An1lam

• An1lam 16 Aug 2019 0:53 UTC

  3 points

  in reply to: Matthew Barnett's comment on: Matthew Bar­nett’s Shortform

  I per­son­ally try to bal­ance keep­ing my­self hon­est by hav­ing some goal out­side but also trust­ing my­self enough to know when I should de­pri­ori­tize the origi­nal goal in fa­vor of some­thing else.

  For ex­am­ple, let’s say I set a goal to write a blog post about a topic I’m learn­ing in 4 hours, and half-way through I re­al­ize I don’t un­der­stand one of the key un­der­ly­ing con­cepts re­lated to the thing I in­tended to write about. Dur­ing an ac­tual test, the right thing to do would be to do my best given what I know already and finish as many ques­tions as pos­si­ble. But I’d ar­gue that in the blog post case, I very well may be bet­ter off say­ing, “OK I’m go­ing to go learn about this other thing un­til I un­der­stand it, even if I don’t end up finish­ing the post I wanted to write.”

  The pithy way to say this is that tests are ba­si­cally pure Good­hardt, and it’s dan­ger­ous to turn ev­ery real life task into a game of max­i­miz­ing leg­ible met­rics.

• An1lam 16 Aug 2019 0:35 UTC

  8 points

  in reply to: Matthew Barnett's comment on: Matthew Bar­nett’s Shortform

  I’ve also been read­ing text­books more and ex­pe­rien­cied some frus­tra­tion, but I’ve found two things that, so far, help me get less stuck and feel less guilt.

  After try­ing to learn math from text­books on my own for a month or so, I started pay­ing a tu­tor (DM me for de­tails) with whom I meet once a week. Like you, I strug­gle with get­ting stuck on hard ex­er­cises and/​or con­cepts I don’t un­der­stand, but hav­ing a tu­tor makes it eas­ier for me to move on know­ing I can dis­cuss my con­fu­sions with them in our next ses­sion. Un­for­tu­nately, a pay­ing a tu­tor re­quires ac­tu­ally hav­ing $ to spare on an on­go­ing ba­sis, but I also sus­pect for some peo­ple it just “feels weird”. If some­one read­ing this is more de­terred by this lat­ter rea­son, con­sider that ba­si­cally ev­ery­one who wants to se­ri­ously im­prove at any phys­i­cal ac­tivity gets 1-on-1 in­struc­tion, but for some rea­son do­ing the same for men­tal ac­tivi­ties as an adult is weirdly un­com­mon (and per­haps a lit­tle low sta­tus).

  I’ve also started to fol­low MIT OCW courses for things I want to learn rather than try­ing to read en­tire text­books. Yes, this means I may not cover as much ma­te­rial, but it has helped me bet­ter gauge how much time to spend on differ­ent top­ics and al­low me to feel like I’m pro­gress­ing. The ma­jor down­side of this strat­egy is that I have to re­mind my­self that even though I’m learn­ing based on a course’s ma­te­ri­als, my goal is to learn the ma­te­rial in a way that’s use­ful to me, not to mem­o­rize pass­words. Also, be­cause I know how long the courses would take in a uni­ver­sity con­text, I do oc­ca­sion­ally feel guilt if I fall be­hind due to spend­ing more time on a spe­cific topic. Still, on net, us­ing courses as loose guides has been work­ing bet­ter for me than just try­ing to 100 per­cent en­tire math text­books.

• An1lam 16 Aug 2019 0:23 UTC

  5 points

  in reply to: Matthew Barnett's comment on: Matthew Bar­nett’s Shortform

  From my own ex­pe­rience go­ing through a similar re­al­iza­tion and try­ing to ap­ply it to my own pro­duc­tivity, I found that cer­tain things I tried ac­tu­ally helped me sus­tain­ably work more pro­duc­tively but oth­ers did not.

