Problems with Many Worlds Interpretation

[JordanL]

This quantum suicide thought experiment doesn't actually provide a way of discerning between MWI and CI, right?

It provides a way to help explain conceptual differences, but does not discern them or define them.

 

[Hurkyl]

In MWI every suicide attempt must fail - you always live in a world, where Russian rulette gun misfired.

This is a level slip -- it's confusing the frog's eye and the bird's eye view. No matter how much you want to pay attention to

P(You live) > 0​

you can't get around the fact that

P(You live | You died) = 0.​

You point out there are "worlds" where people live to attend your funeral. In those same "worlds" you get to attend your funeral dead.

 

[Dmitry67]

Yes, it is a confusion of 2 meanings of "YOU" - YOU as a worldline (frog) and YOU as a tree (Bird). These 2 meanings are different in MWI. "Observer" has 2 meanings as well in MWI.

 

[Hurkyl]

It is very gauche and aggressive of me to resort to reductio ad kick-in-the-balls, after such a dainty intellectual argument, but the line you are taking is just incredibly absurd if you try to apply it to anything real. Reality consists of a series of definite situations.

Yes it is. What does your proposal accomplish other than to make you look like a fool? The variable "X or not X" has only one outcome -- "true". Checking for that result has no power to distinguish between any alternatives at all, let alone the ones under consideration.

 

[mitchell porter]

Hilarious... Hurkyl, if I decide that I will or will not kick you in the balls on the basis of a coin toss, and then I toss a coin and act accordingly, you do not end up in a state of "50% kicked in the balls, 50% not kicked in the balls". At the end, either I left you alone, or you are doubled over in pain.

It is very gauche and aggressive of me to resort to reductio ad kick-in-the-balls, after such a dainty intellectual argument, but the line you are taking is just incredibly absurd if you try to apply it to anything real. Reality consists of a series of definite situations. Maybe there are other realities defined by a different series of situations. But if your model of reality contains no such definiteness anywhere, it is false.

Yes it is. What does your proposal accomplish other than to make you look like a fool?

It is supposed to make you remember the existence of pain, which in turn is supposed to make you realize that reality has definite properties.

Maybe I'm misunderstanding you; but Fredrik asked you directly, if he flips a coin, what is the state of that coin. You could have said that it's actually heads, or actually tails, and you don't know which; you could even have said that it's actually heads in one world, and actually tails in another world; but instead you said that the actual state is a probability distribution.

Your subsequent brag (about how your conceptual facility with mathematics and computers has equipped you to accept things as they are, in all their counterintuitiveness) reinforces my initial suspicion that there's something seriously lacking in your concept of what physics is about. Physics, including quantum mechanics, is supposed to describe the real world of stars and planets and living things. It is not just an abstraction living in your imagination or on your hard drive. One of the features of the real world is that specific things happen.

I have, I hope, been fairly clear about how I think quantum mechanics should be interpreted. Observables are real, wavefunctions are not, they just provide probabilities for the behavior of observables; that is how QM is meant to be used. I have so far failed to discern any similarly clear position from you. You objected when I said that wavefunctions aren't real, but your subsequent remarks have been entirely formal in character, and do not indicate where you think QM connects to reality. Well, there was the nonsensical affirmation that a probability distribution over the possible outcomes of a coin toss is the actual state of the coin. This is where the kick in the balls comes in. It is supposed to make you understand that things actually happen.

 

[Hurkyl]

you could even have said that it's actually heads in one world, and actually tails in another world; but instead you said that the actual state is a probability distribution.

Yes. (Aside: you do notice that, in classical mechanics "probability distribution over N states" is essentially identical to "N worlds (with weights that sum to 1)" right?)

Well, there was the nonsensical affirmation that a probability distribution over the possible outcomes of a coin toss is the actual state of the coin.

I maintain there is no experiment that can possibly be done to distinguish between definite outcomes and indefinite outcomes.

If there is no experiment that can distinguish between possibilities, there is no physical content in the assertion that reality has definite outcomes rather than indefinite outcomes.

If there is no physical content to the assertion that reality has definite outcomes rather than indefinite outcomes, then the claim that you're talking about reality and I'm talking about nonsense is just you blowing smoke.


