Alicorn

• Alicorn 10 Aug 2019 18:03 UTC

  6 points

  in reply to: rmoehn's comment on: Which of these five AI al­ign­ment re­search pro­jects ideas are no good?

  Are you aware that peo­ple’s votes are worth differ­ent amounts? I do not think there’s a way to vote less than one’s de­fault vote amount.

• Alicorn 29 Jun 2019 23:37 UTC

  58 points

  on: Do chil­dren lose ‘childlike cu­ri­os­ity?’ Why?

  I have a kiddo whose “why phase” is in full swing and I am not ac­tu­ally con­fi­dent that it’s mo­ti­vated by cu­ri­os­ity. It’s also not the most effi­cient way to learn things, or even the most effi­cient sim­ple way (that’d prob­a­bly be some­thing like “tell me stuff about $TOPIC”), nor is it ob­vi­ously geared at that goal.

  In par­tic­u­lar, my kid (I don’t know how com­mon this is) will typ­i­cally for­mu­late his ques­tions by re-gram­ma­tiz­ing what­ever state­ment was most re­cently made in his vicinity (“it’s a nice day” “why is it a nice day?” “be­cause it’s a good tem­per­a­ture” “why is it a good tem­per­a­ture?”). This will sure keep the con­ver­sa­tion go­ing, but:

  • He doesn’t re­tain the in­for­ma­tion well, some­times ask­ing the ex­act same ques­tion more than once in a pe­riod of just a few min­utes, even when the an­swer isn’t com­pli­cated com­pared to things he un­der­stands eas­ily.

  • He doesn’t seem to care what kind of an­swer he gets—he will pro­ceed al­most iden­ti­cally if the an­swer to the tem­per­a­ture ques­tion above has to do with it hav­ing been a similar tem­per­a­ture yes­ter­day, or about the sea­son, or about cloud cover, or if the an­swer is “I don’t know” (he’ll ask “why do you don’t know”).

  • He hasn’t no­ticed any com­mon pat­terns that end the line of ques­tion­ing (if he ever asks why he did some­thing, he gets, “I don’t know, why did you do that?”, but hasn’t given up on such ques­tions).

  • Be­cause of how he gen­er­ates new ques­tions, he can be led around con­cept-space in what­ever way is most con­ve­nient for his in­ter­locu­tor. He doesn’t cir­cle back to stuff he’s been in­ter­ested in be­fore ex­cept when he’s re­peat­ing ques­tions he for­got and set­tling for the same an­swers as last time ver­ba­tim. There isn’t a sense, talk­ing to him, that he’s aware of the ex­is­tence of a con­cept out there he re­ally wants to grasp.

  This isn’t to say that he isn’t cu­ri­ous, but I don’t think the “why phase” is strongly re­lated. When he’s re­ally in­ter­ested in learn­ing about some­thing he wants to go in­ter­act with it. He also has other lan­guage abil­ities that he seems to use when what he wants re­ally is in­for­ma­tion, like “I want to talk about it” and non-why-ques­tions. Why ques­tions seem to be just a but­ton-mash for “make the adults talk to me”.

• Alicorn 17 Jun 2019 6:29 UTC

  8 points

  on: Is there a guide to ‘Prob­lems that are too fast to Google’?

  When I did ju­jitsu we learned to fall. The im­por­tant things are to tuck in your head, and to strike the ground—like, slap it as hard as you can with your hands—be­fore your land­ing to re­duce the force with which you hit.

• Alicorn 16 Jun 2019 5:30 UTC

  12 points

  on: Rea­son­able Explanations

  I’ll start.

  A few years ago, I re­ceived a hand-ad­dressed pack­age with my cor­rect name and ad­dress on it; the re­turn ad­dress was a com­pletely un­fa­mil­iar name in a state I’ve never vis­ited and have no friends in. The con­tents were three As­terix books in the origi­nal French which I had no use for, did not know of any­one who wanted, and could not in fact read.

  N sevraq unq hfrq obbx-fjnc­c­vat jrofvgr Ob­bxzbbpu gb trg zr fbzr cerfragf n ju­vyr cerivb­h­fyl naq unqa’g erzr­zor­erq gb hc­qngr gur nqqerff jura trgg­vat gurfr sbe ure uh­fonaq.

