Links

AI:

• Mat­ters Of Scale:

  • The Bit­ter Les­son: “Ac­cu­racy and Per­for­mance Com­par­i­son of Video Ac­tion Recog­ni­tion Ap­proaches”, Hutchin­son et al 2020 (sim­ple large 2D CNNs per­formed best when com­pared un­der con­trolled con­di­tion­s); “A Met­ric Learn­ing Re­al­ity Check”, Mus­grave et al 2020 (the dou­bling of met­ric learn­ing per­for­mance since 2006 ap­pears al­most en­tirely due to larger neural nets + more com­pute used in train­ing, not fancy loss func­tion­s/the­o­ry)
  • “Self­-su­per­vised learn­ing through the eyes of a child”, Orhan et al 2020 (dataset; it’d be fun to see this used for rats or dogs)
  • “The re­mark­able, yet not ex­tra­or­di­nary, hu­man brain as a scaled-up pri­mate brain and its as­so­ci­ated cost”, Her­cu­lano-Houzel 2012
  • “The Node Is Non­sense: There are bet­ter ways to mea­sure progress than the old Moore’s law met­ric”, Moore 2020
  • “Mea­sur­ing hard­ware over­hang”, hippke (“with to­day’s al­go­rithms, com­put­ers would have beat the world chess cham­pion al­ready in 1994 on a con­tem­po­rary desk com­puter”)

Ge­net­ics:

• Every­thing Is Her­i­ta­ble:

  • “Sib­ling val­i­da­tion of poly­genic risk scores and com­plex trait pre­dic­tion”, Lello et al 2020
  • “Poly­genic Scores for Cog­ni­tive Abil­i­ties and their As­so­ci­a­tion with Differ­ent As­pects of Gen­eral In­tel­li­gence—a Deep Phe­no­typ­ing Ap­proach”, Genç et al 2020
  • “Ge­netic test­ing of dogs pre­dicts prob­lem be­hav­iors in clin­i­cal and non­clin­i­cal sam­ples”, Za­p­ata et al 2020 (pre­vi­ous­ly: MacLean et al 2019)

Sta­tis­tic­s/meta-science/­math­e­mat­ics:

• “Ob­ject­ing to ex­per­i­ments even while ap­prov­ing of the poli­cies or treat­ments they com­pare”, Heck et al 2020

Psy­chol­o­gy/bi­ol­o­gy:

• “Epi­ge­netic clocks: A re­view”, José Luis 2020; “TRIIM: Re­ver­sal of epi­ge­netic ag­ing and im­munose­nes­cent trends in hu­mans”, Fahy et al 2019; “Re­ver­sal of Epi­ge­netic Age with Diet and Lifestyle in a Pi­lot Ran­dom­ized Clin­i­cal Trial”, Fitzger­ald et al 2020

• “The quest to slow age­ing through drug dis­cov­ery”, Par­tridge et al 2020

• “Atrib­ac­te­ria re­pro­duc­ing over mil­lions of years in the At­lantic abyssal sub­seafloor”, Vuillemin et al 2020

• “Ex­plor­ing the per­some: The power of the item in un­der­stand­ing per­son­al­ity struc­ture”, Rev­elle et al 2020 (“every­thing is cor­re­lated”)

• How Com­pe­tent Are Hu­mans?

  • “Bet­ter All the Time: How the “per­for­mance rev­o­lu­tion” came to ath­let­ic­s—and be­yond"; “Per­sonal Best: Top ath­letes and singers have coach­es. Should you?”
  • “Does Man­age­ment Mat­ter? Ev­i­dence from In­dia”, Bloom et al 2012; “The Im­pact of Con­sult­ing Ser­vices on Small and Medium En­ter­pris­es: Ev­i­dence from a Ran­dom­ized Trial in Mex­ico”, Bruhn et al 2018
  • “95%-ile is­n’t that good”, Dan Luu (per­sis­tent in­com­pe­tence in Over­watch play­ers)
  • “Worker Rush: De­scent to Bronze”, Gheed 2011 (and in Star­craft play­ers)
  • “Play­ing to Win”, Sir­lin 2006

• “The Scent of the Nile: Jean-Claude El­lena cre­ates a new per­fume”, Chan­dler Burr 2005

Tech­nol­o­gy:

• Your Com­puter Is Re­ally Fast:

  • “There’s plenty of room at the Top: What will drive com­puter per­for­mance after Moore’s law?”, Leis­er­son et al 2020 (gains from end-to-end sys­tems de­sign; AI scal­ing can con­tinue even if semi­con­duc­tors do not, par­tic­u­larly with self­-op­ti­miz­ing stacks; John Car­mack notes, apro­pos of Sey­mour Cray, that “Hy­per­scale data cen­ters and even na­tional su­per­com­put­ers are loosely cou­pled things to­day, but if chal­lenges de­manded it, there is a world with a zetta[flops] scale, tightly in­te­grat­ed, low la­tency ma­trix dis­si­pat­ing a gi­gawatt in a swim­ming pool of cir­cu­lat­ing flu­o­rinert.”)
  • “Scal­a­bil­i­ty! But at what COST?”, Mc­Sh­erry et al 2015 (when lap­tops > clus­ters: the im­por­tance of good base­li­nes)
  • “You’re Do­ing It Wrong: Think you’ve mas­tered the art of server per­for­mance? Think again.”, Poul-Hen­ning Kamp 2010 (on us­ing cache-obliv­i­ous B-heaps to op­ti­mize Var­nish per­for­mance 10×)
  • “Sam­ple Fac­to­ry: Ego­cen­tric 3D Con­trol from Pix­els at 100,000 FPS with Asyn­chro­nous Re­in­force­ment Learn­ing”, Pe­trenko et al 2020
  • “What Every Pro­gram­mer Should Know About Mem­ory”, Drep­per 2007
  • “What it takes to run Stack Over­flow”, Nick Craver
  • see pre­vi­ously on la­ten­cy/UI: “It’s the La­ten­cy, Stu­pid.”/“Com­puter la­ten­cy: 1977–2017”/“Key­board La­tency”/“Ter­mi­nal La­tency”/“Lo­cal-first soft­ware: You own your data, in spite of the cloud”/Klepp­mann et al 2019/“Web Bloat”

• “Fälschungser­schw­erende Schrift (forgery-im­ped­ing type­face)”

• “Inuk­shuk: Cari­bou Drive Lanes on South­ern Vic­to­ria Is­land, Nunavut, Canada”, Brink 2005

• “The case of the bar­na­cled crys­tal”, Kohman 1950 (on dis­ap­pear­ing poly­morphs)

Eco­nom­ics:

• “Mov­ing off the Map: How Knowl­edge of Or­ga­ni­za­tional Op­er­a­tions Em­pow­ers and Alien­ates”, Huis­ing 2019 (“The CEO, after be­ing walked through the map, sat down, put his head on the table, and said, ‘This is even more fucked up than I imag­ined.’”)
• “The Age of Rea­son: Fi­nan­cial De­ci­sions over the Life-Cy­cle with Im­pli­ca­tions for Reg­u­la­tion”, Agar­wal et al 2009

Fic­tion:

• “An Au­to­bi­o­graph­i­cal Es­say”, Jorge Luis Borges 1971
• “Sin­gu­lar: Pos­si­ble fu­tures of the sin­gu­lar­ity”, James Yu & GPT-3

Misc:

• List of games the Bud­dha would not play

Music

Touhou:

• “Moon­bow” (Kirin feat. Chata; Sprout In­ten­tion {C88}) [trance]
• “Har­vester’s Dance” (Ganemes; Sis­ter’s 3 {TK12}) [folk]