mr-hire(Matt Goldenberg)

• mr-hire 6 Apr 2020 20:54 UTC

  4 points

  in reply to: Dagon's comment on: Si­varam Ko­nanki’s Shortform

  Google has had a com­pet­ing product in google hang­outs for a while, but it’s not as high qual­ity.

• mr-hire 5 Apr 2020 13:30 UTC

  2 points

  in reply to: ricraz's comment on: How spe­cial are hu­man brains among an­i­mal brains?

  Koko the go­rilla had par­tial lan­guage com­pe­tency.

  AFAICT this is highly dis­puted. Many peo­ple think that her han­dlers had an agenda, and that the pur­ported ex­am­ples of her com­bin­ing words were her ran­domly spam­ming sign lan­guage to get treats. Raw data was never realeased, and no one was al­lowed to in­ter­act with or see them in­ter­act with Koko ex­cept her han­dlers.

  It seems plau­si­ble that the pur­ported ex­am­ples are a case of se­lec­tive re­port­ing, wish­ful think­ing, and the Clever Hans effect.

• mr-hire 3 Apr 2020 22:50 UTC

  5 points

  in reply to: NaiveTortoise's comment on: An1lam’s Short Form Feed

  I’m not su­per fa­mil­iar with the com­pet­i­tive math cir­cuit, but my un­der­stand­ing is that this is part of it? Peo­ple are given a hard prob­lem and ei­ther in­di­vi­d­u­ally or as a team solve it as quickly as pos­si­ble.

• mr-hire 30 Mar 2020 16:49 UTC

  5 points

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

  In­creased sense of re­lat­ed­ness seems a big one missed here.

• mr-hire 28 Mar 2020 1:47 UTC

  5 points

  on: Matt Gold­en­berg’s Short Term Feed

  Some­thing else in the vein of “things EAs and ra­tio­nal­ists should be pay­ing at­ten­tion to in re­gards to Corona.”

  There’s a com­mon failure mode in large hu­man sys­tems where one out­lier causes us to cre­ate a rule that is a worse equil­ibrium. In the Per­son­alMBA, Josh Kauf­man talks about some­one tak­ing ad­van­tage of a “buy any book you want” rule that a com­pany has—so you make it so that you can no longer get any free books.

  This same pat­tern has hap­pened be­fore in the US, af­ter 9-11 - We cre­ated a whole bunch of se­cu­rity the­ater, that caused more suffer­ing for ev­ery­one, and gave gov­ern­ment way more power and way less over­sight than is safe, be­cause we over-re­acted to pre­vent one bad event, not con­sid­er­ing the coun­ter­fac­tual in­visi­ble things we would be los­ing.

  This will hap­pen again with Corona, things will be put in place that are maybe good at pre­vent­ing pan­demics (or worse, mak­ing peo­ple think they’re safe from pan­demics), but cre­ate a mil­lion triv­ial con­ve­niences ev­ery day that add up to more strife than they’re worth.

  Th­ese types of rules are very hard to re­peal af­ter the fact be­cause of ab­sence blind­ness—some­one needs to do the work of calcu­lat­ing the cost/​benefit ra­tio BEFORE they get im­ple­mented, then build a con­vinc­ing enough nar­ra­tive to what seems ob­vi­ous/​com­mon sense mea­sures given the cli­mate/​dev­as­ta­tion.

• mr-hire 26 Mar 2020 0:09 UTC

  8 points

  on: Oc­cam’s Guillotine

  Real­ity ac­tu­ally ex­ists and has prop­er­ties you can de­ter­mine through study and ex­per­i­men­ta­tion.

  Sure, agreed.

  Con­clu­sions fol­low from their premises and it’s un­rea­son­able to ex­pect a plu­ral­ity of truths.

  Sure agreed.

  Our uni­verse is con­sis­tent

  Yep

  your un­der­stand­ing of the pieces should fit to­gether.

  Why? This seems like a clas­sic mis­take of naive ra­tio­nal­ism. Models ex­ist for rea­sons, and brains have limits. Depend­ing on my use case, It may make sense for me to have a set of heuris­tics I know don’t fit to­gether be­ca­sue they’re use­ful.

  To use an ex­am­ple from el­se­where in the post, ten­sile strength is a leaky ab­strac­tion com­pared to some­thing like the uni­ver­sal wave func­tion, but the former is much more use­ful for build­ing bridges. Mean­while, a fire­fighter is go­ing to be us­ing heuris­tics like “close to break­ing” and “how much weight it can bear” or even a vague feel­ing of “dan­ger” that is only a leaky ab­strac­tion of things like ten­sile strength.

  Now, ‘should’ my un­der­stand­ing of all those pieces fit to­gether? Not re­ally, de­pends on what I want to use those mod­els for.

