I think most responses trivializes and ridicules a potentially interesting and very valid question.
Quantum mechanics works on ensembles or statistics, and that very framework simply becomes ambigous at best when you consider cosmological theories or open systems. Quantum mechanics as it's verified mainly predicts scatting matrices, which are perfectly well defined observables as long as you look at a subsystem, such as a collider embraced by a classical laboratory. Here there is no problem to encode the hilbert space and generate statistics in the environmnet.
But now instead consider that we aren't looking into an atom, we are looking into our environment. Here we can not get away with the same trick.
This is not really a stupid question at all. It has to do with how to define proper observables and how the observer can verify his own theories. If you think a little bit about this, things we do get away with for scattering experiments will not work out in a cosmologicla model.
Smolin has in that paper (and in other talks
http://pirsa.org/08100049/) made the case that quantum mechanics and in fact ant theory with a fixed timeless state space, is likely to be the limiting case of another model (the limit where you consider a subsystem, and the theory lives in the environment). Smolins arguments and ways might not be the ultimate arguments but at least it's an example of reasoning.
/Fredrik