SatvikBeri

• SatvikBeri 22 Apr 2019 16:07 UTC

  1 point

  in reply to: Dr_Manhattan's comment on: Crypto quant trad­ing: Intro

  Yes

• SatvikBeri 10 Feb 2019 19:01 UTC

  18 points

  on: When should we ex­pect the ed­u­ca­tion bub­ble to pop? How can we short it?

  I think the prior that bub­bles usu­ally pop is in­cor­rect. We tend to call some­thing a bub­ble in ret­ro­spect, af­ter it’s popped.

  But if you try to define bub­bles with purely for­ward-look­ing mea­sures, like a pe­riod of un­usu­ally high growth, they’re more fre­quently fol­lowed by pe­ri­ods of un­usu­ally slow growth, not rapid de­cline. For ex­am­ple, Ama­zon’s stock would pass just about any test of a bub­ble over most points in its his­tory.

  I ex­pect some­thing similar with ed­u­ca­tion, spend­ing will likely re­main high, but grow more slowly than it did in the last 20 years. That’s es­pe­cially true be­cause of the struc­ture of stu­dent loans, peo­ple can’t re­ally just de­fault.

  But to an­swer the more di­rect ques­tion: as­sum­ing that there is a rapid drop in ed­u­ca­tion spend­ing, how could we profit from it? Vo­ca­tional schools seem like the most ob­vi­ous bet, e.g. to be­come a pro­gram­mer, den­tal as­sis­tant, mas­sage ther­a­pist, elec­tri­cian, and so on.

  Cer­tifi­ca­tion ser­vices that man­age to de­velop a rep­u­ta­tion will be­come strong as well, e.g. SalesForce cer­tifi­cates are pretty valuable.

  You could di­rectly short lenders such as Sal­lie Mae.

  Re­cruit­ment agen­cies that spe­cial­ize in plac­ing re­cent col­lege grad­u­ates will likely suffer.

  Man­age­ment con­sult­ing firms rely heav­ily on col­lege grad­u­ates, and so do hedge funds to a lesser ex­tent.

• SatvikBeri 28 Nov 2018 2:19 UTC

  9 points

  AF

  on: Iter­a­tion Fixed Point Exercises

  #6:

  As­sume WLOG $f\left(x\right)>=x$Then by mono­ton­ic­ity, we have $x<=f\left(x\right)<=f^2\left(x\right)<=...<=f^{|P|}\left(x\right)$If this chain were all strictly greater, than we would have $|P|+1$istinct el­e­ments. Thus there must be some $k$uch that $f^k\left(x\right)=f^{k+1}\left(x\right)$By in­duc­tion, $f^{n+1}\left(x\right)=f^n\left(x\right)=f^k\left(x\right)$or all $n>k$

  #7:

  As­sume $f\left(x\right)>=x$nd con­struct a chain similarly to (6), in­dexed by el­e­ments of $\alpha$If all in­equal­ities were strict, we would have an in­jec­tion from $\alpha$o L.

  #8:

  Let F be the set of fixed points. Any sub­set S of F must have a least up­per bound $x$n L. If x is a fixed point, done. Other­wise, con­sider $f^{\alpha }\left(x\right)$ which must be a fixed point by (7). For any q in S, we have $f\left(q\right)\le x\Rightarrow f^{\alpha }\left(q\right)\le f^{\alpha }\left(x\right)\Rightarrow q\le f^{\alpha }\left(x\right)$ Thus $f^{\alpha }\left(x\right)$s an up­per bound of S in F. To see that it is the least up­per bound, as­sume we have some other up­per bound b of S in F. Then $x<=b\Rightarrow f^{\alpha }\left(x\right)<=f^{\alpha }\left(b\right)=b$

  To get the lower bound, note that we can flip the in­equal­ities in L and still have a com­plete lat­tice.

  #9:

  P(A) clearly forms a lat­tice where the up­per bound of any set of sub­sets is their union, and the lower bound is the in­ter­sec­tion.

  To see that in­jec­tions are mono­tonic, as­sume $A_0\subseteq A_1$nd $f$s an in­jec­tion. For any func­tion, $f\left(A_0\right)\subseteq f\left(A_1\right)$ If $a\notin A_0$nd $f\left(a\right)\in f\left(A_0\right)$that im­plies $f\left(a\right)=f\left(a^{\prime }\right)$or some $a^{\prime }\in A_0$which is im­pos­si­ble since $f$s in­jec­tive. Thus $f$s (strictly) mono­tonic.

