hairyfigment

• hairyfigment 18 Sep 2017 21:09 UTC

  0 points

  on: Unusual med­i­cal event led to con­clud­ing I was most likely an AI in a simu­lated world

  Sort of re­minds me of that time I missed out on a lu­cid dream be­cause I thought I was in a simu­la­tion. In prac­tice, if you see a glitch in the Ma­trix, it’s always a dream.

  I find it in­ter­est­ing that we know hu­mans are in­clined to an­thro­po­mor­phize, or see hu­man-like minds ev­ery­where. You be­gan by talk­ing about “en­tities”, as if you re­mem­bered this pit­fall, but it doesn’t seem like you looked for ways that your “de­cep­tion” could stem from a non-con­scious en­tity. Of course the real an­swer (sce­nario 1) is ba­si­cally that. You have delu­sions, and their ori­gin lies in a non-con­scious Uni­verse.

• hairyfigment 18 Sep 2017 20:58 UTC

  0 points

  in reply to: Erfeyah's comment on: In­trin­sic prop­er­ties and Eliezer’s metaethics

  The sec­ond set of brack­ets may be the dis­con­nect. If “their” refers to moral val­ues, that seems like a cat­e­gory er­ror. If it refers to sto­ries etc, that still seems like a tough sell. Noth­ing I see about Peter­son or his work looks en­courag­ing.

  Rather than look­ing for value you can sal­vage from his work, or an ‘in­ter­pre­ta­tion con­sis­tent with mod­ern sci­ence,’ please imag­ine that you never liked his ap­proach and ask why you should look at this view­point on moral­ity in par­tic­u­lar rather than any of the other view­points you could ex­am­ine. As­sume you don’t have time for all of them.

  If that still doesn’t help you see where I’m com­ing from, con­sider that re­al­ity is con­stantly chang­ing and “the evolu­tion­ary pro­cess” usu­ally hap­pened in en­vi­ron­ments which no longer ex­ist.

• hairyfigment 18 Sep 2017 20:11 UTC

  0 points

  in reply to: Erfeyah's comment on: In­trin­sic prop­er­ties and Eliezer’s metaethics

  Without us­ing terms such as “ground­ing” or “ba­sis,” what are you say­ing and why should I care?

• hairyfigment 18 Sep 2017 19:48 UTC

  0 points

  in reply to: Erfeyah's comment on: Stupid Ques­tions Septem­ber 2017

  I re­peat: show that none of your neu­rons have con­scious­ness sep­a­rate from your own.

  Why on Earth would you think Searle’s ar­gu­ment shows any­thing, when you can’t es­tab­lish that you aren’t a Chi­nese Gym? In or­der to even cast doubt on the idea that neu­rons are peo­ple, don’t you need to rely on func­tion­al­ism or a similar premise?

• hairyfigment 18 Sep 2017 6:01 UTC

  0 points

  in reply to: Erfeyah's comment on: Stupid Ques­tions Septem­ber 2017

  What about it seems worth re­fut­ing?

  The Zom­bie se­quence) may be re­lated. (We’ll see if I can ac­tu­ally link it here.) As far as the Chi­nese Room goes:

  • I think a nec­es­sary con­di­tion for con­scious­ness is ap­prox­i­mat­ing a Bayesian up­date. So in the (ridicu­lous) ver­sion where the rules for speak­ing Chi­nese have no abil­ity to learn, they also can’t be con­scious.

  • Searle talks about “un­der­stand­ing” Chi­nese. Now, the way I would in­ter­pret this word de­pends on con­text—that’s how lan­guage works—but nor­mally I’d in­cline to­wards a Bayesian in­ter­pre­ta­tion of “un­der­stand­ing” as well. So this again might de­pend on some­thing Searle left out of his sce­nario, though the ques­tion might not have a fixed mean­ing.