Nothing QM specific is relevant to anything you've said in this post. You are familiar with probability distributions on classical state spaces, right? I challenge you to find an experiment that can distinguish between reality is a single point of phase space versus reality is a probability distribution on phase space.

 

[mitchell porter]

Yes. (Aside: you do notice that, in classical mechanics "probability distribution over N states" is essentially identical to "N worlds (with weights that sum to 1)" right?)

If you mean "possible worlds" then I think I accept the equivalence.

I maintain there is no experiment that can possibly be done to distinguish between definite outcomes and indefinite outcomes.

If there is no experiment that can distinguish between possibilities, there is no physical content in the assertion that reality has definite outcomes rather than indefinite outcomes.

If there is no physical content to the assertion that reality has definite outcomes rather than indefinite outcomes, then the claim that you're talking about reality and I'm talking about nonsense is just you blowing smoke.

Every second of your waking life tells you that at least one "definite outcome" exists. This whole idea of an "indefinite outcome" is a contradiction. Either something is happening or it isn't happening.

Many worlds makes sense as a belief in "many definite outcomes". It then founders on the problem of reproducing the Born probabilities, but at least there's a definite hypothesis.

 

[Hurkyl]

If you mean "possible worlds" then I think I accept the equivalence.

I don't know what you're trying to connote with "possible". My best guess is that you're using it to mean "reality is in one of these worlds", which is very much not what I'm talking about.

Every second of your waking life tells you that at least one "definite outcome" exists. This whole idea of an "indefinite outcome" is a contradiction. Either something is happening or it isn't happening.

Show me an experiment that can tell the difference. Anything at all. The criterion you're suggesting is to observe:

P(X or not X) = 1​

or possibly

P(X and not X) = 0​

but neither criterion differentiates between X being a definite value that is either "true" or "false" and X being a random variable with sample space {"true", "false"}.

Let me repeat that for emphasis. In an indefinite outcome interpretation of classical mechanics, if X is any proposition tested by experiment, then you are guaranteed to observe either X or not X. "X or not X" is definitely true, whether X is definite or not.

 

[Demystifier]

There is a famous Tegmark's "quantum suicide" experiment. But it may convince only you...

First, it will not "convince" only me, but also all my neighbors living in the branch in which I survive each time.

Second, I doubt that it will really convince anybody; it could also be that I was just lucky to survive each time, or that there is some secret mechanism providing my survival in a deterministic manner (for example, perhaps someone found a way to control the initial conditions of Bohmian hidden variables).

 

[mitchell porter]

I don't know what you're trying to connote with "possible".

I don't want to presuppose the actuality of all the worlds appearing in the probability distribution.

Every second of your waking life tells you that at least one "definite outcome" exists. This whole idea of an "indefinite outcome" is a contradiction. Either something is happening or it isn't happening.

Show me an experiment that can tell the difference. Anything at all.

The criterion you're suggesting is to observe:

P(X or not X) = 1​

or possibly

P(X and not X) = 0​

but neither criterion differentiates between X being a definite value that is either "true" or "false" and X being a random variable with sample space {"true", "false"}.

Let me repeat that for emphasis. In an indefinite outcome interpretation of classical mechanics, if X is any proposition tested by experiment, then you are guaranteed to observe either X or not X. "X or not X" is definitely true, whether X is definite or not.

The radical insanity of your position makes it difficult to rebut. Or rather, it is easy to rebut, but it is difficult to make you acknowledge the rebuttal. I can't say exactly what your problem is, but it has something to do with misuse of formal concepts like "random variable".

First you ask for an experiment that can tell the difference between a definite outcome and an indefinite outcome. I repeat that an indefinite outcome is an incoherent concept, a logical contradiction. The experiment has an outcome, or it doesn't. If it has an outcome, it is necessarily a definite outcome. There is no such thing as an experiment with an "indefinite outcome" in the sense you are discussing.

Your distinction between

X being a definite value that is either "true" or "false" and X being a random variable with sample space {"true", "false"}

appears to be a confusion of levels. I'll just quote Wikipedia:

In probability and statistics, a random variable or stochastic variable is, roughly speaking, a variable whose value results from a measurement on some type of random process. Formally, it is a function from a probability space, typically to the real numbers, which is measurable.