  What has worked for me based on my ex­pe­rience with exam-like situ­a­tions is hav­ing clear goals and time boxes for work ses­sions, e.g. the blog post ex­am­ple you de­scribed. What hasn’t worked for me is try­ing to im­pose ag­gres­sively short dead­lines on my­self all the time to in­cen­tivize my­self to fo­cus more in­tensely. Per­son­ally, the level of fo­cus I have dur­ing ex­ams is driven by an un­sus­tain­able level of stress, which, if ap­plied con­tin­u­ously, would prob­a­bly lead to burnout and/​or pro­cras­ti­na­tion bing­ing. That said, oc­ca­sion­ally ar­tifi­cially im­pos­ing dead­lines has helped me en­gage exam-style fo­cus when I need to do some­thing that might oth­er­wise be bor­ing be­cause it mostly in­volves ex­e­cut­ing known strate­gies rather than do­ing more open, ex­plo­ra­tory think­ing. For hard think­ing though, I’ve ac­tu­ally found that giv­ing my­self con­ser­va­tively long time boxes helps me fo­cus bet­ter by al­low­ing me to re­lax and take my time. I saw you men­tioned strug­gling with read­ing text­books above, and while I still strug­gle try­ing to read them too, I have found that not ex­pect­ing mirac­u­lous progress helps me get less frus­trated when I read them.

  Re­lated to all this, you used the term “deep work” a few times so you may already be fa­mil­iar with Cal New­port’s work. But, if you’re not I recom­mend a few of his rele­vant posts (1, 2) de­scribing how he pro­duces work ar­ti­facts that act as a forc­ing func­tion for learn­ing the right stuff and stay­ing fo­cused.

• An1lam 12 Aug 2019 19:23 UTC

  2 points

  in reply to: Chris Cooper's comment on: AI Safety Read­ing Group

  Thanks!

• An1lam 11 Aug 2019 17:46 UTC

  4 points

  in reply to: Søren Elverlin's comment on: AI Safety Read­ing Group

  I’m very in­ter­ested in his re­sponses to the fol­low­ing ques­tions:

  1. The ques­tion ad­dress­ing Gw­ern’s post about Tool AIs want­ing to be Agent AIs.

  2. The ques­tion ad­dress­ing his op­ti­mism about progress with­out the­o­ret­i­cal break­throughs (re­lated to NNs/​DL).

• An1lam 8 Aug 2019 16:55 UTC

  9 points

  on: Why do hu­mans not have built-in neu­ral i/​o chan­nels?

  I re­ally like this ques­tion but don’t feel con­fi­dent about any of my cur­rent an­swer ideas.

  I did want to note that a good test for a rea­son­able an­swer is whether it passes a “would this have ruled out hu­man speech?” test. That is, any an­swer that la­bels “com­mu­ni­cat­ing by ma­nipu­lat­ing vibra­tional waves with tongues” im­pos­si­ble prob­a­bly should be dis­qual­ified.

  To­tally un­re­lated to my prior point, Chil­dren of Time, a biotech/​evolu­tion heavy sci-fi book which I highly recom­mend, ac­tu­ally fo­cuses on

  a species that com­mu­ni­cates com­plex knowl­edge and mem­o­ries through gen­er­a­tions and, sub­se­quently, be­tween in­di­vi­d­u­als, via DNA viruses.

• An1lam 17 Jul 2019 17:37 UTC

  1 point

  in reply to: An1lam's comment on: What are we pre­dict­ing for Neu­ral­ink event?

  Scor­ing my pre­dic­tions, I’d say I was right on 1 and 3, and mostly right on 2. On 2, Musk dis­cussed his idea of merg­ing with AI, which is pretty clearly aug­men­ta­tion, but no con­crete timelines for im­plan­ta­tion in healthy hu­mans was given.

• An1lam 16 Jul 2019 21:15 UTC

  1 point

  in reply to: Ruby's comment on: What makes peo­ple in­tel­lec­tu­ally ac­tive?

  Yeah, I al­most wish I’d ex­cluded that refer­ence to my own post. Re-read­ing it now, I re­al­ize it comes off as “woe is me, no one read my post” but I more meant it as “just re­mov­ing hur­dles has its own prob­lems, like some­times pub­lish­ing stuff that isn’t very good. There­fore, I’m not yet sure where the right bal­ance lies.”

• An1lam 14 Jul 2019 20:01 UTC

  3 points

  on: What are we pre­dict­ing for Neu­ral­ink event?

  I mostly agree with Vaniver’s pre­dic­tions. That said, I wanted to add three spe­cific pre­dic­tions of my own that par­tially over­lap with Vaniver’s.

  1. I pre­dict with medium con­fi­dence (60%) we’ll see some­thing promised in the near term (where near term is defined on med­i­cal/​clini­cal trial scales not soft­ware com­pany time scales, so next few [3] years) tar­get­ing neuro-de­gen­er­a­tive dis­eases. For ex­am­ple, Alzheimer’s, Park­in­son’s, and strokes all seem like de­cent can­di­dates.

