Richard_Kennaway

• Richard_Kennaway 6 Jul 2020 22:45 UTC

  4 points

  in reply to: shminux’s comment on: Causal­ity and its harms

  Imag­ine the world as fully de­ter­minis­tic. Then there is no “real causal­ity” to speak of, ev­ery­thing is set in stone, and there is no differ­ence be­tween cause and effect.

  If cau­sa­tion is un­der­stood in terms of coun­ter­fac­tu­als — X would have hap­pened if Y had hap­pened — then there is still a differ­ence be­tween cause and effect. A model of a world im­plies mod­els of hy­po­thet­i­cal, coun­ter­fac­tual wor­lds.

• Richard_Kennaway 6 Jul 2020 22:12 UTC

  2 points

  on: Has any­one writ­ten up a con­sid­er­a­tion of Downs’s “Para­dox of Vot­ing” from the per­spec­tive of MIRI-ish de­ci­sion the­o­ries (UDT, FDT, or even just EDT)?

  the chance that your vote (along with ev­ery­one else’s) would be pivotal be­cause the mar­gin was 1 vote,

  I have never un­der­stood this crite­rion for your vote “mat­ter­ing”. It has the con­se­quence that if (as will al­most always be the case for a large elec­torate) the win­ner has a ma­jor­ity of at least 3, then no-one’s vote mat­tered. If a com­mit­tee of 5 peo­ple votes 4 to 1, then no-one’s vote mat­tered. Two votes mat­tered, but no-one’s vote mat­tered. If one of the yes vot­ers had stayed at home that day, then ev­ery yes vote would mat­ter, but the no vote wouldn’t mat­ter.

  This does not seem like a sen­si­ble con­cept to at­tach to the word “mat­ter”. If some­one on that com­mit­tee was very anx­ious that the vote should go they way they would like, they will have done ev­ery­thing they could to per­suade ev­ery other per­suad­able mem­ber to vote their way. Far from no-one’s vote mat­ter­ing, ev­ery vote in that situ­a­tion mat­ters. This is a fre­quent oc­cur­rence in par­li­a­men­tary votes, when there is any doubt be­fore­hand whether the mo­tion will pass, and the re­sult is of great im­por­tance to both sides. In the forth­com­ing US pres­i­den­tial elec­tion, both par­ties will be mak­ing tremen­dous efforts to “get out the vote”. Yet no-one’s vote “mat­ters”?

• Richard_Kennaway 6 Jul 2020 21:57 UTC

  9 points

  on: What are your thoughts on ra­tio­nal wiki

  I once summed up my judge­ment of Ra­tion­alWiki as “Ra­tion­al­ity is their flag, not their method.” I have paid it no at­ten­tion since form­ing that opinion. When I last looked at it, their method was sneer­ing, ev­ery ar­ti­cle was nega­tive, there was no ra­tio­nal con­tent, and no new ideas. It is not worth even the min­utes of my time it would take to look again and see if the leop­ard has changed its spots.

• Richard_Kennaway 3 Jul 2020 16:45 UTC

  3 points

  in reply to: Sunny from QAD’s comment on: The alle­gory of the hospital

  I think that’s what I had in mind. One of the “image en­hance­ment” de­mos takes a heav­ily pix­e­lated face and gives a high qual­ity image of a face — which may look lit­tle like the real face. Another takes the top half of a pic­ture and fills in the bot­tom half. In both cases it’s just mak­ing up some­thing which may be plau­si­ble given the in­put, but no more plau­si­ble than any of countless pos­si­ble ex­trap­o­la­tions.

• Richard_Kennaway 3 Jul 2020 16:29 UTC

  3 points

  in reply to: Sunny from QAD’s comment on: The alle­gory of the hospital

  Even if my guess is wrong (see other com­ment), I think this story works well as it is. It has some­thing of the spirit of Mul­lah Nasred­din.

• Richard_Kennaway 3 Jul 2020 16:15 UTC

  2 points

  in reply to: Sunny from QAD’s comment on: The alle­gory of the hospital

  The in­ter­nal links on your web site are hav­ing the same prob­lem.

