# Stuart_Armstrong

Karma: 20,226
NewTop
Page 1
• Again, we have to be clear about the ques­tion. But if it’s “what pro­por­tions of ver­sions of me are likely to be in a large uni­verse”, then the an­swer is close to 1 (which is the SIA odds). Then you up­date on your birthrank, no­tice, to your great sur­prise, that it is suffi­ciently low to ex­ist in both large and small uni­verses, so up­date to­wards small and end up at 50:50.

• So even with meta-prefer­ences, likely there are mul­ti­ple ways

Yes, al­most cer­tainly. That’s why I want to pre­serve all meta-prefer­ences, at least to some de­gree.

• No, the event “we sur­vived” is “we (the ac­tual peo­ple now con­sid­er­ing the an­thropic ar­gu­ment and past xrisks) sur­vived”.

Over enough draws, you have .

So we up­date the lot­tery odds based on whether we win or not; we up­date the dan­ger odds based on whether we live. If we die, we alas don’t get to do much up­dat­ing (though note that we can con­sider hy­po­thet­i­cal with bets that pay out to sur­viv­ing rel­a­tives, or have a chance of re­viv­ing the hu­man race, or what­ever, to get the up­dates we think would be cor­rect in the wor­lds where we don’t ex­ist).

• The DA, in it’s SSA form (where it is rigor­ous) comes as a pos­te­rior ad­just­ment to all prob­a­bil­ities com­puted in the way above—it’s not an ar­gu­ment that doom is likely, just that doom is more likely than ob­jec­tive odds would im­ply, in a pre­cise way that de­pends on fu­ture (and past) pop­u­la­tion size.

How­ever my post shows that the SSA form does not ap­ply to the ques­tion that peo­ple gen­er­ally ask, so the DA is wrong.

• There are two ver­sions of the DA; the first is “we should roughly be in the mid­dle”, and the sec­ond is “our birth rank is less likely if there were many more hu­mans in the fu­ture”.

I was more think­ing of the sec­ond case, but I’ve changed the post slightly to make it more com­pat­i­ble with the first.

• Maybe: larger refer­ence classes make the uni­verses more likely, but make it less likely that you would be a spe­cific mem­ber of that refer­ence class, so when you up­date on who you are in the class, the two effects can­cel out.

More con­cep­tu­ally: in SAI, the defi­ni­tion of refer­ence class com­mutes with re­stric­tions on that refer­ence class. So it doesn’t mat­ter if you take the refer­ence class of all hu­mans, then spe­cial­ise to the ones al­ive to­day, then spe­cial­ise to you; or take the refer­ence class of all hu­mans al­ive to­day, then spe­cial­ise to you; or just take the refer­ence class of you. SIA is, in a sense, sen­si­ble with re­spect to up­dat­ing.

Does that help?

• You are cor­rect, I dropped a in the proof, thanks! Put it back in, and the proof is now shorter.

• SSA is not refer­ence class in­de­pen­dent. If it uses , then the SSA prob is (rather that ), which is , which is not in­de­pen­dent of (con­sider dou­bling the size of in one world only—that makes that world less likely rel­a­tive to all the oth­ers).

• This is not the stan­dard sleep­ing beauty para­dox, as the in­for­ma­tion you are cer­tain to get does not in­volve any am­ne­sia or du­pli­ca­tion be­fore you get it.

• I was mainly won­der­ing since the web­page must run the la­tex code and throw an er­ror if it de­cides an ex­pres­sion is malformed (and then not show the out­put), whether it would be easy to see the fact there was an er­ror?