  2. I pre­dict with lower con­fi­dence (~30%) there will be some sort of dis­cus­sion of some­thing longer term that promises to en­hance or aug­ment healthy hu­mans, with no con­crete product or timeline speci­fied. I add the sec­ond clause, even though it makes the over­all pre­dic­tion less likely, be­cause I re­al­ized I’d be sur­prised if an en­hance­ment was in de­vel­op­ment and wanted to spec­ify that as part of my pre­dic­tion.

  3. I pre­dict with lower con­fi­dence (30%) that all prod­ucts dis­cussed will be in­va­sive, where in­va­sive can in­clude things that don’t re­quire surgery but still have to be in­gested/​in­jected.

• An1lam 1 Apr 2019 23:34 UTC

  1 point

  in reply to: Sameerishere's comment on: What are effec­tive strate­gies for miti­gat­ing the im­pact of acute sleep de­pri­va­tion on cog­ni­tion?

  Thanks! Th­ese are good recom­men­da­tions!

  I used to do cold show­ers but stopped a while back. Maybe I’ll re­vive the prac­tice.

  Eat­ing low carb and as lit­tle as you can

  I’m cu­ri­ous about this one. You find that eat­ing low carb/​less keeps you sharper on days when you sleep very lit­tle? I’ve anec­do­tally no­ticed that fast­ing gives me an en­ergy boost on days when I sleep well, al­though at the cost of mak­ing me a bit more jit­tery, but haven’t no­ticed the same effects on days when I don’t sleep enough. Is this purely anec­do­tal or can you share some ar­ti­cles/​pa­pers about this?

• An1lam 1 Apr 2019 23:33 UTC

  1 point

  in reply to: rohinmshah's comment on: What are effec­tive strate­gies for miti­gat­ing the im­pact of acute sleep de­pri­va­tion on cog­ni­tion?

  Good to know I’m not the only one!

  For me, the strat­egy you de­scribe works some­times but fails when I block out my sched­ule for a day to al­low my­self to fo­cus and then un­ex­pect­edly sleep poorly the night be­fore.

• An1lam 1 Apr 2019 23:31 UTC

  1 point

  in reply to: G Gordon Worley III's comment on: What are effec­tive strate­gies for miti­gat­ing the im­pact of acute sleep de­pri­va­tion on cog­ni­tion?

  I re­mem­ber read­ing some­where that med­i­ta­tion has been shown to re­duce the need for sleep but was skep­ti­cal. If there’s any liter­a­ture on this that you con­sider trust­wor­thy, please share!

• An1lam 1 Apr 2019 23:30 UTC

  2 points

  in reply to: CheerfulWarrior's comment on: What are effec­tive strate­gies for miti­gat­ing the im­pact of acute sleep de­pri­va­tion on cog­ni­tion?

  Sit­ting next to a big win­dow and spend­ing a lot of time peo­ple-watch­ing. I don’t un­der­stand why it worked, but I no­ticed it would put me in a rhythm where I would make slow but con­sis­tent progress with my work.

  Go­ing for a long walk similarly seemed to help tem­porar­ily.

  You seem to as­sume that your low­ered abil­ity is caused by sleep de­pri­va­tion. Is that an as­sump­tion? If so, I would en­courage you to track your sleep qual­ity and your cog­ni­tive perfor­mance and see if they re­ally cor­re­late, if you can think of a way to do it.

  This is an as­sump­tion in the sense that I haven’t com­pared any for­mal cog­ni­tive test re­sults be­tween days when I do/​don’t sleep well. How­ever, there are two fac­tors that make me skep­ti­cal that it’s placebo-like, as you seem to be in­di­cat­ing:

  1. I’ll of­ten get up and be sur­prised to find I’m strug­gling more than usual with hard cog­ni­tive tasks, so it can’t just be a self-fulfilling prophecy.

  2. I do feel fairly con­fi­dent that on days when I sleep un­der a cer­tain amount, I strug­gle more to make progress on math & pro­gram­ming, which at the very least mea­sure my abil­ity to con­cen­trate for long pe­ri­ods of time and think with­out ex­ter­nal aids. Speci­fi­cally, I’ve no­ticed that on days when I sleep less, I can’t man­age com­pli­cated trains of thought as well with­out writ­ing stuff down.

