5 ms·

Thanks for the writeup, I enjoyed it very much. How does Deutsch connect this view to experimental results? Is the idea that all results are realized, or is my

Thanks for the writeup, I enjoyed it very much. How does Deutsch connect this view to experimental results? Is the idea that all results are realized, or is my choice of experiment itself predetermined?

kd5bjo 7y ago
My layman’s understanding is that all results are realized many times. The values that we usually interpret as probabilities are really the proportion of universes in which the event occurs; a 95% probably of an experimental outcome means that if we sampled 100 worlds that are successors of the one in which the experiment occurred, 95 of them would have produced that outcome and 5 wouldn’t. We perceive this as randomness due to the weak anthropoic principle: from our perspective, there is only one universe and things actually occurred in only one way, because there’s no interaction between worlds after the split; at every opportunity, we effectively draw one random result from the distribution because all of the other worlds don’t contain us, they contain copies of us.
- TheOtherHobbes 7y ago
  That means branch counting matters, which means there has to be some way for the universes to communicate the results of branch counting with each other. Otherwise there's no mechanism for probability to arise. If you try to fall back to the position that the probabilities are somehow inherent in the interaction between all the universes, then you may as well dispense with the universes and just say the probabilities are defined by a hidden non-local (i.e. not Bell-violating) interaction in this universe. Because otherwise you still have to explain how a split (or differential propagation or whatever you want to call it) in one location generates an entire universe instantaneously - i.e. faster than c - while preserving state that is beyond the event horizon of the current location. There's a kind of sleight of hand about the argument. You can just about imagine multiple copies of a single particle existing along a foliated timeline. But when MWI people say "What splits is the wavefunction of the universe" they're just handwaving. The wavefunction of the universe doesn't exist as a testable abstraction, and even if it did, are we supposed to believe that it's the only physical entity discovered so far that can ignore relativistic limits? And even if that were true, it implies that the wavefunction of the universe is somehow perfectly deterministic - which is unproven at best. Those are all very strong claims to be making. In fact (IMO) there's a strong whiff of handwavy nonsense about the idea. But I have the luxury of not doing this for a living, and I'm very happy to be proved wrong by people who do. Observer independence - and observer relevance - seems to be a separate problem.
  - kd5bjo 7y ago
    > But I have the luxury of not doing this for a living, and I'm very happy to be proved wrong by people who do. I'm in a similar (but likely even less knowledgable) position; I'm having trouble following your logic, presumably because it's relying on terms and results I'm unfamiliar with. > That means branch counting matters, which means there has to be some way for the universes to communicate the results of branch counting with each other. Otherwise there's no mechanism for probability to arise. I don't know what branch counting is, but I don't see how the various universes need to communicate with each other. Observers in most of the universes will agree with us about the probabilities, as that's what the history of those universes has shown. In a small but nonzero number of them, however, everything will have happened the way that classical physics predicts and they will have never come up with quantum mechanical theory in the first place. > Because otherwise you still have to explain how a split (or differential propagation or whatever you want to call it) in one location generates an entire universe instantaneously - i.e. faster than c - while preserving state that is beyond the event horizon of the current location. I don't see a problem conceptually with envisioning each split like a stress fracture that propagates at the speed of light. As information can't travel faster than that (per relativity), there's no need to envision the two new worlds separating from each other faster than that anyway.