• Richard_Kennaway 3 Jul 2020 16:13 UTC

  2 points

  in reply to: Richard_Kennaway’s comment on: The alle­gory of the hospital

  I won­der when some­one in­ves­ti­gat­ing a crime will try feed­ing all the ev­i­dence to some­thing like GPT-3 and ask­ing it to con­tinue the sen­tence “There­fore the guilty per­son is...” Then they pre­sent this as ev­i­dence in court.

• Richard_Kennaway 3 Jul 2020 15:41 UTC

  3 points

  on: The alle­gory of the hospital

  Is this about re­cent de­mos of Hol­ly­wood-level image en­hance­ment, and how they’re not dis­cov­er­ing what’s in the image, but mak­ing stuff up that’s con­sis­tent with it? And similar de­mos with GPT-3, that one might call “text en­hance­ment”?

• Richard_Kennaway 28 Jun 2020 16:18 UTC

  2 points

  in reply to: abramdemski’s comment on: Rad­i­cal Prob­a­bil­ism [Tran­script]

  Jeffrey wanted to han­dle the case where you some­how be­come 90% con­fi­dent of X, in­stead of fully confident

  How does this differ from a Bayesian up­date? You can up­date on a new prob­a­bil­ity dis­tri­bu­tion over X just as you can on a point value. In fact, if you’re up­dat­ing the prob­a­bil­ities in a Bayesian net­work, like you de­scribed, then even if the ev­i­dence you are up­dat­ing on is a point value for some ini­tial vari­able in the graph, the prop­a­ga­tion steps will in gen­eral be up­dates on the new prob­a­bil­ity dis­tri­bu­tions for par­ent vari­ables.

• Richard_Kennaway 27 Jun 2020 10:17 UTC

  2 points

  on: Atem­po­ral Eth­i­cal Obligations

  This is sav­ing your­self from the mob by run­ning ahead of it.

• Richard_Kennaway 26 Jun 2020 20:53 UTC

  11 points

  in reply to: johnswentworth’s comment on: Ab­strac­tion, Evolu­tion and Gears

  I heard it a long long time ago in a physics lec­ture, but I since ver­ified it. The vari­a­tion in where a ball is struck is mag­nified by the ra­tio of (dis­tance to the next col­li­sion) /​ (ra­dius of a ball), which could be a fac­tor of 30. Seven col­li­sions gives you a fac­tor of about 22 billion.

  I also tried the same calcu­la­tion with the mo­tion of gas molecules. If the am­bi­ent grav­i­ta­tional field is varied by an amount cor­re­spond­ing to the dis­place­ment of one elec­tron by one Planck length at a dis­tance equal to the ra­dius of the ob­serv­able uni­verse, I think I got about 30 or 40 col­li­sions be­fore the ex­trap­o­la­tion breaks down.

• Richard_Kennaway 26 Jun 2020 9:37 UTC

  5 points

  in reply to: johnswentworth’s comment on: Ab­strac­tion, Evolu­tion and Gears

  To ex­pand on the billiard ball ex­am­ple, the de­gree of sen­si­tivity is not always re­al­ised. Sup­pose that the con­di­tions around the billiard table are changed by hav­ing a player stand on one side of it rather than the other. The differ­ence in grav­i­ta­tional field is suffi­cient that af­ter a ball has un­der­gone about 7 col­li­sions, its tra­jec­tory will have de­vi­ated too far for fur­ther ex­trap­o­la­tion to be pos­si­ble — the ball will hit balls it would have missed or vice versa. Be­cause of ex­po­nen­tial di­ver­gence, if the change were to move just the cue chalk from one edge of the table to an­other, the pre­dic­tion hori­zon would be not much in­creased.

• Richard_Kennaway 23 Jun 2020 14:40 UTC

  2 points

  in reply to: Signer’s comment on: The “hard” prob­lem of con­scious­ness is the least in­ter­est­ing prob­lem of consciousness

  But if we started with two prob­lems and ended with one, then one of them is solved.

  You won’t es­cape an ex­cess bag­gage charge by putting both your suit­cases into one big case.