  • Some ver­sions of the “Chi­nese Gym” have many peo­ple work­ing to­gether to im­ple­ment the al­gorithm. Now, your neu­rons are all tech­ni­cally al­ive in one sense. I gen­uinely feel un­sure how much con­scious­ness a sin­gle neu­ron can have. If I de­cide to claim it’s com­pa­rable to a man blindly fol­low­ing rules in a room, I don’t think Searle could re­fute this. (I also don’t think it makes sense to say one neu­ron alone can un­der­stand Chi­nese; neu­rol­o­gists, feel free to cor­rect me.) So what is his ar­gu­ment sup­posed to be?

• hairyfigment 16 Sep 2017 22:44 UTC

  0 points

  in reply to: DragonGod's comment on: Open thread, Septem­ber 11 - Septem­ber 17, 2017

  Do you know what the Elec­toral Col­lege is? If so, see here:

  The sin­gle most im­por­tant rea­son that our model gave Trump a bet­ter chance than oth­ers is be­cause of our as­sump­tion that pol­ling er­rors are cor­re­lated.

• hairyfigment 16 Sep 2017 19:35 UTC

  0 points

  in reply to: DragonGod's comment on: Open thread, Septem­ber 11 - Septem­ber 17, 2017

  Ar­guably claims about Don­ald Trump win­ning enough states—but Nate Silver didn’t as­sume in­de­pen­dence, and his site still gave the out­come a low prob­a­bil­ity.

• hairyfigment 19 Aug 2017 22:41 UTC

  0 points

  in reply to: ImmortalRationalist's comment on: Open thread, Jul. 17 - Jul. 23, 2017

  Not ex­actly. MIRI and oth­ers have re­search on log­i­cal un­cer­tainty, which I would ex­pect to even­tu­ally re­duce the sec­ond premise to in­duc­tion. I don’t think we have a clear plan yet show­ing how we’ll reach that level of prac­ti­cal­ity.

  Jus­tify­ing a not-su­per-ex­po­nen­tially-small prior prob­a­bil­ity for in­duc­tion work­ing feels like a cat­e­gory er­ror. I guess we might get a kind of jus­tifi­ca­tion from bet­ter un­der­stand­ing Teg­mark’s Math­e­mat­i­cal Macro­cosm hy­poth­e­sis—or, more likely, un­der­stand­ing why it fails. Such an ar­gu­ment will prob­a­bly lack the in­tu­itive force of ‘Clearly the prior shouldn’t be that low.’

• hairyfigment 19 Aug 2017 22:31 UTC

  0 points

  in reply to: ImmortalRationalist's comment on: What Are The Chances of Ac­tu­ally Achiev­ing FAI?

  I would only ex­pect the lat­ter if we started with a hu­man-like mind. A psy­chopath might care enough about hu­mans to tor­ture you; an uFAI not built to mimic us would just kill you, then use you for fuel and build­ing ma­te­rial.

  (At­tempt­ing to pro­duce FAI should the­o­ret­i­cally in­crease the prob­a­bil­ity by try­ing to make an AI care about hu­mans. But this need not be a sig­nifi­cant in­crease, and in fact MIRI seems well aware of the prob­lem and keen to sniff out er­rors of this kind. In the­ory, an uFAI could de­cide to keep a few hu­mans around for some rea­son—but not you. The chance of it want­ing you in par­tic­u­lar seems effec­tively nil.)

• hairyfigment 20 Mar 2017 18:32 UTC

  0 points

  in reply to: SnowSage4444's comment on: Dou­ble Crux — A Strat­egy for Re­solv­ing Disagreement

  Yes, but as it hap­pens that kind of differ­ence is un­nec­es­sary in the ab­stract. Be­sides the point I men­tioned ear­lier, you could have a log­i­cal set of as­sump­tions for “self-hat­ing ar­ith­metic” that proves ar­ith­metic con­tra­dicts it­self.

  Com­pletely un­nec­es­sary de­tails here.

• hairyfigment 14 Mar 2017 1:10 UTC

  0 points

  in reply to: gjm's comment on: Dou­ble Crux — A Strat­egy for Re­solv­ing Disagreement

  Not if they’re suffi­ciently differ­ent. Even within Bayesian prob­a­bil­ity (tech­ni­cally) we have an ex­am­ple in the hy­po­thet­i­cal lem­ming race with a strong Gam­bler’s Fal­lacy prior. (“Lem­ming” be­cause you’d never meet a species like that un­less some­one had played games with them.)