- lisper 7y ago
  Yes, that is almost exactly right. The only issue is that what distinguishes one universe from another is not actually well-defined, so you can't actually "count universes". You have to base the calculation of what proportion of universes experience one outcome versus another on something else. That something else turns out to be "branch weight", which, it turns out, is where the quantum randomness hides in MWI because branch weights can't be measured, not even in principle.
  - kd5bjo 7y ago
    Both you and the other reply refer to counting universes/branches, which is obviously a term of art that I’m not understanding. Where I talk about sampling successor universes, are you just emphasizing that we’re drawing from a continuous probability distribution instead of a discrete one, or is there something else?
    - lisper 7y ago
      > a term of art that I’m not understanding Actually, it's a hand-wavy term that is not well-defined, because "universe" and "branch" and "world" are not well-defined. > sampling successor universes The problem with this is that in order to sample something you have to specify the procedure by which you are going to do the sampling because your results are going to depend on that procedure. So even leaving aside the fact that the thing that you're trying to sample is not even well-defined, there are other issues. Sampling finite sets is straightforward, but for infinite sets it gets tricky [1], and for infinite sets without a total ordering it gets very tricky. Consider, for example, generating a random complex number. There are at least two plausible ways to do this: 1. Generate two random reals x and y, and combine them to produce the random complex number x + iy. 2. Generate two random reals, r and theta and combine them to produce the random complex number r * (sin(theta) + i cos(theta)). Depending on which of these methods you choose, you'll get different-looking distributions. Neither one is "correct". If you're going to talk about sampling worlds you have all these difficulties in spades because worlds are rays in Hilbert spaces, i.e. vector spaces with an infinite number of dimensions. There are an infinite number of ways to generate random distributions over a Hilbert space. The trick is to make an argument for why one particular way is better than all the others. The obvious argument is to present a way of sampling that reproduces the Born rule and argue that it's better because it reproduces the experimental results. But in order to not be begging the question, this sampling procedure can't have the Born rule hidden within it. That is what Wallace and Deutsch claim to have done, but which I (and many others) dispute. [1] https://math.stackexchange.com/questions/997173/can-you-pick-a-random-natural-number-and-a-random-real-number https://math.stackexchange.com/questions/997173/can-you-pick...
lisper 7y ago
> How does Deutsch connect this view to experimental results? MWI makes exactly the same predictions as regular QM, so it's connected to experimental results in the exact same way that regular QM is. The only difference is rhetoric. The canonical interpretation (Copenhagen) says, "The outcome is random, we don't know why, and we can't know why." MWI says, "The outcome is not random, all possible outcomes actually happen in point of physical fact. It appears random because when you make a quantum observation you split into multiple copies of yourself, and every one of those copies has the subjective sensation of seeing a random outcome even thought the actual outcomes were deterministic." So... > Is the idea that all results are realized, Yes. > or is my choice of experiment itself predetermined? No. This is yet another interpretation of QM called superdeterminism. https://en.wikipedia.org/wiki/Superdeterminism https://en.wikipedia.org/wiki/Superdeterminism

