Problems with Many Worlds Interpretation

[Ken G]

We already have wonderful examples of the benefit of 'observer in the system'. The symmetries of the laws of mechanics put severe constraints on what sorts of quantities are physically meaningful... but only on the hypothesis that observers obey the laws of mechanics as well.

Indeed, symmetries are an excellent access point to both the value, and the drawbacks, of rationalistic thinking. Symmetries are the rationalist dream-- reasoning by what has to be correct, the straight line between two points. This is also why rationalistic thinking is so powerful, and is responsible for every physical theory we have ever had. But it's also the "poison pill" of rationalism-- because symmetries were made to be broken. Consider the history of symmetries-- symmetry with time was once seen as fundamental, but then came the Big Bang model and symmetry with time was seen as only an effective symmetry. Time reversal symmetry seems quite fundamental in physics, but not in thermodynamics. Parity reversal was promising, then combinations like CT. You know the story, the history of science is a tale of one sliding rationalistic anchor right after another. This doesn't mean the rationalistic anchors are bad-- they are the very stuff of physics theories, one just needn't take them too seriously.

As for the observers obeying the laws, relativity has some nice comments on this. Relativity is a different kind of physical law, it is a meta-law-- a law that constrains laws. That right there tells you that there is something different about the role of observers in physical theory. It's not that the observer has to obey the laws, it is that truths about observers constrain the laws those observers could ever find useful. I see relativity as a kind of flashlight for looking for your keys-- it is what saves you from having to look under the lamppost, because relativity insures that the "lamp" comes with you. We look for laws like that, we need laws like that, but their purpose is to explain observations-- it is the observations that are the rock of physics, and we should expect the laws we use to understand them to constantly change as the observations broaden (but never change).

But if we try to say that the observers aren't part of the system and aren't described by the laws of SR, we no longer have a compelling argument -- there's no longer any reason why simultaneity must be relative for observers.

We cannot have that observers are part of the system in SR. Just look at the postulate: "the laws of physics shall be the same for all (inertial-- fixed in GR) observers." The postulate requires a definition of an observer be already in place, and then the meta-law constrains all the laws that could possibly be useful to an empiricist. The idea of relativity is just saying "let empiricism be an objective approach, and seek laws that can be useful within that approach", and the rest is about which brand of relativity accomplishes that (the constant c brand, for now). The observer is not "part of the system" there, or we wouldn't need relativity in the first place, we'd just have the laws of systems, and observers would automatically be covered by them, rather than appearing directly in the meta-law.

 

[Dmitry67]

But collapse is what we perceive-- it does not require a cause.

No. We don't perceive a collapse. To perceive a collapse one must at first observe a fuzzy combination of dead and alive cats, and then how it switches to a definite outcome. Nobody had ever observed it, so your claim is wrong.

For the same reason nobody had ever perceived a neutrino, instead, scientists perceive a list of numbers and retrospectively decode the behavior of the neutrino. In both cases it is the same, we perceive a shadow and we try to decode how an object look like.

 

[Hurkyl]

I think I see your point and I think it's a great example indeed. This is the connection between observer invariance or covariance and symmetries in nature that effectively encode the laws of nature.

But I think I've expressed my view on this before (here I think I am far more radical then Ken though).

There are two ways to understand this, and my hunch is that you adopt the first, I adopt the second. Both solve the problem you address but in different way. Way(1) is the realist way, way(2) is the solipsist way.

It's only a problem if you postulate that observers are external, which I don't. It's an empirical fact that I and all other observers obey Newton's laws^* (in its domain of validity^**). If there was any evidence that observers did not obey Newton's laws, scientists would be studying said observers to look for new physics.

*: Or, at least, something indistinguishable from them
**: I mean if we avoid questions like the topic of the energy levels of electrons in atoms

It's not that an observer must obey Newton's laws -- it's that we have mountains of empirical evidence for Newton's laws. The claim that observers cannot observe an absolute notion of 'at rest' is relative is merely a prediction of Newton's laws. That it is borne out in reality reflects the high level of confidence we have in Newton's laws, and that it would be violated is both very unlikely and very interesting.


