# cousin_it comments on [SEQ RERUN] Against Modal Logics

• The over­all mes­sage of the post sounds rea­son­able to me, but doesn’t it ap­ply to Pearl’s causal­ity as well? If the world is built from com­pu­ta­tion rather than causal ar­rows, how do you get to causal ar­rows?

• If the world is built from com­pu­ta­tion rather than causal ar­rows, how do you get to causal ar­rows?

I had to do a bit of search­ing, but it seems that Eliezer (or at least Eliezer_2008) con­sid­ers causal ar­rows to be more fun­da­men­tal than com­pu­ta­tions:

And, if we have “there­fore” back, if we have “cause” and “effect” back—and sci­ence would be some­what for­lorn with­out them—then we can hope to re­trieve the con­cept of “com­pu­ta­tion”. We are not forced to grind up re­al­ity into dis­con­nected con­figu­ra­tions; there can be glue be­tween them. We can re­quire the am­pli­tude re­la­tions be­tween con­nected vol­umes of con­figu­ra­tion space, to carry out some kind of time­less com­pu­ta­tion, be­fore we de­cide that it con­tains the time­less Now of a con­scious mind.

So here’s my un­der­stand­ing of Eliezer_2008′s guess of how all the re­duc­tions would work out: mind re­duces to com­pu­ta­tion which re­duces to causal ar­rows which re­duces to some sort of similar­ity re­la­tion­ship be­tween con­figu­ra­tions, and the uni­verse fun­da­men­tally is a (time­less) set of con­figu­ra­tions and their am­pli­tudes.

In­ter­est­ingly, Pearl him­self doesn’t seem nearly as am­bi­tious about how far to push the re­duc­tion of “causal­ity” and ex­plains that his the­ory

takes the phys­i­cal no­tions of “mechanisms”, “vari­ables”, “mea­sure­ments” and “in­ter­ven­tions” as the ba­sic prim­i­tives.

which bears al­most no re­sem­blance to Eliezer’s idea of re­duc­ing causal­ity to similar­ity.

• I still don’t un­der­stand what Bar­bour’s the­ory ac­tu­ally says, and if it says any­thing at all. It seems to be one of Eliezer’s more bizarre en­dorse­ments.

• Does this ex­plain it for you, or are you look­ing for some­thing more de­tailed?

Bar­bour is spec­u­lat­ing that if we solve the Wheeler-DeWitt equa­tion, we’ll get a sin­gle prob­a­bil­ity dis­tri­bu­tion over the con­figu­ra­tion space of the uni­verse, and all of our ex­pe­riences can be ex­plained us­ing this dis­tri­bu­tion alone. Speci­fi­cally, we don’t need a prob­a­bil­ity dis­tri­bu­tion for each in­stant of time, like in stan­dard QM.

• Yeah, that ap­proach seems over­com­pli­cated to me. We shouldn’t ask whether a chunk of mat­ter or in­for­ma­tion “con­tains” a con­scious mind, we should ask how much it con­tributes to the ex­pe­riences of a con­scious mind. The most ob­vi­ous an­swer is that the con­tri­bu­tion de­pends on how easy it is to com­pute the mind given that chunk of mat­ter or in­for­ma­tion, or vice versa. Of course I’m hand­wav­ing a lot here, but hav­ing ac­tual causal ar­rows in the ter­ri­tory doesn’t seem to be re­quired, you just need laws of physics that are sim­ple to com­pute.