kd5bjo 7y ago

My layman’s understanding is that all results are realized many times. The values that we usually interpret as probabilities are really the proportion of universes in which the event occurs; a 95% probably of an experimental outcome means that if we sampled 100 worlds that are successors of the one in which the experiment occurred, 95 of them would have produced that outcome and 5 wouldn’t. We perceive this as randomness due to the weak anthropoic principle: from our perspective, there is only one universe and things actually occurred in only one way, because there’s no interaction between worlds after the split; at every opportunity, we effectively draw one random result from the distribution because all of the other worlds don’t contain us, they contain copies of us.

TheOtherHobbes 7y ago
That means branch counting matters, which means there has to be some way for the universes to communicate the results of branch counting with each other. Otherwise there's no mechanism for probability to arise. If you try to fall back to the position that the probabilities are somehow inherent in the interaction between all the universes, then you may as well dispense with the universes and just say the probabilities are defined by a hidden non-local (i.e. not Bell-violating) interaction in this universe. Because otherwise you still have to explain how a split (or differential propagation or whatever you want to call it) in one location generates an entire universe instantaneously - i.e. faster than c - while preserving state that is beyond the event horizon of the current location. There's a kind of sleight of hand about the argument. You can just about imagine multiple copies of a single particle existing along a foliated timeline. But when MWI people say "What splits is the wavefunction of the universe" they're just handwaving. The wavefunction of the universe doesn't exist as a testable abstraction, and even if it did, are we supposed to believe that it's the only physical entity discovered so far that can ignore relativistic limits? And even if that were true, it implies that the wavefunction of the universe is somehow perfectly deterministic - which is unproven at best. Those are all very strong claims to be making. In fact (IMO) there's a strong whiff of handwavy nonsense about the idea. But I have the luxury of not doing this for a living, and I'm very happy to be proved wrong by people who do. Observer independence - and observer relevance - seems to be a separate problem.
- kd5bjo 7y ago
  > But I have the luxury of not doing this for a living, and I'm very happy to be proved wrong by people who do. I'm in a similar (but likely even less knowledgable) position; I'm having trouble following your logic, presumably because it's relying on terms and results I'm unfamiliar with. > That means branch counting matters, which means there has to be some way for the universes to communicate the results of branch counting with each other. Otherwise there's no mechanism for probability to arise. I don't know what branch counting is, but I don't see how the various universes need to communicate with each other. Observers in most of the universes will agree with us about the probabilities, as that's what the history of those universes has shown. In a small but nonzero number of them, however, everything will have happened the way that classical physics predicts and they will have never come up with quantum mechanical theory in the first place. > Because otherwise you still have to explain how a split (or differential propagation or whatever you want to call it) in one location generates an entire universe instantaneously - i.e. faster than c - while preserving state that is beyond the event horizon of the current location. I don't see a problem conceptually with envisioning each split like a stress fracture that propagates at the speed of light. As information can't travel faster than that (per relativity), there's no need to envision the two new worlds separating from each other faster than that anyway.
lisper 7y ago
Yes, that is almost exactly right. The only issue is that what distinguishes one universe from another is not actually well-defined, so you can't actually "count universes". You have to base the calculation of what proportion of universes experience one outcome versus another on something else. That something else turns out to be "branch weight", which, it turns out, is where the quantum randomness hides in MWI because branch weights can't be measured, not even in principle.
- kd5bjo 7y ago
  Both you and the other reply refer to counting universes/branches, which is obviously a term of art that I’m not understanding. Where I talk about sampling successor universes, are you just emphasizing that we’re drawing from a continuous probability distribution instead of a discrete one, or is there something else?
  - lisper 7y ago
    > a term of art that I’m not understanding Actually, it's a hand-wavy term that is not well-defined, because "universe" and "branch" and "world" are not well-defined. > sampling successor universes The problem with this is that in order to sample something you have to specify the procedure by which you are going to do the sampling because your results are going to depend on that procedure. So even leaving aside the fact that the thing that you're trying to sample is not even well-defined, there are other issues. Sampling finite sets is straightforward, but for infinite sets it gets tricky [1], and for infinite sets without a total ordering it gets very tricky. Consider, for example, generating a random complex number. There are at least two plausible ways to do this: 1. Generate two random reals x and y, and combine them to produce the random complex number x + iy. 2. Generate two random reals, r and theta and combine them to produce the random complex number r * (sin(theta) + i cos(theta)). Depending on which of these methods you choose, you'll get different-looking distributions. Neither one is "correct". If you're going to talk about sampling worlds you have all these difficulties in spades because worlds are rays in Hilbert spaces, i.e. vector spaces with an infinite number of dimensions. There are an infinite number of ways to generate random distributions over a Hilbert space. The trick is to make an argument for why one particular way is better than all the others. The obvious argument is to present a way of sampling that reproduces the Born rule and argue that it's better because it reproduces the experimental results. But in order to not be begging the question, this sampling procedure can't have the Born rule hidden within it. That is what Wallace and Deutsch claim to have done, but which I (and many others) dispute. [1] https://math.stackexchange.com/questions/997173/can-you-pick-a-random-natural-number-and-a-random-real-number https://math.stackexchange.com/questions/997173/can-you-pick...

lisper 7y ago

> How does Deutsch connect this view to experimental results? MWI makes exactly the same predictions as regular QM, so it's connected to experimental results in the exact same way that regular QM is. The only difference is rhetoric. The canonical interpretation (Copenhagen) says, "The outcome is random, we don't know why, and we can't know why." MWI says, "The outcome is not random, all possible outcomes actually happen in point of physical fact. It appears random because when you make a quantum observation you split into multiple copies of yourself, and every one of those copies has the subjective sensation of seeing a random outcome even thought the actual outcomes were deterministic." So... > Is the idea that all results are realized, Yes. > or is my choice of experiment itself predetermined? No. This is yet another interpretation of QM called superdeterminism. https://en.wikipedia.org/wiki/Superdeterminism https://en.wikipedia.org/wiki/Superdeterminism