At some point in the past various of us have argued about whether the simulation argument and / or the multiple worlds interpretation of quantum mechanics implies an "every possible world" (EPW) interpretation, i.e. one in which highly improbable events, laws of physics, etc. obtain.
Stumbled across an interesting if tangential paper that has something to say about this. First some terminology: let's call events that are highly improbable "white rabbits" and universes in which such events happen frequently (or universes with entirely inscrutable laws of physics) "white rabbit worlds."
Let's further adopt the term "djinni" or (to follow Gott's nomenclature) "jinni" to refer to closed time-like (causally cyclic) curves, and "jinn worlds" as worlds (n-dimensional "spacetime" slices of the higher-order spacetime, or rather n-m dimensional phase-space volumes where n is the total dimensionality of the phase space) that contain such causal cycles. In order to explain what this means: these are causally consistent chains of events in which there is no ultimate cause, but rather a closed causal chain that traverses both forward and backward along the time dimension. A peculiarity of this idea is that, in such a world, information "appears" without cause. For example a computer employing a closed time-like curve as a register can compute "hard" problems, but when one examines the execution history of the computer through time one finds that it never actually executes the computation! Cf.:
http://arxiv.org/pdf/gr-qc/0209061
Anyway, "jinni" are these little closed curves of causality in the presence of time travel that are consistent but defy common sense.
David G. Boulware of the University of Washington published this paper in PRD:
http://arxiv.org/abs/hep-th/9207054
...in which he studies the behavior of quantum fields in spaces with closed time-like curves. What he finds is that probabilities are not "conserved", i.e. not unitary, in such spaces. That is, the Feynman sum-over-histories approach always yields precisely 1 --- except when space contains one or more jinn. In such cases, there are quantum events that simply cannot occur.
So: jinn defeat white rabbits. If any world-line through the phase space is cyclic / allowed to self-intersect, the overall phase-space is constrained, presumably to those set of configurations which are of higher probability. The very existence of such causal cycles may indeed be --- meta-paradoxically ;-) --- essential in stabilizing the overall structure of the phase space. It would seem that these cycles act as a kind of strange attractor around which probable configurations (universes) coalesce.
Speculation: it may be that through studying the impact of such closed time-like curves in various spacetimes that we ultimately reconcile Cramer's transactional interpretation (retarded waves moving forward in time, advance waves reaching back to "handshake" on each quantum event, producing a kind of causal contract) of QM with MWI --- and ultimately COMP. Indeed, each retarded wave-advance wave pair *is* a jinni. Cramer doesn't just embrace jinn in his interpretation --- he bases the whole idea on their existence! (FWIW: this seems to me an embarrassment of riches. Why should *every* quantum event require a jinni, when a few --- acting as strange attractors --- might suffice? Though admittedly the latter leads to the questions which few, and why?)
The implication ala Boulware is that if this is a real physical effect, then this provides a kind of global probabilistic censorship that makes the world the predictable place that it is! And --- connectionism --- it's rather ironic that Cramer's transactional hypothesis is based in part on some of Feynman's own speculation, when Feynman probably didn't realize the essential seemingly paradoxical consequences of pairing the histories approach with cyclic causality.
So that's all well and good for physics, but what about the more algorithmic cosmologies? One school of thought regarding the COMP hypothesis is that it is easier to simulate all possible worlds than it is to simulate any subset of them. (Cf. previously-discussed Champernowne machine / "everything" algorithm.) But what if the dynamics of the simulation are such that these jinni exist as a priori structural parameters, "roots" if you will of the computation? In such an environment, "every computable universe" is NOT every possible universe.
Curiouser and curiouser,
jb
