It’s an interesting idea, with some intuitive appeal. Also reminds me of a science fiction novel I read as a kid, the title of which currently escapes me, so the concept feels a bit mundane to me, in a way. The complexity argument is problematic, though—I guess one could assume some sort of per-universe Kolmogorov weighting of subjective experience, but that seems dubious without any other justification.
Suppose we had a G.O.D. that takes N bits of input, and uses the input as a starting-point for running a simulation. If the input contains more than one simulation-program, then it runs all of them.
Now suppose we had 2^N of these machines, each with a different input. The number of instantiations of any given simulation-program will be higher the shorter the program is (not just because a shorter bit-string is by itself more likely, but also because it can fit multiple times on one machine). Finally, if we are willing to let the number of machines shrink to zero, the same probability distribution will still hold. So a shorter program (i.e. more regular universe) is “more likely” than a longer/irregular one.
It’s an interesting idea, with some intuitive appeal. Also reminds me of a science fiction novel I read as a kid, the title of which currently escapes me, so the concept feels a bit mundane to me, in a way. The complexity argument is problematic, though—I guess one could assume some sort of per-universe Kolmogorov weighting of subjective experience, but that seems dubious without any other justification.
Suppose we had a G.O.D. that takes N bits of input, and uses the input as a starting-point for running a simulation. If the input contains more than one simulation-program, then it runs all of them.
Now suppose we had 2^N of these machines, each with a different input. The number of instantiations of any given simulation-program will be higher the shorter the program is (not just because a shorter bit-string is by itself more likely, but also because it can fit multiple times on one machine). Finally, if we are willing to let the number of machines shrink to zero, the same probability distribution will still hold. So a shorter program (i.e. more regular universe) is “more likely” than a longer/irregular one.
(All very speculative of course.)