• Richard_Kennaway 22 Jun 2020 18:28 UTC

  2 points

  on: Utility need not be bounded

  (I also posted this to the Open Thread—I’m not sure which is more likely to be seen.)

  Since post­ing the OP, I’ve re­vised my pa­per, now called “Un­bounded util­ity and ax­io­matic foun­da­tions”, and elimi­nated all the place­hold­ers mark­ing work still to be done. I be­lieve it’s now ready to send off to a jour­nal. If any­one wants to read it, and es­pe­cially if any­one wants to study it and give feed­back, just drop me a mes­sage. As a taster, here’s the in­tro­duc­tion.

  Sev­eral ax­io­ma­ti­sa­tions have been given of prefer­ence among ac­tions, which all lead to the con­clu­sion that these prefer­ences are equiv­a­lent to nu­mer­i­cal com­par­i­son of a real-val­ued func­tion of these ac­tions, called a “util­ity func­tion”. Among these are those of Ram­sey [11], von Neu­mann and Mor­gen­stern [17], Nash [8], Marschak [7], and Sav­age [13, 14].

  Th­ese ax­io­ma­ti­sa­tions gen­er­ally lead also to the con­clu­sion that util­ities are bounded. (An ex­cep­tion is the Jeffrey-Bolker sys­tem [6, 2], which we shall not con­sider here.) We ar­gue that this con­clu­sion is un­nat­u­ral, and that it arises from a defect shared by all of these ax­iom sys­tems in the way that they han­dle in­finite games. Tak­ing the ax­ioms pro­posed by Sav­age, we pre­sent a sim­ple mod­ifi­ca­tion to the sys­tem that ap­proaches in­finite games in a more prin­ci­pled man­ner. All mod­els of Sav­age’s ax­ioms are mod­els of the re­vised ax­ioms, but the re­vised ax­ioms ad­di­tion­ally have mod­els with un­bounded util­ity. The ar­gu­ments to bounded util­ity based on St. Peters­burg-like gam­bles do not ap­ply to the re­vised sys­tem.

• Richard_Kennaway 22 Jun 2020 18:26 UTC

  8 points

  on: Open & Wel­come Thread—June 2020

  Since post­ing this, I’ve re­vised my pa­per, now called “Un­bounded util­ity and ax­io­matic foun­da­tions”, and elimi­nated all the place­hold­ers mark­ing work still to be done. I be­lieve it’s now ready to send off to a jour­nal. If any­one wants to read it, and es­pe­cially if any­one wants to study it and give feed­back, just drop me a mes­sage. As a taster, here’s the in­tro­duc­tion.

  Sev­eral ax­io­ma­ti­sa­tions have been given of prefer­ence among ac­tions, which all lead to the con­clu­sion that these prefer­ences are equiv­a­lent to nu­mer­i­cal com­par­i­son of a real-val­ued func­tion of these ac­tions, called a “util­ity func­tion”. Among these are those of Ram­sey [11], von Neu­mann and Mor­gen­stern [17], Nash [8], Marschak [7], and Sav­age [13, 14].

  Th­ese ax­io­ma­ti­sa­tions gen­er­ally lead also to the con­clu­sion that util­ities are bounded. (An ex­cep­tion is the Jeffrey-Bolker sys­tem [6, 2], which we shall not con­sider here.) We ar­gue that this con­clu­sion is un­nat­u­ral, and that it arises from a defect shared by all of these ax­iom sys­tems in the way that they han­dle in­finite games. Tak­ing the ax­ioms pro­posed by Sav­age, we pre­sent a sim­ple mod­ifi­ca­tion to the sys­tem that ap­proaches in­finite games in a more prin­ci­pled man­ner. All mod­els of Sav­age’s ax­ioms are mod­els of the re­vised ax­ioms, but the re­vised ax­ioms ad­di­tion­ally have mod­els with un­bounded util­ity. The ar­gu­ments to bounded util­ity based on St. Peters­burg-like gam­bles do not ap­ply to the re­vised sys­tem.

