I’m not sure the sibling comment is correct, so I’ll take another stab at it. First, I also think the “this favors the Everett interpretation” argument is inco

I’m not sure the sibling comment is correct, so I’ll take another stab at it. First, I also think the “this favors the Everett interpretation” argument is incoherent, but for different reasons. I won’t be defending that view. But I can try to answer your question about the nature of computation and “where” that computation is taking place. A quantum computer is essentially a random number generator, but with a non-uniform distribution that can be programmatically controlled. Let’s start with the random part: a qubit can be either a 1 or a 0. When first initialized, when you then read it you will get either a zero or a one with 50% probability. (I am oversimplifying to the point of error! This is technically incorrect, but please excuse me for the purpose of a simplified explanation that makes sense to computer scientists rather than quantum physicists.) In the Copenhagen interpretation the wave function randomly collapsed in a way that can be measured as a 0 or as a 1 with 50% probability for either outcome. In the Everett interpretation there are two universes (really, two disjoint sets of universes): one where you see a 0, and the other where you see a 1, and your likelihood of ending up in either is 50%. For example, with three qubits you can read a random value between 0 to 7, where each qubit provides a 0 or 1 of a classical 3-bit binary value. If you're making a D&D simulator, you can take three of these registers (9 qubits total) and read them to get three values between 0 to 7, then increment each and add together to get the simulated value of three 8-sided dice, for the attack value of some game weapon. But so far we're assuming even and independent odds for a 0 or 1 in each qubit. A quantum gate operation lets you modify that probability distribution in interesting ways. Specifically you can take multiple qubits and entangle their state such that the eventual outcome (still random) is constrained to a non-uniform probability distribution. (The details don't matter, but it has to do with the fact that you can add wave amplitude, which is complex-valued and therefore can constructively or destructively interfere.) So instead of reading three separate 1d8 registers and adding them together, we can use quantum gate operations to perform the addition: entangling the state of a new 4 qubit register (large enough to hold the sum of two 3 qubit registers) such that it represents the sum of two separate, randomly chosen with independent probability values between 0..7. The combined sum will be a random value between 0 and 14, but with the value 7 being most likely, 6 or 8 next most likely, etc. (You can see the probabilities here: https://anydice.com https://anydice.com) We then add in the third 3-bit register, getting a 5-qubit result. For good measure, we then add a constant 3 value using quantum gate operations so the result will be what we expect for adding 3 dice rolls, which start counting at 1 rather than 0. Now we have a great random number generator for D&D games. It's a 5-bit register that when you read it will give you a value between 3 and 24, with exactly the probabilities you would expect of three rolls of a fair 8-sided die. You can interpret what this means in Copenhagen vs Everett, but there isn't a way in which it makes material difference. I will say that the Everett interpretation is easier to think about: you setup the entangled operations such that there simply isn't a universe in which the register reads 31, or some value not representative of a 3d8 dice roll, and if you count the number of possible universes under the born rules and holding the rest of the multiverse constant, the counts match the 3d8 dice probability distribution. So you don't know which universe you will end up in when you finally read the register and, under the many worlds interpretation, split off into one of many possibilities. But you know it will be a good 3d8 dice roll for your D&D game. Now imagine you have two 1,024 qubit registers (each with random probability for each bit), and you run a full set of shift-adders to multiply them together into a 2,048 qubit register. If you read the result, it'll be a random non-prime number. Why? Because it's necessarily the product of two 1,024 composites. What's interesting is that you can kinda do the reverse, as Peter Shor figured out how to do in the 90's: start with a 2,048 bit register, run the shift-adder in reverse (really, the method of continued fractions, but I'm going for an intuition pump here), then read the two 1,024 registers. What you will get is two values that when multiplied together will result in the input, the factors. If you put "60" in the input, you will maybe get the outputs (15, 4), or (5, 12), or (2, 30), or (1, 60). It's random which will pop out, but what you get WILL be a factorization of 60. Now put your bank's SSL public key in the register, and read the factors out. The factors are the private key. I don't think any talk of multiverses is needed to understand this. It's simply mucking around with the probability distribution of a random number generator in clever enough ways that the random number generated, whatever it is, will be A solution (perhaps one of many) to your problem. You do this by destructively interfering the register with wave functions that represent non-solutions, and keep doing this until only solutions are left as possible values. Then you read your random number register and know the result is drawn from the solution set.