It's as if you're saying "Your experiment can't distinguish between the possibility that X is a value of a function, and the possibility that X is a function."

I am not sure about the following attempt at cognitive debugging, I can only guess at the details of the thought process which causes you to say what you do; but your remark may be arising from an inconsistent use of the concept "random variable". Either you are talking about it as a mathematical concept, or you are using it to refer to a physical quantity. If you are talking about an experiment, then your random variable is a physical quantity, in which case it necessarily satisfies your first option - 'a definite value that is either "true" or "false"'. The "function from a probability space", etc, specifies the mathematical formalism we use to describe the physical variable, and it says something about the properties of the possible values of the physical variable (their probabilities).

 

[Fredrik]

First, it will not "convince" only me, but also all my neighbors living in the branch in which I survive each time.

If that's true, there's no need to have the experiment involve a suicide. Just fire the gun into a barrel of water, or throw away the gun and just measure the spin of a bunch of silver atoms. But does getting the result "up" the first 20 times really make the CI less likely to be true than the MWI? I don't see how, and apparently neither does Tegmark. What he said (I just checked) is that

• a quantum suicide is the only experiment he can think of that allows us to distinguish between MWI and Copenhagen.
• it will only convince the person in front of the gun, not the person behind it.
• the probability that the person in front of the gun will hear a click (indicating that the gun isn't going to fire) is 1.

 

[Hurkyl]

Let P be the phase space of a universe-describing physical theory. Any theory will work here (even hypothetical yet-to-be-discovered theories)

Suppose we've solved the measurement problem for P -- for a configuration x, we are able to ask the question

E(x) = "does the configuration x include a person named mitchell porter that flipped a coin?"​

Similarly,

H(x) = "does the configuration x include a person named mitchell porter that flipped a coin that turned up heads?"
T(x) = "does the configuration x include a person named mitchell porter that flipped a coin that turned up tails?"
​

where E, H, and T take values in the boolean algebra {true, false}. And, for example in our hypothetical universe, we have a logical identity E(x) ==> (H(x) or T(x)) expressing the fact that if a configuration includes a coin-flipping experiment, then it either comes out heads or it comes out tails.



Now, construct a new physical theory whose phase space is PxP (i.e. a configuration of the new theory is a pair of configurations of the old theory). The laws of physics operate pointwise (so the time evolution of (y,z) can be computed by evolving y and z individually according to the old theory, then pairing them back together).

(PxP could be replaced with other interesting things, like a product of more copies of P, or by the space of probability distributions on P, or by a space of arbitrary distributions on P -- everything behaves in essentially same way, so I've used PxP for simplicity)

The solution to the measurement problem works the same way -- with E, H, T, or any other proposition that asserts something about what goes on in the universe taking values in the boolean algebra { (true, true), (true, false), (false, true), (false, false) }.

And in this new theory, the logical identity E(x) ==> (H(x) or T(x)) for any configuration x (= (y,z)) in the phase space PxP.


All of the physical laws in the theory I described with phase space PxP have exactly the same form as the physical laws on the original theory with phase space P. All of our propositions about what happens in the universe decompose into a pair describing what happens on each component.


The difference between the original theory and the new theory are completely imperceptible to anything or anyone described by the universe. If it is possible to talk about whether the state of the universe includes an experiment and what outcome occurred, then it is impossible to use this experiment to distinguish between the two theories described above.

And if no experiment can distinguish, then there is no scientific basis for insisting that one does a better job of describing reality than the other.

I don't want to presuppose the actuality of all the worlds appearing in the probability distribution.

Whether you want to presuppose something shouldn't affect your ability to acknowledge it equivalent or otherwise indistinguishable to something else.

 

[kith]

it will only convince the person in front of the gun, not the person behind it.

I haven't read the paper(s), but doesn't the answer to the question 'who will be convinced?' also depend on the branch, in which the surviving experimentalist ends up? There will be a number of branches where he survives, and in some the spectators will be convinced of MWI, and in some not. Depending on the unlikelihood of his survival, on the question whether the spectators are aware of his experiment, etc.

So although it is only the person who commits the suicides who can conduct the experiment, the question 'how likely will others be convinced of MWI?' should be subject to the experimental setup.