Aside: symmetry is not a restriction on observers. The argument that there is no absolute notion of 'at rest' is as follows:

Consider the claim that there is some proposition P(v) of an observers velocity (v) with the property that

• P(0) = true
• P(v) = false if v is not zero

P can be thought of as the proposition that the observer is truly at rest.

We could apply a Galilean boost that adds v₀ to velocity, and we'd have a new proposition P' with the property

• P'(v₀) = true
• P'(v) = false if v is not v₀

(I've used the fact 'true', 'false', '=', 'not', and 'if ... then ...' are all invariants)

This is all still fine. We haven't contradicted anything -- P' is a different proposition than P. Nor have we refuted the claim -- this just means "at rest" corresponds to a velocity of v₀ in this representation.

But if the proposition P was actually constructed entirely out of physical laws (or any other invariants under symmetry), we must have P' = P, and so we derive a contradiction:

false = P(v₀) = P'(v₀) = P(0) = true​

 

[Hurkyl]

I think Bohr's points on this matter are clear-- there is no quantum realm, and physics is what we can say about nature.

The "quantum realm" is the part of nature that we talk about with wave-functions and unitary evolution.

Whatever ontological beliefs you may have about physics in general, you still have to recognize that collapse-based quantum physics posits that there are some things you can talk about using wave-functions and unitary evolution, some things that you can't talk about that way, and there is an interface between the two that operates via collapses and the Born rule.

The "quantum realm" is, by definition, the aspect you can talk about using wave-functions and unitary evolution.

it is the subject/object dichotomy that is central to all scientific models because it is the beating heart of the scientific method.

Your version anyways. (the phrase 'subject/object' still doesn't mean very much to me, so I'm guessing at the meaning)

I, personally, have never seen "now, make sure you have a subject/object dichotomy in your philosophical beliefs" listed in the steps of the scientific method.

Nor have I ever recall seeing anyone justify making such a dichotomy -- they just assert "do it that way".

But if you're right -- if subject/object dichotomy really is part of what "empiricism" means -- then tell me this: why should I care? Why am I not better off adopting a new philosophical position that doesn't insist on the dichotomy but is otherwise the same as empiricism?

CI is nothing but the interpretation that is required by empiricism, with no added bells or whistles to make it seem like a more rationalistic world.

The cut is a bell and/or whistle.

Indeed, they were both throwbacks to old ways--

CI is a throwback to definiteness -- to non-parmness. Well, it's not really a throwback, it's more of a carry-over from classical mechanics.

Everett is a throwback to ancient Greek rationalism, in which the truth is the mathematics.

This doesn't make sense. You keep saying it, but it doesn't make sense. The idea is that

[physics] is the general analysis of nature, conducted in order to understand how the universe behaves.

Everett wasn't trying to operate in some strange Platonic world of absolute truth. Everett was taking our general analysis of nature, and finding a way applying it to understand how the universe behaves.

The difference is between "I observed X" and "X was observed." Science always treats those sentences as identical, hence the problem.

By consistency of observation, it's not a problem -- if "O observed X" is a proposition independent of O (i.e. that "O observed X"="O' observed X" whenever both sides are well-defined), then there is no need to include O in the proposition.

 

[Hurkyl]

Consider the history of symmetries-- symmetry with time was once seen as fundamental, but then came the Big Bang model and symmetry with time was seen as only an effective symmetry.

I'm not aware of anything in BBT that suggests you can't reverse the arrow of time and get a big crunch.

Time reversal symmetry seems quite fundamental in physics, but not in thermodynamics.

Eh? It's there in thermodynamics too. It's just as ridiculously unlikely for entropy to decrease with time running forwards as it is to decrease with time running backwards.

Symmetry of the laws and symmetry of the physical configuration are very different things.

You know the story, the history of science is a tale of one sliding rationalistic anchor right after another.

... and this suddenly comes out of nowhere.

You do get the idea that our confidence in pieces of knowledge depends on empirical verification? And when new experiments are done, our confidence in some piece of knowledge can diminish? Or that our confidence in it may be surpassed by our confidence in some new piece of knowledge? And even when surpassed, our old piece of knowledge is still useful? That working with reason doesn't entail having some strange idea of absolute platonic truth and being trapped in a cycle of building a house of cards and having to completely scrap it for a new house of cards?