  Now $h:=g\circ f$s an in­jec­tion $A\to A$Let $X$e the set of all points not in the image of $g$and let $A^{\prime }=X\cup h\left(X\right)\cup h^2\left(X\right)\cup ...$ote that $h\left(A^{\prime }\right)=h\left(X\right)\cup h^2\left(X\right)\cup h^3\left(X\right)\cup ...=A^{\prime }-X$since no el­e­ment of $X$s in the image of $h$Then $g(B-f(A^{\prime }\left)\right)=g\left(B\right)-h\left(A^{\prime }\right)=g\left(B\right)-(A^{\prime }-X)=g\left(B\right)-A^{\prime }+g\left(B\right)\cap X=g\left(B\right)-A^{\prime }$On one hand, ev­ery el­e­ment of A not con­tained in $g\left(B\right)$s in $A^{\prime }$y con­struc­tion, so $A-A^{\prime }\subseteq g\left(B\right)$ On the other, clearly $g\left(B\right)\subseteq A$so $g\left(B\right)-A^{\prime }\subseteq A-A^{\prime }$QED.

  #10:

  We form two bi­jec­tions us­ing the sets from (9), one be­tween A’ and B’, the other be­tween A—A’ and B—B’.

  Any in­jec­tion is a bi­jec­tion be­tween its do­main and image. Since $B^{\prime }=f\left(A^{\prime }\right)$nd $f$s an in­jec­tion, $f$s a bi­jec­tion where we can as­sign each el­e­ment $b^{\prime }\in B^{\prime }$o the $a^{\prime }\in A^{\prime }$uch that $f\left(a^{\prime }\right)=b^{\prime }$Similarly, $g$s a bi­jec­tion be­tween $B-B^{\prime }$nd $A-A^{\prime }$Com­bin­ing them, we get a bi­jec­tion on the full sets.

• SatvikBeri 27 Oct 2018 3:39 UTC

  3 points

  in reply to: nshepperd's comment on: Kal­man Filter for Bayesians

  Thanks! Edited.

• SatvikBeri 27 Oct 2018 3:38 UTC

  3 points

  in reply to: rohinmshah's comment on: Kal­man Filter for Bayesians

  Thanks! Edited. Yeah, I speci­fi­cally fo­cused on var­i­ance be­cause of how Bayesian up­dates com­bine Nor­mal dis­tri­bu­tions.

Kal­man Filter for Bayesians

Sys­tem­iz­ing and Hacking

• SatvikBeri 20 Mar 2018 16:17 UTC

  7 points

  in reply to: StefanDeYoung's comment on: In­fer­ence & Empiricism

  Yes.

In­fer­ence & Empiricism

• SatvikBeri 14 Feb 2018 21:16 UTC

  20 points

  in reply to: Raemon's comment on: Ra­tion­al­ist Lent

  I’m speci­fi­cally giv­ing up games that en­courage many short check-ins, e.g. most phone games and idle games. Binges aren’t a big is­sue for me, they tend to give me joy and re­newal. But fre­quent check-in games make me less happy and less pro­duc­tive.

• SatvikBeri 29 Jan 2018 23:21 UTC

  19 points

  on: Ham­mers and Nails

  “Pre­fer a few large, sys­tem­atic de­ci­sions to many small ones.”

  1. Pick what per­centage of your port­fo­lio you want in var­i­ous as­sets, and re­bal­ance quar­terly, rather than mak­ing reg­u­lar buy­ing/​sel­l­ing decisions

  2. Pri­ori­tize once a week, and by de­fault do what­ever’s next on the list when you com­plete a task.

  3. Set up re­cur­ring hang­outs with friends at what­ever fre­quency you en­joy (e.g. weekly). Can­cel or resched­ule on an ad-hoc ba­sis, rather than schedul­ing ad-hoc

  4. Ri­gor­ously de­cide how you will judge the re­sults of ex­per­i­ments, then run a lot of them cheaply. Ma­chine Learn­ing ex­am­ple: pick one eval­u­a­tion met­ric (might be a com­pos­ite of sev­eral sub-met­rics and rules), then au­to­mat­i­cally run lots of differ­ent mod­els and do a deeper dive into the 5 that perform par­tic­u­larly well

  5. Make a pack­ing check­list for trips, and use it repeatedly

  6. Figure out what crite­ria would make you leave your cur­rent job, and only take in­ter­views that plau­si­bly meet those criteria

  7. Pick a rou­tine for your com­mute, e.g. listen­ing to pod­casts. Test new ideas at the rou­tine level (e.g. pod­casts vs books)

  8. Find a spe­cific method for de­cid­ing what to eat—for me, this is query­ing sys­tem 1 to ask how I would feel af­ter eat­ing cer­tain foods, and pick­ing the one that re­turns the best answer

  9. Ac­cept­ing ev­ery time a coworker asks for a game of ping-pong, as a way to get ex­er­cise, un­less I am about to en­ter a meeting

  10. Always sug­gest­ing the same small set of places for coffee or lunch meetings

• SatvikBeri 29 Jan 2018 22:54 UTC

  52 points

  in reply to: SquirrelInHell's comment on: A LessWrong Crypto Autopsy

  A good gen­eral rule here is to think in terms of what per­centage of your port­fo­lio (or net worth) you want in a spe­cific as­set class, rather than mak­ing buy­ing/​sel­l­ing a bi­nary de­ci­sion. Then re­bal­ance ev­ery 3 months.

