carado.moe

tammy's blog about
AI alignment, utopia, anthropics, and more;

also on twitter, lesswrong, rss, and at

(2020-04-27 edit: actually Greg Egan already made this argument previously (see Q5 if you have read Permutation City))

(2021-04-28 edit: this post might not be the best job at explaining its idea; see an alternate explanation in this other post)

Limiting Real Universes

The following is an argument for thinking that the set of universes that can "be real" (what that means is covered) is limited.

Notably, not all Tegmark 4 (i.e. mathematically possible universes) are real, nor even all considerable states of universes based on our current physics.

I. Limiting Many-Worlds

A universe being "real" is defined here as "one could observe being in it".

Suppose all possible configurations of particles under our current laws of physics are real.

Then, out of all the universes that contain an exact physical copy of you, the vast majority of them should be universes that do not descend from a coherent history and thus everything that surrounds the copy of you should look like random particle soup.

(If such a distinction even makes sense), then You cannot tell if you're "the original" that comes from a coherent history or if you were "just created" as-is, because your memory could also "just have been created" as-is.

Yet, when you look around, everything looks very coherent.

Therefore, either you're extremely lucky, or only universe-states that descend from a coherent history are real. As per bayesianism, you should think the latter.

II. Limiting Computing Ability

Suppose all universes based on our current physics, but with arbitrary amounts of "computing power" (i.e. how of its stuff can be turned into computers i.e. how much it has stuff) are "real".

Then some of those universes would end up making simulations of random universe-states, some of which happen to contain an exact copy of you.

However, if that were possible, because of the number of amounts of computing powers possible, you should be more likely to exist in one of these randomly created simulations with a universe with much more computing power than ours.

Yet, when you look around, everything looks very coherent.

Therefore, there must be some limit on the amount of computing power universes can have; and then, so that the sheer number of these universe can't compete with the meagre set of history-coherent from which our reality descends, there must either be a limit in the number of initial configurations other universes can have, or on the total computing power allocated to all universes.

Even better: suppose all states of Conway's Game of Life are real. Then out of this infinity, a smaller infinity should happen to be running perfect simulations of subsets of this universe that happen to have you in them. But, reusing my argument again, you observe probably not being in them; therefore there must be a limit on the amount of computing power even universes with other physics have (and, again, a limit on either the number of configurations these other universes would be in, or the total computing power allocated to them all collectively).

III. Conclusion

At this point it seems easier to just assume that only universe-states that descend from our history exist, or at least that the number of such histories is limited.

That certainly seems simpler than imagining there being a set of various systems by which other universes with other rules of physics would have a fixed amount of computing power allocated amongst themselves.

Not that all of this is mostly true even if you're a dualist: even if you have a soul (or equivalent), if there were infinite universes with infinite computing power, there's no reason the soul of you reading this right now should happen to be the soul of the original you and not a soul "created in a just-then created universe-state", unless you also assume complex soul mechanics.