­fect Heist: Recipes from Around the World”, Lafleur et al 2014

Film/TV

Live-ac­tion:

• L.A. Con­fi­den­tial (one of the best noir films; Ex­ley is par­tic­u­larly in­ter­est­ing as nei­ther hero nor an­ti-hero)
• All About Eve (re­view)
• Amy (2015) (re­view)
• 8 1⁄2 (a lame waste of time; not re­motely funny nor in­sight­ful nor ren­dered suffi­ciently in­ter­est­ing by the now-ex­otic set­ting of 1960s Catholic Italy, as the mak­ing of 8 1⁄2 is Fellini’s ex­cuse for mak­ing 8 1⁄2 in a third-rate break­ing-the-fourth-wall ex­er­cise)
• An­nie Hall (a few one-lin­ers aside, Woody Al­len’s neu­rotic hu­mor is un­bear­able to watch. It would seem that Allen movies are not for me.)

Music

Touhou:

• “冬中火葬” (PIROPARU; Send_Off {C80}) [folk]
• “人別離苦” (PIROPARU; Send_Off {C80}) [folk]
• “廃獄のロア” (Cororo feat. Sara Mat­sumo­to; Edis re­vir -River­side vo­cal best al­bum & ryuno’s art works vol.1- {R8}) [folk]
• “揺籃曲 -Pure Japan­ese Ver.-” (Cororo feat. Sara Mat­sumo­to; Edis re­vir -River­side vo­cal best al­bum & ryuno’s art works vol.1- {R8}) [folk]
• “Aisya -ye ru am­ma-” (Cororo feat. Sara Mat­sumo­to; Edis re­vir -River­side vo­cal best al­bum & ryuno’s art works vol.1- {R8}) [folk]

Dou­jin:

• “Ke­mono Friends BGM” (Kou Ogata; {2017}) [Celtic]