  On the other hand, if an episte­molog­i­cal dis­pute ac­tu­ally stems from fac­tual dis­agree­ments, we might ap­proach the prob­lem by look­ing for the ac­tual rea­sons peo­ple adopted their differ­ent be­liefs be­fore hav­ing an ex­plicit episte­mol­ogy. Dis­cussing a re­li­gious be­liever’s faith in their par­ents may not be pro­duc­tive, but at least progress seems math­e­mat­i­cally pos­si­ble.

• hairyfigment 3 Mar 2017 1:29 UTC

  0 points

  in reply to: Elo's comment on: Why Don’t Ra­tion­al­ists Win?

  How could cor­rect­ing gram­mar be good epistemics? The only ques­tion of fact there is a prac­ti­cal one—how var­i­ous peo­ple will re­act to the gram­mar com­ing out of your word-hole.

• hairyfigment 13 Feb 2017 21:31 UTC

  0 points

  in reply to: entirelyuseless's comment on: Forc­ing An­throp­ics: Boltz­mann Brains

  1. I’m us­ing prob­a­bil­ity to rep­re­sent per­sonal un­cer­tainty, and I am not a BB. So I think I can le­gi­t­i­mately as­sign the the­ory a dis­tri­bu­tion to rep­re­sent un­cer­tainty, even if be­liev­ing the the­ory would make me more un­cer­tain than that. (Note that if we try to in­clude rad­i­cal log­i­cal un­cer­tainty in the dis­tri­bu­tion, it’s hard to ar­gue the num­bers would change. If a uniform dis­tri­bu­tion “is wrong,” how would I know what I should be as­sign­ing high prob­a­bil­ity to?)

  2. I don’t think you as­sign a 95% chance to be­ing a BB, or even that you could do so with­out se­vere men­tal ill­ness. Be­cause for starters:

  3. Hu­mans who re­ally be­lieve their ac­tions mean noth­ing don’t say, “I’ll just pre­tend that isn’t so.” They stop func­tion­ing. Per­haps you meant the bar is liter­ally 5% for mean­ingful ac­tion, and if you thought it was 0.1% you’d stop typ­ing?

  4. I would agree if you’d said that evolu­tion hard­wired cer­tain premises or ap­prox­i­mate pri­ors into us ‘be­cause it was use­ful’ to evolu­tion. I do not be­lieve that hu­mans can use the sort of pas­calian rea­son­ing you claim to use here, not when the is­sue is BB or not BB. Nor do I be­lieve it is in any way nec­es­sary. (Also, the link doesn’t make this clear, but a true prior would need to in­clude con­di­tional prob­a­bil­ities un­der all the­o­ries be­ing con­sid­ered. Hu­mans, too, start life with a sketch of con­di­tional prob­a­bil­ities.)

• hairyfigment 13 Feb 2017 8:41 UTC

  0 points

  in reply to: Douglas_Knight's comment on: In­ter­view with Nas­sim Taleb ‘Trump makes sense to a gro­cery store owner’

  OK, they gave him a greater chance than I thought of win­ning the pop­u­lar vote. I can’t tell if that ap­plies to the polls-plus model which they ac­tu­ally seemed to be­lieve, but that’s not the point. The point is, they had a model with a lot of un­cer­tainty based on rec­og­niz­ing the world is com­pli­cated, they ex­plic­itly as­signed a dis­turb­ing prob­a­bil­ity to the ac­tual out­come, and they praised Trump’s state/​Elec­toral Col­lege strat­egy for that rea­son.

• hairyfigment 13 Feb 2017 8:02 UTC

  0 points

  in reply to: entirelyuseless's comment on: Forc­ing An­throp­ics: Boltz­mann Brains

  Seems like you’re us­ing a con­fus­ing defi­ni­tion of “be­lieve”, but the point is that I dis­agree about our rea­sons for re­ject­ing the claim that you’re a BB.

