Suppose in QM you have a wavefunction which recognizably evolves into a superposition of wavefunctions. I'll write that psi0, the initial wavefunction, becomes m.psi' + n.psi'', where m and n are coefficients, and psi' and psi'' are basis wavefunctions. Something slightly analogous to the MWI interpretation of this, could be seen in a Turing machine which started with one copy of a bitstring, PSI0, and which replaced it with M copies of the bitstring PSI' and N copies of the bitstring PSI''. That would be a deterministic computation which replaces one world, the single copy of PSI0, with many worlds, the multiple copies of PSI' and PSI''. So it's straightforward enough for a deterministic state machine to invent rules corresponding to a proliferation of worlds. In fact, in the abstract theory of computation, this is one of the standard ways to model nondeterministic computation - have a deterministic computation which deterministically produces all the possible paths that could be produced by the nondeterministic process. However, the way that QM works, and thus the way that a MWI theory would have to work, is rather more complicated, because the coefficients are complex numbers, the probabilities (which one might suppose correspond to the number of copies of each world) are squares of the absolute values of those complex numbers, and probability waves can recombine and destructively interfere, so you would need worlds / bitstrings to be destroyed as well as created. In particular, it seems that you couldn't reproduce QM with a setup in which the only fact about each world / bitstring was the number of current copies - you need the "phase information" (angle in the complex plane) of the complex numbers, in order to know what the interference effects are. So your Turing machine's representation of the state of the multiverse would be something like: (complex coefficient associated with the PSI' worlds) (list of M copies of the PSI' bitstring); (complex coefficie