 

[Fra]

It's only a problem if you postulate that observers are external, which I don't.

a) I think we use "external" in different ways. I try to take very seriously that observers are real physical systems, an not just sitting in some mathematical realm. So the observers are indeed in this sense "inside observers".

Note that this is not taken seriously in mainstream physics. But then it's no wonder that pretty much the only observables anyone has defined in quantum gravity theories are scattering matrices where the observer is dismissed to the boundary of the system at infinity. I see this as a symptom that is greatly amplifier in QG, but that is rooted in the foundations of QM. The very point is that QM (not just interpretations) are not respecting thte fact that the observer only encodes finite amount of information. This IMO forces the conclusions that any inferrable laws can't possibly be timeless platonic. They are more like interaction tools.

b) However, the distinction you seem to have problems with (which I think is unavoidable) is between the observed and and observer. Actually this distinction is in my view simply follows from the decision perspective that what the observer KNOWs is distinct from what the observers is trying to guess. The whole quest is; to predict the future given a memory of the past. Here I like to cite Zurek which said "what the observer IS; is indistinguishable from what the observer KNOWS".

This means that the distinction observer vs observed; is simply the distinction between what you know and what you don't know.

I always see "the system" as a black box, from the point of view of the observer. Actually, the complete picture means that the observers entire environment is a black box. The decomposition of the environment into a small subsystem and the remainder is artificial, except in obvious cases where the observer effectivel controls and monitors the entire boundary of the system - like in a particle collider.

(in its domain of validity^**)

Yes. This was part of my point as well, but I think this requires more attention than a footnote ;)

You can put the same thing is two equivalent ways, either you say that the theory constraints the observers for which it applies, or holds for all observers "in it'd domain of validity", which is just two sides of the same coin).

/Fredik

 

[Ken G]

No. We don't perceive a collapse.

Actually, we do. "Collapse" doesn't mean a single thing other than "definite state of a subsystem."

To perceive a collapse one must at first observe a fuzzy combination of dead and alive cats, and then how it switches to a definite outcome.

If that were true, neither Bohr nor Heisenberg would have held to their own ideas.

In both cases it is the same, we perceive a shadow and we try to decode how an object look like.

Certainly, the "neutrino" concept is a rationalist notion, no question there.

 

[Ken G]

Your version anyways. (the phrase 'subject/object' still doesn't mean very much to me, so I'm guessing at the meaning)

No, it is not "my version." Go look up "scientific method", provide any text you find, and I will gladly point out where the subject/object dichotomy appears in that text. I'm sure you could too. The "subject" is the scientist, the "object" is the experiment. Now find me some science that doesn't make that dichotomy, unless you count psychology as science.

I, personally, have never seen "now, make sure you have a subject/object dichotomy in your philosophical beliefs" listed in the steps of the scientific method.

Then you have chosen to overlook it, and need to try a little harder. Again, find any text you like on the subject of the scientific method.

Nor have I ever recall seeing anyone justify making such a dichotomy -- they just assert "do it that way".

Yes, it is quite easy to overlook what we are doing. So what? So we shouldn't notice what we are doing?

But if you're right -- if subject/object dichotomy really is part of what "empiricism" means -- then tell me this: why should I care? Why am I not better off adopting a new philosophical position that doesn't insist on the dichotomy but is otherwise the same as empiricism?

Because you risk failure.

The cut is a bell and/or whistle.

To a rationalist, yes.

CI is a throwback to definiteness -- to non-parmness. Well, it's not really a throwback, it's more of a carry-over from classical mechanics.

No, definiteness is what we perceive. You are trying very hard to ignore that fact, but it is the crux of empiricism. Science is built on the back of definite outcomes to experiments, and only the extreme rationalist modifies that to "science is built on the back of the appearance of definite outcomes." The word "appearance" in that sentence is not falsifiable, but whenever an appearance is absolutely consistent, there is no difference between appearance and fact.

Everett wasn't trying to operate in some strange Platonic world of absolute truth. Everett was taking our general analysis of nature, and finding a way applying it to understand how the universe behaves.

That sounds like someone describing a theory, not an interpretation of a theory.

By consistency of observation, it's not a problem -- if "O observed X" is a proposition independent of O (i.e. that "O observed X"="O' observed X" whenever both sides are well-defined), then there is no need to include O in the proposition.