 

[Fyzix]

First, it will not "convince" only me, but also all my neighbors living in the branch in which I survive each time.

Second, I doubt that it will really convince anybody; it could also be that I was just lucky to survive each time, or that there is some secret mechanism providing my survival in a deterministic manner (for example, perhaps someone found a way to control the initial conditions of Bohmian hidden variables).

Yeah and not to mention: if MWI had been true it would have to somehow obey Born Rule, so whether it would be many worlds or only one world the probability of surviving would have to be the same, so it wouldn't be a proof of anything to you.

In a single universe the chance of you surviving a shot to the head would be identical to the probability in a functioning model of MWI which could somehow make sense of Born Rule.

QS is just a fallacy anyway, even if MWI miraculously made sense you would not get immortality with it.
Like I started earlier in the thread, it just says that "a" version of you survives, not that "you" survive.
If you are on one of the branches where you die, you die just like a single universe model, what difference does it make that a twin of you survived? Your consciousness doesn't "jump" into this parallel universe like some sort of magical soul.

I'm actually pretty astonished that someone like Tegmark considered this view to ever be legit.

 

[Fredrik]

I haven't read the paper(s),

In case you or someone would want to, this is the one: http://arxiv.org/abs/quant-ph/9709032. It's right at the end.

but doesn't the answer to the question 'who will be convinced?' also depend on the branch, in which the surviving experimentalist ends up? There will be a number of branches where he survives, and in some the spectators will be convinced of MWI, and in some not. Depending on the unlikelihood of his survival, on the question whether the spectators are aware of his experiment, etc.

I don't think we need to complicate things by considering spectators that don't have all the information. The experimenter's assistant knows exactly what the experiment is about, and the question is whether he will be convinced by a low probability result. As I said in my previous post, I don't see why a low probability result would favor a MWI over the CI, and neither does Tegmark.

Tegmark's argument is that the suicide experiment will convince the person who's trying to kill himself, because the MWI says that he/she is certain to survive. (That's his argument, not mine).

 

[Samshorn]

Let P be the phase space of a universe-describing physical theory... Suppose we've solved the measurement problem for P... Now, construct a new physical theory whose phase space is PxP (i.e. a configuration of the new theory is a pair of configurations of the old theory)... The time evolution of (y,z) can be computed by evolving y and z individually according to the old theory, then pairing them back together...

What do you mean when you say "suppose we've solved the measurement problem for P"? The purpose of MWI was to solve the measurement problem. If you concede that it doesn't, I suspect there is no argument here. No one disputes that, given a satisfactory theory P of the universe, we could metaphysically postulate a theory consisting of the Cartesian product PxPx... of multiple independent universes. But it would be pointless. If we can already solve the measurement problem in P, what would be the point of postulating a Cartesian product of independent Ps?

 

[Demystifier]

Tegmark's argument is that the suicide experiment will convince the person who's trying to kill himself, because the MWI says that he/she is certain to survive. (That's his argument, not mine).

I criticize this in another (new) thread:
https://www.physicsforums.com/showthread.php?t=526213

 

[Hurkyl]

No one disputes that, given a satisfactory theory P of the universe, we could metaphysically postulate a theory consisting of the Cartesian product PxPx... of multiple independent universes. But it would be pointless.

The issue at the moment is mitchell porter's assertion that it's obvious from direct observation that reality has definite outcomes, versus my assertion that it is absolutely impossible to empirically distinguish between definite and the type of indefinite outcome I'm describing.

(and also my assertion that being able to understand that indistinguishability is a prerequisite for being able to understand a decoherence-based interpretation of QM)


If you already recognize PxP is empirically equivalent to P, then you have no need to pay attention to that post.

 

[Hurkyl]

On a different note, I worked out a thought experiment I wanted to come up with a couple days ago.

Scenario 1: I start with a qubit in a carefully chosen state and measure its spin along the Z axis. The state chosen so that my measurement will result in Z+ 60% of the time and Z- 40% of the time.

Scenario 2: I start with a qubit in a carefully chosen state and measure its spin along the Z axis. If I get spin down, I then measure its spin along the Y axis. The state was chosen so that my end result will be Z+ 20% of the time, Y+ 40% of the time, and Y- 40% of the time.