• Richard_Kennaway 20 Jun 2020 20:37 UTC

  2 points

  on: Me­mory is not about the past

  Few ac­tivi­ties are as quintessen­tially hu­man as be­ing on the cusp of fal­ling asleep and sud­denly be as­saulted by a mem­ory that has us re­live an em­bar­rass­ing epi­sode that we thought long for­got­ten.

  Really? *does not raise hand*

• Richard_Kennaway 20 Jun 2020 19:57 UTC

  2 points

  on: When is it Wrong to Click on a Cow?

  “Only one thing is se­ri­ous for all peo­ple at all times. A man may be more aware of it or less aware of it but the se­ri­ous­ness of things will not al­ter on this ac­count.

  “If a man could un­der­stand all the hor­ror of the lives of or­di­nary peo­ple who are turn­ing round in a cir­cle of in­signifi­cant in­ter­ests and in­signifi­cant aims, if he could un­der­stand what they are los­ing, he would un­der­stand that there can be only one thing that is se­ri­ous for him—to es­cape from the gen­eral law, to be free. What can be se­ri­ous for a man in prison who is con­demned to death? Only one thing: How to save him­self, how to es­cape: noth­ing else is se­ri­ous.”

  Gur­d­jieff, as quoted in Ous­pen­sky, “In Search of the Mirac­u­lous”.

• Richard_Kennaway 20 Jun 2020 19:56 UTC

  2 points

  on: When is it Wrong to Click on a Cow?

  Well, what do you want? What will you do to get it?

  Per­son­ally, I have no in­cli­na­tion to read trashy nov­els or watch the Kar­dashi­ans (or in­form my­self of who they might be), so the is­sue of whether to do that does not ex­ist for me.

  When is it wrong to click on a cow? When your bet­ter self (the one that is smarter and bet­ter in­formed than you, your per­sonal co­her­ent ex­trap­o­lated vo­li­tion) would not.

• Richard_Kennaway 17 Jun 2020 11:09 UTC

  2 points

  in reply to: Wei_Dai’s comment on: Tips/​tricks/​notes on op­ti­miz­ing investments

  In­fer­en­tial dis­tance? Or sim­ply knowl­edge dis­tance.

  You lose me at “With port­fo­lio mar­gin”. You’re talk­ing about fi­nan­cial in­stru­ments that, so I un­der­stand, you have a lot of pro­fes­sional ex­pe­rience in us­ing, but I know noth­ing about these things. I googled “box spread fi­nanc­ing”, and it turns out to be a com­pli­cated in­stru­ment in­volv­ing four sep­a­rate op­tions that, I’m still not sure what the pur­pose is. No crit­i­cism of your­self in­tended, but if a com­plete stranger started talk­ing to me about box spread fi­nanc­ing, de­spite it be­ing a real thing I’d as­sume they were tout­ing a scam. I don’t know what “with­draw­ing ex­cess “equity” from my mar­gin ac­count” means, nor the quote from Gold­man Sachs (which would not come to my at­ten­tion any­way).

  And per­son­ally, I’m in the UK and a lot of what you’re talk­ing about is US-spe­cific, but I can’t even tell which parts are and which aren’t. CD? FDIC? I do not know of a UK bank offer­ing more than de­ri­sory in­ter­est on a sav­ings ac­count (typ­i­cally 0.01% for in­stant ac­cess, 0.35% if you never with­draw money), but per­haps the banks I know of (re­tail banks) are not the sort of banks you’re talk­ing about. The Wikipe­dia page for Gold­man Sachs sug­gests it is not in­volved in re­tail bank­ing.

• Richard_Kennaway 13 Jun 2020 8:19 UTC

  2 points

  in reply to: Signer’s comment on: The “hard” prob­lem of con­scious­ness is the least in­ter­est­ing prob­lem of consciousness

  You can’t “make ev­ery­thing be con­scious”. The thing we have ex­pe­rience of and call con­scious­ness works how­ever it works. It is pre­sent wher­ever it is pre­sent. It takes what­ever differ­ent forms it takes. How it works, where it is pre­sent, and what forms it takes can­not be af­fected by point­ing at ev­ery­thing and say­ing “it’s con­scious!”