  That said, hope­fully get­ting a Dreem as Chris­ti­anKI sug­gested will help me bet­ter mea­sure my sleep qual­ity and get a bet­ter sense of whether low sleep qual­ity ac­tu­ally al­igns with days on which I strug­gle.

• An1lam 1 Apr 2019 23:25 UTC

  1 point

  in reply to: ChristianKl's comment on: What are effec­tive strate­gies for miti­gat­ing the im­pact of acute sleep de­pri­va­tion on cog­ni­tion?

  Thanks, I’ll try this out! Zeo had already been shut down when I went to buy one, so I’m ex­cited to see there’s a similar product on the mar­ket.

  I’m not perfect about avoid­ing blue light but I do usu­ally sleep in a cold room, stop eat­ing well be­fore bed, and wear an eye mask (al­though my room isn’t as dark as it should be).

  Hope­fully the Dreem will still help me get a bet­ter un­der­stand­ing of the fac­tors that im­pact my sleep though.

• An1lam 31 Mar 2019 23:31 UTC

  7 points

  on: An1lam’s Short Form Feed

  I’ve re­cently been ob­sess­ing over the idea of try­ing to “make math more like pro­gram­ming”. I’m not sure if it’s just be­cause I feel fluent at pro­gram­ming and still not very fluent at ab­stract math or also be­cause pro­gram­ming re­ally does have a feed­back loop that you don’t get in math.

  Re­gard­less, based on my read­ing it seems like there’s a gen­eral con­sen­sus in math that even the most mod­ern the­o­rem provers, like Lean and Coq, are much less effi­cient than typ­i­cal “in­for­mal” math rea­son­ing. That said, I won­der if this ig­nores some of the benefits that pro­gram­mers get from writ­ing in a for­mal lan­guage, i.e. au­to­matic re­fac­tor­ing tools, fast feed­back loops, and code anal­y­sis/​search tools. Also, it seems like a suffi­ciently user-friendly math the­o­rem prov­ing tool could be use­ful for ed­u­ca­tion. If kids can learn how to pro­gram in Javascript, I have to be­lieve they can learn to prove the­o­rems, even if the learn­ing curve’s rel­a­tively steep.

  Maybe once I play around with Lean more, I’ll change my mind, but for now, I’m stick­ing to my con­trar­ian view­point.

[Question] What are effec­tive strate­gies for miti­gat­ing the im­pact of acute sleep de­pri­va­tion on cog­ni­tion?

• An1lam 31 Mar 2019 18:15 UTC

  1 point

  on: Why Plan­ning is Hard: A Mul­ti­faceted Model

  The dis­cus­sion of plan­ning across do­mains seems to ig­nore the fact that of­ten the best solu­tion to plan­ning a pro­ject in a do­main with which you’re not fa­mil­iar is to hire/​get help from an­other per­son. Of course, once you turn plan­ning into a multi-per­son ac­tivity, which I think any com­pre­hen­sive model of plan­ning should treat it as, you also need to fac­tor in un­cer­tainty about oth­ers’ plans, which com­pli­cates the model quite a bit.

So you want to be a wizard

• An1lam 6 Feb 2019 22:05 UTC

  7 points

  in reply to: Darkar Dengeno's comment on: If Ra­tion­al­ity can be likened to a ‘Mar­tial Art’, what would be the Forms?

  I like Pre­dic­tionBook.com for this sort of thing. You make pre­dic­tions with con­fi­dences and then set a date on which they’ll be ready to be judged. By de­fault, pre­dic­tions you make are pub­lic, but you can also eas­ily make pri­vate pre­dic­tions.

• An1lam 2 Feb 2019 23:53 UTC

  4 points

  on: Build­ing up to an In­ter­nal Fam­ily Sys­tems model

  Really en­joyed the post, thanks!

  I started the Ear­ley book and it’s definitely a strug­gle. I usu­ally can han­dle “soft skills” books like this one with­out get­ting frus­trated by the vague, hand-wavy mod­els—I re­ally en­joyed Gendlin’s Fo­cus­ing, for ex­am­ple—but this one’s been es­pe­cially hard. That said, hav­ing your model in mind while I’m read­ing has kept me go­ing as I’m us­ing it as a sort of Rosetta’s stone for some of Ear­ley’s claims.