It is necessary when X has no meaning without O, as in science.

 

[Ken G]

I'm not aware of anything in BBT that suggests you can't reverse the arrow of time and get a big crunch.

I wasn't referring to the arrow of time, but to the symmetry of time translation. I mentioned the arrow in a different context-- thermodynamics.

Eh? It's there in thermodynamics too. It's just as ridiculously unlikely for entropy to decrease with time running forwards as it is to decrease with time running backwards.

Yet the second law makes a very specific claim about the arrow of time. Yes, we have a law of physics, used constantly (indeed the patent office uses it daily), that breaks time reversal symmetry.

Symmetry of the laws and symmetry of the physical configuration are very different things.

Obviously. The second law is a law, it doesn't require specification of anything about the system.

... and this suddenly comes out of nowhere.

You can't see the slide of rationalistic anchors in the history of science? I think you must be trying pretty hard not to.

You do get the idea that our confidence in pieces of knowledge depends on empirical verification? And when new experiments are done, our confidence in some piece of knowledge can diminish? Or that our confidence in it may be surpassed by our confidence in some new piece of knowledge? And even when surpassed, our old piece of knowledge is still useful?

Certainly. And none of that has anything to do with those "rationalistic anchors" that litter the history of science. There's nothing wrong with them, indeed they are essential-- until you take them literally, like the postulates of quantum mechanics.

That working with reason doesn't entail having some strange idea of absolute platonic truth and being trapped in a cycle of building a house of cards and having to completely scrap it for a new house of cards?

No, "working with reason" certainly doesn't require that, which is how Bohr and Heisenberg were able to do it too.

 

[Hurkyl]

Yet the second law makes a very specific claim about the arrow of time. Yes, we have a law of physics, used constantly (indeed the patent office uses it daily), that breaks time reversal symmetry.

I realized this part needed clarification after I had already left.

One thing we have that breaks the symmetry is strong prior odds on the past being in a low entropy state. Given knowledge that the entropy an hour ago is less than it is now, it's considerably more likely that the entropy of five minutes ago is somewhere in-between than any of the other alternatives.

(conversely, if we somehow had knowledge that some specific point in the future was in a low entropy state, the second law would have to reverse somewhere along the way -- it would just be too unlikely to predict otherwise)But, of course, I don't know how we got knowledge of the past being in a lower entropy state to begin with. (Nor, of course, have I argued that searching for violations of time reversal is a worthwhile pursuit in physics)

 

[Ken G]

One thing we have that breaks the symmetry is strong prior odds on the past being in a low entropy state. Given knowledge that the entropy an hour ago is less than it is now, it's considerably more likely that the entropy of five minutes ago is somewhere in-between than any of the other alternatives.

The second law of thermodynamics is a classic example of how human intelligence interacts with our own laws. There is really no such thing as entropy, in a physical sense, the system just goes from one state to another. All the same, we can tell which way the movie is running quite easily, because the concept of time comes from us in the first place-- from what we choose to care about. We choose to care about things that change with time in a recognizable way, we care about various types of information. We "prepare the system", and watch what happens next. The way we think about reality has an arrow to it, but a universe without intelligence cannot establish any arrow to time. This is what I mean about the duality of the observer and the observed-- neither is a meaningful concept without the other. Bohr said it: physics is what we can say (using our intelligence and abilities to perceive) about nature. It's clear enough where physics comes from, just watch the process. But is it true outside of us, even to worlds in which we are neither present nor even possible? I don't think one can say that it is, without entering into a kind of pretense-- unfalsifiable, but pretense all the same.

 

[Ken G]

b) However, the distinction you seem to have problems with (which I think is unavoidable) is between the observed and and observer. Actually this distinction is in my view simply follows from the decision perspective that what the observer KNOWs is distinct from what the observers is trying to guess. The whole quest is; to predict the future given a memory of the past. Here I like to cite Zurek which said "what the observer IS; is indistinguishable from what the observer KNOWS".