In both cases, I hide my results and pass the qubit along to you to experiment with however you want.


Fact: you get identical experimental results in both scenarios.

Or put differently, if we repeat this 1000 times and I use the same scenario each time and you are meant to figure out which one, you can do no better than guessing blindly.


From a collapse interpretation's point of view, I think this result is very surprising. In the first scenario, I'm sometimes giving you a Z- qubit, and never giving you a qubit in a definite Y spin state. In the second scenario, I'm never giving you a Z+ qubit, and sometimes giving you qubits in a definite Y spin state.

 

[mitchell porter]

The difference between the original theory and the new theory are completely imperceptible to anything or anyone described by the universe. If it is possible to talk about whether the state of the universe includes an experiment and what outcome occurred, then it is impossible to use this experiment to distinguish between the two theories described above.

And if no experiment can distinguish, then there is no scientific basis for insisting that one does a better job of describing reality than the other.

Neither theory contains an "indefinite outcome". Theory P has definite outcomes, and the new theory is just PxP - a doubling of definite outcomes.

you do notice that, in classical mechanics "probability distribution over N states" is essentially identical to "N worlds (with weights that sum to 1)" right?

If you mean "possible worlds" then I think I accept the equivalence.

Whether you want to presuppose something shouldn't affect your ability to acknowledge it equivalent or otherwise indistinguishable to something else.

"Actual" is different from "possible", and "essentially identical" is different from "indistinguishable".

 

[Hurkyl]

Neither theory contains an "indefinite outcome". Theory P has definite outcomes, and the new theory is just PxP - a doubling of definite outcomes.

Please clarify. The outcomes seem pretty clearly indefinite to me -- e.g. if H(x) and T(y) hold for theory P, then (x,y) is pretty clearly indefinite in PxP. (despite the fact that E(x,y) holds -- the coin flip definitely occurred)

In case it wasn't clear, remember that I am talking about experiments done internally to the universe.

 

[mitchell porter]

Please clarify. The outcomes seem pretty clearly indefinite to me -- e.g. if H(x) and T(y) hold for theory P, then (x,y) is pretty clearly indefinite in PxP. (despite the fact that E(x,y) holds -- the coin flip definitely occurred)

In case it wasn't clear, remember that I am talking about experiments done internally to the universe.

What do you mean by "(x,y)"?

 

[mitchell porter]

Actually, forget it. This whole discussion is nonsense. There is no such thing as an "indefinite outcome", it's tantamount to speaking of a "non-outcome outcome". Your attempts to provide examples of an indefinite outcome are just inappropriate uses of formalism, inappropriate because they don't have any meaning when you use them that way.

If, against all appearances, by an indefinite outcome you just mean an uncertain outcome, in the sense of experimental uncertainty: experimental uncertainty derives from a number of practical factors like low resolution of the pointer variable and unreliability of the process which correlates the pointer variable with the measured property. It would make no sense to reify uncertainty, and speak e.g. of a particle that is "objectively somewhere between 2 meters and 3 meters away, but also objectively not at any specific location between those two extremes". Concepts come with conditions of use attached, and if we really can't apply them according to those conditions, then we need new concepts. If it's wrong to speak of particles having exact positions and momenta, that means that they have some other properties, not that they have objectively inexact positions and momenta.

So, this is my last post in the thread. I'm going back to thinking about concrete physical questions, which include realistic explanations of QM, but which do not include logic-chopping debates about whether the concept of indefinite outcome makes sense or not. Hurkyl, though I have expressed my disdain for your philosophy of QM in this thread, obviously you can apply the machinery of QM sensibly enough. It's often like this with "philosophical" issues; you find people who, for philosophical/ideological reasons, will say bizarre things, but then it turns out that, yes, on a mundane level the bizarreness doesn't interfere with practical activities in the way you think it might, if it was taken literally. I think it's the same with you, you can use QM, you can get along in the world like any other intelligent practical person, but then you have this ideological superstructure in which you say nonsensical things because you believe them to be logically implied by QM - or for some other reason. Maybe I have you completely wrong, it doesn't matter now because I'm on my way out.