  For ex­am­ple, you might de­cide you want 2.5%-5% in crypto. If the price quadru­pled, you would well about 75% of your stake at the end of the quar­ter. If it halved, you would buy more.

  The ma­jor benefit is that this moves you from mak­ing many small de­ci­sions to one big de­ci­sion, which is usu­ally eas­ier to get right.

• SatvikBeri 30 Dec 2017 16:32 UTC

  16 points

  in reply to: Benito's comment on: Why did ev­ery­thing take so long?

  It’s widely be­lieved that hu­mans ba­si­cally lived in <200 per­son tribes that didn’t in­ter­act with each other too much be­fore agri­cul­ture, so one might won­der how any­thing got any amount of adop­tion.

  And only about 100 years ago, hu­man­ity es­sen­tially for­got the cure for scurvy: http://​​idle­words.com/​​2010/​​03/​​scott_and_scurvy.htm

Ex­am­ples of Miti­gat­ing As­sump­tion Risk

• SatvikBeri 29 Nov 2017 19:16 UTC

  12 points

  in reply to: Benito's comment on: Gears Level & Policy Level

  One thing I’d add to that list is that the post fo­cuses on re­fin­ing ex­ist­ing con­cepts, which is quite valuable and gen­er­ally doesn’t get enough at­ten­tion.

• SatvikBeri 29 Nov 2017 19:11 UTC

  24 points

  in reply to: Chris Hibbert's comment on: Gears Level & Policy Level

  Pro­vid­ing Slack at the pro­ject level in­stead of the task level is a re­ally good idea, and has worked well in many fields out­side of pro­gram­ming. It is analo­gous to the con­cept of in­surance: the RoI on Slack is higher when you ag­gre­gate many events with at least par­tially un­cor­re­lated er­rors.

  One ma­jor prob­lem with try­ing to fix es­ti­mates at the task level is that there are strong in­cen­tives not to finish a task too early. For ex­am­ple, if you es­ti­mated 6 weeks, and are al­most done af­ter 3, and some­thing mod­er­ately ur­gent comes up, you’re more likely to switch and fix that ur­gent thing since you have time. On the other hand, if you es­ti­mated 4 weeks, you’re more likely to de­lay the other task (or ask some­one else to do it).

  As a re­sult, I’ve found that teams are liter­ally likely to finish pro­jects faster with higher qual­ity if you es­ti­mate the pro­ject as, say, 8 3-week tasks with 24 weeks of over­all slack (so 48 weeks to­tal) than if you es­ti­mate the pro­ject as a 8 6-week tasks.

  This is some­what coun­ter­in­tu­itive but re­ally easy to ap­ply in prac­tice if you have a bit of so­cial cap­i­tal.

• SatvikBeri 4 Nov 2017 14:38 UTC

  3 points

  in reply to: jacobjacob's comment on: The Coper­ni­can Revolu­tion from the Inside

  Nit­pick: as I un­der­stand, Fey­er­abend would agree. His main ar­gu­ment seems to be “any sim­ple method­ol­ogy for de­cid­ing whether a sci­en­tific the­ory is true or false (such as falsifi­ca­tion­ism) would have missed im­por­tant ad­vances such as he­lio­cen­trism, New­ton’s the­ory of grav­ity, and rel­a­tivity, there­fore philoso­phers of sci­ence should stop try­ing to for­mu­late sim­ple ac­cept/​re­ject method­olo­gies.”

• SatvikBeri 2 Nov 2017 17:47 UTC

  24 points

  in reply to: ESRogs's comment on: Com­pet­i­tive Truth-Seeking

  I could have been clearer—hiring is definitely a case where you get some points for fol­low­ing con­sen­sus, un­like, say, ac­tive in­vest­ing where you’re typ­i­cally mea­sured on alpha. And fol­low­ing con­sen­sus on some parts of your pro­cess is fine if you have an edge el­se­where (e.g. Google and Face­book pay more than most, so hav­ing con­sen­sus-level as­sess­ment is fine.) But I would ar­gue that for most star­tups you’ll see some­thing like or­der-of-mag­ni­tude im­prove­ments through Pro­cess C.

Com­pet­i­tive Truth-Seeking

• SatvikBeri 30 Oct 2017 13:43 UTC

  17 points

  in reply to: Zvi's comment on: Inad­e­quacy and Modesty

  I think the main cause is that peo­ple who view them­selves are solv­ing a prob­lem are of­ten us­ing the pro­ce­dure “look at the cur­rent pat­tern and try to find is­sues with it.” A pro­cess that com­ple­ments this well is “look at what’s worked his­tor­i­cally, and do more of it.”

  Some ex­am­ples I wrote about a while back: less­wrong.com/​​lw/​​iro/​​sys­tem­atic_lucky_breaks/​​