That's an interesting way to frame the subject/object dichotomy-- the subject is what is taken as known, the object is the foray into the unknown. It resonates with your earlier points about the importance of "surprise" in anything we could call an actual interaction. I believe that is what both Parmenides, and the postulates of QM, missed-- that there is something very fundamental about doing science that involves not knowing the outcome of the experiment in advance, which we tend to overlook when we put all the stress on science as a predictive tool. Yes the goal of science is to be able to predict, but the act of doing science requires something that is not known, something that is the object, something that is "outside" the observer, or outside the axioms being used by the observer. This is a kind of fundamental paradox of science, related to the fact that every discovery opens a new frontier of questions, and the only mistake is thinking that the next answer is going to be an endpoint, a "TOE."

 

[Dmitry67]

Actually, we do. "Collapse" doesn't mean a single thing other than "definite state of a subsystem."

Sorry, but no.

Do you agree that observing (perceiving) car wreckage is not the same as observing a car crash?

For the very same reason observing the definite outcome is not the same as observing the collapse.

For example, in the 'mascoscopic realism' interpretation there are always definite outcomes. However, collapse in such theories is just a purely mathematical concept (without physical existence) which helps us to calculate probabilities of the macroscopic events. In other words, if you observe car wreckage, you can't be sure that there was a time when car was intact - it could be produced that way.

 

[Ken G]

Do you agree that observing (perceiving) car wreckage is not the same as observing a car crash?

Yes. But collapse is the wreckage. The crash is not observed in quantum mechanics.

For the very same reason observing the definite outcome is not the same as observing the collapse.

Yes, I know, that's why your concept of collapse is incorrect. Collapse is an outcome of a measurement. CI is empirical. If you see a red Porsche enter a tunnel, and enter the tunnel moments later to find a smashed red Porsche, you say the car crashed. That's quantum mechanics, that's how the term "collapse" is used in CI.

 

[Dmitry67]

No, based on a public definition:

http://en.wikipedia.org/wiki/Wave_function_collapse

In quantum mechanics, wave function collapse (also called collapse of the state vector or reduction of the wave packet) is the phenomenon in which a wave function—initially in a superposition of several different possible eigenstates—appears to reduce to a single one of those states after interaction with an observer.

So it is process of transforming (A) initially in a superposition of several different possible eigenstates into (B) a single one

You perceive only (B), how can you claim (A)?
Contre-example: in classical mechanics, we observe only definite outcomes. Does it mean that collapse is a part of classical mechanics?

 

[Fra]

So it is process of transforming (A) initially in a superposition of several different possible eigenstates into (B) a single one

You perceive only (B), how can you claim (A)?

I see no problem in claiming (A). Not though that (A) is not an "event", (A) is simply a state. The information you have may simply give you a system "in superpositon". If this is the information the observer has then that's what he will "perceive".

But if I may rephrase the question. I sense that that bugs you is rather this: How can we by observing say the cat understand the process whereby you ARRIVE at at superposition in the first place? Since its seems that you never at one single "observation event" get one datapoint corresponding the superposition?

This is a more sensible question, and in MY view the answer is that a general STATE, is NOT "prepared" simply by a singular observation, it rather incorporates the history of the observer. When one "combines" rationally information that are not commuting, in a way where you weight the corresponding evidences properly you end up with a kind of "information state" that simply can't be described in terms of a classical probability distribution where the set of future events belong to a simple event space. You are forced into a new framework where there is instead different event spaces (corresponding to different operators) that generally aren't commuting.

I can give you one credit for critique and it's the fact that even though I personally have a picture of this in the context of my own work, there is to my knowledge no published papers which elaborates this in the way I think is needed to understand it. This is something that I'm working on and wish that one day I'll be able to publish this in the bigger context.

So the missing part is, how can we understand the construction of a superposition not just mathematically since htat's trivial, but in terms of an actual interaction history encoded and processed by a real observer.

This seems like a simple thing, but I've found in my own thinking that although this is a tiny brick in the big picture of trying to reconstruct measurement theory and understand interactions, it's very complicated since several deep issues meet here. For example, my understanding is that in order to properly understand how to defined the new logical AND and OR operators for these more general information structure one needs to understnad the WEIGHTS of the non-commuting informations and in the inference picture this is naturally nothing but the inertia of opinon. This deeply seems to be related to mass in physical interactions, and this is also observer dependent which means this simple thing necessarily couples with the origin of mass as well as a bigger picture of scale depdenent theories. This is why I haven't complete this yet. I've found it to be a somewhat tough problem for the main reason that it's impossible to isolate from other hard problems. So all problems needs to be solve in parallell.