  Note that ac­cord­ing to your rea­son­ing, any the­ory which says you’re a BB must give us a uniform dis­tri­bu­tion for all pos­si­ble ex­pe­riences. So ra­tio­nally com­ing to as­sign high prob­a­bil­ity to that the­ory seems nearly im­pos­si­ble if your ex­pe­rience is not ac­tu­ally ran­dom.

• hairyfigment 11 Feb 2017 22:51 UTC

  0 points

  in reply to: entirelyuseless's comment on: Forc­ing An­throp­ics: Boltz­mann Brains

  One man’s modus po­nens is an­other man’s modus tol­lens. I don’t even be­lieve that you be­lieve the con­clu­sion.

• hairyfigment 9 Feb 2017 23:06 UTC

  0 points

  in reply to: Erfeyah's comment on: Stupid Ques­tions Fe­bru­ary 2017

  I am com­pletely ig­nor­ing the very im­por­tant part of defin­ing God

  That is in­deed the chief prob­lem here. I’m as­sum­ing you’re talk­ing about the prior prob­a­bil­ity which we have be­fore look­ing at the ev­i­dence.

• hairyfigment 9 Feb 2017 11:33 UTC

  2 points

  in reply to: MrMind's comment on: In­ter­view with Nas­sim Taleb ‘Trump makes sense to a gro­cery store owner’

  Not to put too fine a point on it, but I bet all of my money he didn’t pre­dict Trump would win the Elec­toral Col­lege while los­ing the pop­u­lar vote by mil­lions. That ar­ti­cle gives no hint he even knows it hap­pened. Mean­while, Five Thirty-Eight sup­pos­edly based most of the prob­a­bil­ity mass they as­signed to Trump win­ning on a tech­ni­cal EC vic­tory, and they said he had one chance in four (one in three ear­lier).

  The fact that a 25% chance can hap­pen with po­ten­tially dev­as­tat­ing con­se­quences is roughly why Trump the out­sider might blow up Amer­ica, though I’d give that no more than a 13% chance.

• hairyfigment 8 Feb 2017 18:24 UTC

  0 points

  in reply to: entirelyuseless's comment on: Forc­ing An­throp­ics: Boltz­mann Brains

  I don’t think that’s de­scrip­tively true at all. Re­gard­less of whether or not I see a use­ful way to ad­dress it, I still wouldn’t ex­pect to dis­solve mo­men­tar­ily with no warn­ing.

  Now, this may be be­cause hu­mans can’t eas­ily be­lieve in novel claims. But “my” ex­pe­rience cer­tainly seems more co­her­ent than one would ex­pect a BB’s to seem, and this calls out for ex­pla­na­tion.

• hairyfigment 7 Feb 2017 0:22 UTC

  0 points

  in reply to: whpearson's comment on: Why I’m work­ing on satis­fic­ing in­tel­li­gence aug­men­ta­tion (and not blog­ging about it much)

  If it can’t be solved, how will MIRI know?

  For one, they wouldn’t find a sin­gle ex­am­ple of a solu­tion. They wouldn’t see any fsck­ing hu­man be­ings main­tain­ing any goal not defined in terms of their own per­cep­tions—eg, mak­ing oth­ers happy, hav­ing an his­tor­i­cal ar­ti­fact, or vis­it­ing a place where some event ac­tu­ally hap­pened—de­spite chang­ing their un­der­stand­ing of our world’s fun­da­men­tal re­al­ity.

  If I try to in­ter­pret the rest of your re­sponse char­i­ta­bly, it looks like you’re say­ing the AGI can have goals wholly defined in terms of per­cep­tion, be­cause it can avoid wire­head­ing via satis­fic­ing. That seems in­com­pat­i­ble with what you said be­fore, which again in­voked “some ab­stract no­tion of good and bad” rather than sen­sory data. So I have to won­der if you un­der­stand any­thing I’m say­ing, or if you’re con­flat­ing on­tolog­i­cal crises with some less im­por­tant “paradigm shift”—some­thing, at least, that you have made no case for car­ing about.