  Like Cicero, he was do­ing philos­o­phy at a time when philos­o­phy meant liv­ing a bet­ter life. He found sug­ges­tions farmed from differ­ent sys­tems of thought, but which prac­ti­cally helped one live a bet­ter life, and helped form a foun­da­tion for sto­icism, a set of tools and heuris­tics for liv­ing bet­ter. “Should” he have cre­ated a set of prin­ci­ples that were com­pletely log­i­cally con­sis­tent? Well, it may have helped him gen­er­ate more. But, in terms of a set of prac­ti­cal tools and thoughts that helped one live bet­ter, he did quite well cu­rat­ing us­ing an eclec­tic style.

• mr-hire 25 Mar 2020 20:46 UTC

  2 points

  in reply to: Dagon's comment on: Matt Gold­en­berg’s Short Term Feed

  I did up­date from this quite sig­nifi­cantly.

• mr-hire 25 Mar 2020 17:54 UTC

  4 points

  in reply to: ryan_b's comment on: His­tory’s Biggest Nat­u­ral Experiment

  It de­pends on the size of the win­dow. If schizophre­nia shows up be­tween 20-25 years later, then the 1 year effects of the quaran­tine get dis­tributed over that 5 year win­dow, and are much harder to de­tect above other fluc­tu­a­tions.

• mr-hire 25 Mar 2020 17:05 UTC

  2 points

  in reply to: mr-hire's comment on: Matt Gold­en­berg’s Short Term Feed

  It was brought to my at­ten­tion on Less­wrong that de­pres­sions ac­tu­ally save lives.

  Which would make it much harder to build a sim­ple “two curves to flat­ten” nar­ra­tive out of.

• mr-hire 25 Mar 2020 16:02 UTC

  2 points

  in reply to: miro's comment on: Matt Gold­en­berg’s Short Term Feed

  It’s in­ter­est­ing be­cause you would in­tu­itively think this, but there is ac­tu­ally not ter­rible ev­i­dence link­ing pe­ri­ods of eco­nomic growth to in­creased mor­tal­ity.

  Wow that is fas­ci­nat­ing. It does make the case harder to make be­cause you have to start quan­tify­ing hap­piness/​de­pres­sion, etc and trade off against lives. Much much harder to sim­plify enough to make it viral. Up­dates to­wards cap­i­tal­ism be­ing hor­rible.

  Is non-profit fund­ing re­ally that in­elas­tic in de­pres­sion?

  It prob­a­bly varies quite a bit by sec­tor, and where fund­ing comes from for differ­ent non-prof­its. In the case of AI safety I think it’s likely more in­elas­tic than AI ca­pa­bil­ity.

• mr-hire 25 Mar 2020 15:01 UTC

  9 points

  on: Matt Gold­en­berg’s Short Term Feed

  Was think­ing a bit about the how to make it real for peo­ple that the quaran­tine de­press­ing the econ­omy kills peo­ple just like Coron­avirus does.

  Was think­ing about find­ing a sim­ple good enough cor­re­la­tion be­tween eco­nomic de­pres­sion and death, then cre­at­ing a “flat­ten­ing the curve” graphic that shows how many deaths we would save from stop­ping the eco­nomic freefall at differ­ent points. Com­bin­ing this was clear nar­ra­tives about re­ces­sion could be quite effec­tive.

  On the other hand, I think it’s quite plau­si­ble that this par­tic­u­lar prob­lem will take care of it­self. When peo­ple be­gin to ex­pe­rience de­pres­sion, will the young peo­ple who are the eco­nomic en­g­ine of the coun­try re­ally con­tinue to stay home and quaran­tine them­selves? It seems quite likely that we’ll sim­ply be­come strat­ified for a while where young healthy peo­ple break quaran­tine, and the older and im­muno-com­pro­mised stay home.

  But get­ting the time of this right is ev­ery­thing. Strik­ing the right bal­ance of “deaths from eco­nomic freefall” and “deaths from an over­loaded med­i­cal sys­tem” is a bal­anc­ing act, go­ing too far in ei­ther di­rec­tion re­sults in hun­dreds of thou­sands of un­nec­es­sary deaths.

  Then I got to think­ing about the effect of a de­pressed econ­omy on x-risks from AI. Be­cause the fund­ing for AI safety is

  1. Mostly in non-profits

  and

  2. Orders of mag­ni­tude smaller than fund­ing for AI capabilities

  It’s quite likely that the fund­ing for AI safety is more in­elas­tic in de­pres­sions than than the fund­ing for AI ca­pa­bil­ities. This may an­swer the puz­zle of why more EAs and ra­tio­nal­ists aren’t speak­ing co­gently about the trade­offs be­tween de­pres­sion and lives saved from Corona—they have gone through this same train of thought, and de­cided that pre­vent­ing a de­pres­sion is an in­for­ma­tion haz­ard.