Rea­son­able Explanations

Frac­tional Re­serve Charity

New Rat­fic: Nyssa in the Realm of Possibility

• Alicorn 16 Oct 2018 16:39 UTC

  7 points

  in reply to: TruePath's comment on: Nyoom

  I ex­pect that when I re­place my scooter at the end of its use­ful life I will be able to get a much bet­ter scooter. For me, this one is at what I think is cur­rently a good trade­off be­tween porta­bil­ity (more bat­tery is heav­ier, more torque might re­quire a heav­ier ma­chine too), plau­si­bil­ity as a wheelchair (for use in­doors and on the side­walk rather than among cars), and price (I’d pay more know­ing what I know, but when I got it I didn’t ac­tu­ally re­al­ize quite how big a deal it was go­ing to be for me).

• Alicorn 13 Oct 2018 2:44 UTC

  8 points

  in reply to: gwillen's comment on: Nyoom

  I’ve had tests for Ba­sic Defi­cien­cies in Im­por­tant Stuff, but maybe once I get a new pcp I’ll ask about a phys­i­cal ther­a­pist.

• Alicorn 13 Oct 2018 0:57 UTC

  2 points

  in reply to: Manadnock's comment on: Nyoom

  I haven’t tried phys­i­cal ther­a­pists in par­tic­u­lar and my ini­tial re­ac­tion is pretty con­sid­er­able skep­ti­cism. It seems like this would re­quire a lot of in­vest­ment of time in ap­point­ments, I doubt my in­surance would cover it, and then I’d have to pay at­ten­tion to some sort of in­struc­tion they gave me, all the time when­ever I moved, which, if I could do that, I’d have bet­ter pos­ture.

• Alicorn 12 Oct 2018 14:34 UTC

  17 points

  in reply to: Elo's comment on: Nyoom

  The cat­a­lyst was the Costco scooter and from there it was just a mat­ter of hunt­ing down some­thing with the right specs, but what led me to try­ing it that day? I don’t know, prob­a­bly some com­bi­na­tion of “they were right there and ob­vi­ously free to try” and “I re­ally re­ally wanted to sit down right then” (maybe we parked far away).

  I think it’s not un­re­lated to https://​​www.less­wrong.com/​​posts/​​9ZodFr54Ft­pLThHZh/​​ex­pe­ri­en­tial-pica as a con­cept. If you do some­thing weird, that might just be you be­ing weird, but there could also be an un­der­ly­ing prob­lem you could fix.

Nyoom

• Alicorn 26 Sep 2018 21:28 UTC

  4 points

  in reply to: cousin_it's comment on: AI Read­ing Group Thoughts (2/​?): Re­con­struc­tive Psychosurgery

  It seems un­likely that the mem­ory can be re­con­structed (al­though it could be recre­ated, if the AI hap­pens to know who-all was at the gym that day). Your per­spec­tive makes sense to me but for my part I don’t think that kind of de­tail is im­por­tant to me; it’s okay if I wake up miss­ing a lot of minor epi­sodic mem­o­ries like that.

AI Read­ing Group Thoughts (2/​?): Re­con­struc­tive Psychosurgery

Y Couchinator

AI Read­ing Group Thoughts (1/​?): The Man­date of Heaven

• Alicorn 7 Aug 2018 18:01 UTC

  7 points

  in reply to: servy's comment on: Open Thread Au­gust 2018

  The more naive in­ter­pre­ta­tion of the phrase is in­stead rep­re­sented by “self-aware”, if that’s helpful.

• Alicorn 16 Jul 2018 23:47 UTC

  4 points

  in reply to: jimmy's comment on: Ex­pected Pain Parameters

  Hm, I guess that’s a slightly stronger claim, but I do en­dorse it now that I think about the dis­tinc­tion; not to the point where ev­ery sin­gle such piece of ad­vice should come bun­dled with an un­so­lic­ited painful­ness model, but to the point where if you’re giv­ing painful ad­vice you should have the model on hand.

• Alicorn 15 Jul 2018 4:53 UTC

  8 points

  in reply to: Peter Gerdes's comment on: Ex­pected Pain Parameters

  What I mean is more like “if some­one is sug­gest­ing that you do some­thing painful, they should pre­sent you with a model of why and how that pain is okay”. This doesn’t rule out mis­ap­pro­pri­a­tion—I’m sure cult lead­ers and cer­tain brands of in­ter­per­sonal abusers do it hand­ily, es­pe­cially if they’re weaponiz­ing guilt—but it’s at least ro­bust against generic, opaque com­mands to “suck it up”, and if you go in with that ex­pec­ta­tion you’ll have an op­por­tu­nity to no­tice some­thing is wrong if some­one tells you that you shouldn’t be in pain and your pain is in­valid (they don’t have a model that de­scribes the thing you are in fact feel­ing, so they don’t have a good model of the situ­a­tion as a whole).

Ex­pected Pain Parameters