 

[Hurkyl]

If you don't want to think about the appearance of definite outcomes to an observer internal to the universe when the external view doesn't have them, that's fine. But don't tell other people not to think about it just because you don't want to.

If it's wrong to speak of particles having exact positions and momenta, that means that they have some other properties, not that they have objectively inexact positions and momenta.

And one candidate for that "other property" is wavefunction-ness.

But the issue isn't about particles -- the issue is about quantum mechanics applying to systems "large" enough to include things that look like observers and measuring devices. If you are part of a quantum system, whether wave-functions-evolving-via-unitary-evolution can predict that you see a nice happy classical reality.

And on that last question, there were once lots of serious obstacles that have been overcome. I believe the only two remaining obstacles are:

• Develop quantum thermodynamics to a point where we can precisely formulate the question and test it
• Deal with people who refuse to consider the question

(I put "large" in quotes, because IMO all you need is 2 qubits and a CNOT gate)

 

[Hurkyl]

Actually, forget it. This whole discussion is nonsense. There is no such thing as an "indefinite outcome", it's tantamount to speaking of a "non-outcome outcome". Your attempts to provide examples of an indefinite outcome are just inappropriate uses of formalism, inappropriate because they don't have any meaning when you use them that way.

You don't like the phrase "indefinite outcome"? Fine, call them kumquats, I don't care. All I care is that there is no observational difference (for someone in the universe) between a classical universe with kumquats a reality that has a definite state, but we've assigned ignorance probabilities to a distribution of possibilities.

"P or not P" is a tautology. But viewed externally, even in classical logic, it does not follow that there are only two truth values e.g. the four-valued logic consisting of the four truth values

{(true, true), (true, false), (false, true), (false, false)}​

Note that, even in this logic, "P = true or P = false" is a tautology -- e.g. if (externally) we see P = (true, false), then when we compute internally to the logic:

"P = true or P = false" is "(true, false) or (false, true)" is "(true, true)"​

 

[fleem]

If consciousness never ceases, then the person repeatedly attempting and failing suicide could easily inductively prove to himself/herself that the probability of successfully committing suicide is less than what he might have estimated is the probability that there is life after death (in the classic sense of the soul living after the body dies). Yet that person also knows that others had "succeeded" in their suicides as per natural laws of physics. So anyone claiming that consciousness must always survive, is also claiming the probability of life after death is comparable to the probability that the natural laws we've observed are true.

 

[Ken G]

I'm sorry if I missed it in this long thread, but what I never understood about the "quantum suicide" scenario is why we can assume it is our own consciousness that will be the one that survives? It seems to me that in the Many Worlds view, there would need to be many consciousnesses associated with the me-like entity in lots of those other worlds, and I have no perception of them now, so why would I suddenly have a perception of a surviving one? I should no more care about that surviving consciousness than I should care about your consciousness when I attempt suicide. It can be argued that perhaps I shouldn't care any more about my own consciousness, on the grounds that any ownership there is illusory, but if I adopt that philosophy I don't need Many Worlds to tell me I can't succeed at suicide-- I can't as long as anyone else survives.

 

[Fredrik]

The discussion about quantum suicide continued in another thread. Demystifier linked to it in one of his posts above.

 

[Ken G]

OK, thanks. As for the interpretation of Many Worlds, there is never anything wrong with an interpretation, although one can explain one's reasons for the preference. I see "many worlds' as the ultimate tradeoff between sacrificing anything remotely resembling an empirically motivated ontology in favor of a rationalistically motivated ontology. But the empiricists can't really complain-- it gets all the predictions right. No one has suggested any different predictions, have they? (The one just above about failed suicide attempts is invalid, as we wouldn't have anyone in our world with multiple failed attempts, and none of us have done it. Have we?)

 

[Eqblaauw]

I think the poll of this David Raub fellow his highly misleading, the whole poll is unfindable (it's in a french magazine of an issue that's not online anymore), the outcomes are very weird
In no other poll more then half of the physics subscribe to the mwi, when they have the 'I don't know' choise. David Deutsch himself says 20% at best believe this theory. When you search for you David Raub you find out that he has send a letter to Everett, explaining how much he likes is theory. A poll by a fanboy isn't very reliable. David Raub hasn't got anymore hits then involving the many-world on google, that says something.