• mr-hire 25 Mar 2020 13:26 UTC

  2 points

  in reply to: Davidmanheim's comment on: What will the eco­nomic effects of a 3 week quaran­tine be? 3 months?

  I think of we’re talk­ing about the coun­ter­fac­tual with the best trade­offs, it might look some­thing like quaran­tin­ing the most vuln­er­a­ble pop­u­la­tions while hav­ing oth­ers get back to work.

• mr-hire 24 Mar 2020 23:27 UTC

  4 points

  in reply to: Zack_M_Davis's comment on: Can crimes be dis­cussed liter­ally?

  If liter­ally ev­ery­one knew, what would be the func­tion of mak­ing the claim? How do you end up with a sys­tem that wouldn’t work with­out false as­ser­tions, and yet allegedly “ev­ery­one” knows that the as­ser­tions are false?

  This is an­swered in Ben­quo’s last post, take a look at stages 3 and 4 to see how this situ­a­tion can arise.

  https://​​www.less­wrong.com/​​posts/​​fEX7G2N7Ct­mZQ3eB5/​​simu­lacra-and-sub­jec­tivity

• mr-hire 24 Mar 2020 14:43 UTC

  6 points

  in reply to: romeostevensit's comment on: What will the eco­nomic effects of a 3 week quaran­tine be? 3 months?

  You can still make quan­ti­ta­tive pre­dic­tions with un­cer­tainty. Note that beat­ing the mar­ket isn’t im­pos­si­ble (some­one has to make the mar­ket effi­cient) and even if you don’t have enough data to beat the mar­ket, that just means you make your con­fi­dence in­ter­vals larger. The point of Bayesian episte­mol­ogy is that even large un­cer­tainty can be quan­tified.

• mr-hire 24 Mar 2020 0:18 UTC

  2 points

  in reply to: jmh's comment on: Against Dog Ownership

  But I’m say­ing that it does, and that your point is not a counter?

• mr-hire 23 Mar 2020 22:03 UTC

  4 points

  in reply to: jmh's comment on: Against Dog Ownership

  But it’s also a mis­take to think that us­ing an an­i­mal in that way won’t lead to more em­pa­thy for an­i­mals in gen­eral—the ques­tion is in­stead to ask how peo­ple will best main­tain con­sis­tent self—images? In prac­tice for the peo­ple who own pets, do you find them more em­pathic of an­i­mals af­ter own­ing them? I think that

  1. It does seem peo­ple use pets for their own needs, and of­ten fool them­selves that they’re do­ing it for the pet/​ they’re a com­pas­sion­ate pet owner

  2. Peo­ple seem to be more com­pas­sion­ate to­wards an­i­mals when hav­ing owned pets, in or­der to main­tain this image of com­pas­sion­ate pet owner.

  So in my ex­pe­rience the ev­i­dence is that both what you and Chris are say­ing is true.

• mr-hire 23 Mar 2020 17:41 UTC

  4 points

  in reply to: Dagon's comment on: Vanessa Kosoy’s Shortform

  I’m think­ing hu­mans striv­ing for ra­tio­nal choices,

  It feels like there’s bet­ter words for this like ra­tio­nal­ity, whereas bayesi­anism is a more spe­cific philos­o­phy about how best to rep­re­sent and up­date be­liefs.

• mr-hire 23 Mar 2020 17:39 UTC

  2 points

  in reply to: jmh's comment on: Against Dog Ownership

  I don’t see how us­ing a pet as a per­sonal emo­tional need re­lates to de­creas­ing em­pa­thy for an­i­mals. It’s likely to me that would in­crease em­pa­thy.

• mr-hire 23 Mar 2020 3:27 UTC

  2 points

  on: Reflec­tions on a no con­tent Jan­uary 2020 experiment

  I made a pre­dic­tion when you started this that you’d do more pro­cess­ing than cre­at­ing. Cu­ri­ous if it came true

• mr-hire 22 Mar 2020 0:12 UTC

  2 points

  in reply to: orthonormal's comment on: Ra­tion­al­ists, Post-Ra­tion­al­ists, And Ra­tion­al­ist-Adjacents

  There’s a set of post-ra­tio­nal­ist norms where switch­ing frames isn’t a con­ver­sa­tional gam­bit, it’s ex­pected and part of gen­er­a­tive pro­cess for solv­ing prob­lems and cre­at­ing close­ness. I would love to see peo­ple be able to switch be­tween these differ­ent types of norms, as it can be equally frus­trat­ing when you’re try­ing to vibe with peo­ple who can only op­er­ate through ra­tio­nal­ist frames.