= 27d567bc2d48d9f40e9ccc20ea370b15)
MBlume comments on The Fabric of Real Things - Less Wrong
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (305)
I am confused by these posts. On one hand, Eliezer argues for an account of causality in terms of probability, which as we know are subjective degrees of belief. So we should be able to read off whether X thinks A causes B from looking at conditional probabilities in X's map.
But on the other hand, he suggests (not completely sure this is his view from the article) that the universe is actually made of cause and effect. I would think that the former argument instead suggests causality is "subjectively objective". Just as with probability, causality is fundamentally an epistemic relation between me and the universe, despite the fact that there can be widespread agreement on whether A causes B. Of course, I can't avoid cancer by deciding "smoking doesn't cause cancer", just as I can't win the lottery by deciding that my probability of winning it is .9.
For instance, how would an omniscient agent decide if A causes B according Eliezer's account of Pearl? I don't think they would be able to, except maybe in cases where they could count frequencies as a substitute for using probabilities.
OK, let's say you're looking down at a full printout of a block universe. Every physical fact for all times specified. Then let's say you do Solomonoff induction on that printout -- find the shortest program that will print it out. Then for every physical fact in your printout, you can find the nearest register in your program it was printed out of. And then you can imagine causal surgery -- what happens to your program if cosmic rays change that register at that moment in the run. That gives you a way to construe counterfactuals, from which you can get causality.
ETA: There's still some degrees of freedom in how this gets construed though. Like, what if the printout I'm compressing has all its info time-reversed -- it starts out with details about what we'd call the future, then the present, then the past. Then I'd imagine that the shortest program that'd print that out would process everything forward, store it in an accumulator, then run a reversal on that accumulator to print it out, the problem being that the registers printed out from might be downstream from where the value was. It seems like you need some extra magic to be sure of what you mean by "pretend this fact here had gone the other way"