Consider a quantum computer running a classical inference process, specified as a reversible circuit. The inference process is estimating the probability of get

Consider a quantum computer running a classical inference process, specified as a reversible circuit. The inference process is estimating the probability of getting a 1 insead of a 0 from some trial, with trial results provided as a series of inputs into the computation. The output of the process is then written into some specific register. This is meant to be analogous to a human embedded in a quantum world, doing experiments and thinking in classical sort of way and writing down conclusions. Now go compute what happens when, instead of feeding classical trial results into this computation, you feed in a series of qubits each in the state a|0> + b|1>. You will find that the output register ends up storing a state extremely close to the representation of the probability |b|^2, up to some precision limited by the number of samples made available to the inference process. This indicates that "almost all the branches" must be agreeing on this particular value as the probability. That's the sort of way in which quantum mechanics, minus the Born Rule, predicts that almost all agents embedded in a quantum world will conclude that it follows the Born rule.