/Fredrik

 

[Ken G]

No, based on a public definition:

That definition is just precisely what I said-- collapse is simply the detection of a definite state (when quantum formalism indicates a mixed state). That's it, that's a collapse. Nothing about "watching the transition from a superposition to a definite state", there is zero connotation in that definition that we have to "watch" the collapse, the collapse is an endpoint and the "transition" is simply an inference on our part.

Contre-example: in classical mechanics, we observe only definite outcomes. Does it mean that collapse is a part of classical mechanics?

Yes, collapse is most definitely part of classical mechanics, there was just no need for the term there because it seemed obvious. But a collapse in classical mechanics is what we've been talking about quite a lot in this thread-- you flip a coin, and don't look. The state of the coin is now mixed, entirely classically. Now look at the coin. The state of the coin is now "collapsed." There is no difference between CI and MWI that is not accessed in flipping a coin.

 

[Hurkyl]

But a collapse in classical mechanics is what we've been talking about quite a lot in this thread-- you flip a coin, and don't look. The state of the coin is now mixed, entirely classically. Now look at the coin. The state of the coin is now "collapsed." There is no difference between CI and MWI that is not accessed in flipping a coin.

In the example I've been trying to consider all thread, the state started mixed, the mixture persisted through the coin flip, and remained after the flip was observed. That's the only behavior of mixtures that classical mechanics supports.

But in the usual way to interpret classical mechanics the state is never mixed. Probabilities that enter into the analysis are measures of ignorance, not of "reality". Classical mechanics is traditionally interpreted as having counter-factual definiteness: it's a hidden variable theory.

I now feel like you never understood the distinction, and you're interpreting all probabilities as ignorance probabilities. (I'd already started getting the impression you were treating parmness as just another word for expressing ignorance )

 

[Hurkyl]

No, it is not "my version." Go look up "scientific method", provide any text you find, and I will gladly point out where the subject/object dichotomy appears in that text. I'm sure you could too. The "subject" is the scientist, the "object" is the experiment.

I know it will talk about scientists and experiments. The controversy is your claim it will not only talk about dichotomy, but it will be built into the very formulation of the method. (and not in a "oh, I can see how one could interpret things that way" fashion but in a "nobody could possibly interpret it any other way")

Now find me some science that doesn't make that dichotomy,

Pick any science you want. I don't know of any field of science asserting the dichotomy, despite your claims of its prevalence.

Because you risk failure.

Er... so what?

No, definiteness is what we perceive.

AFAIK, you are the only person who uses the term that way.

It's fine if you want to use the word that way, but it's not fine when you don't want to listen to how I'm using the word and instead pretend I'm using the word your way.

It is necessary when X has no meaning without O, as in science.

On the hypothesis of consistency of observation, O doesn't need to be stated. If, for some reason, you can't conceive of X without also picturing an O, you can pick any relevant O you want and use that; it doesn't matter which O you pick because all choices work out the same.

 

[Ken G]

In the example I've been trying to consider all thread, the state started mixed, the mixture persisted through the coin flip, and remained after the flip was observed. That's the only behavior of mixtures that classical mechanics supports.

Yes, but this is also a situation that classical mechanics addresses all the time-- mixed states evolving into mixed states. That's "statistical mechanics." It doesn't make any difference to statistical mechanics if the information "actually exists" somewhere, what matters to the physicist is what information they actually have, and how to treat the information they do not have. That's the origin of laws like the second law of thermodynamics, it's perfectly classical.

The classical description of a coin is this. You flip a coin. The information you would need to predict the outcome is not available, so you treat the initial state as mixed. You also treat the final state as mixed. You then look at the coin, and replace the mixed state with a definite state. That's a "collapse", there just wasn't a need to use the term in classical physics, because it was so close to our daily intuition.

Remember, the term "collapse" is used in CI, and CI is an empiricist interpretation, which means that the whole concept of a "state" in the first place is a way the physicist has chosen to address the situation, based on information. (Observations are the reality, not "states.") In CI, there is no difference between reality and information about reality, there is only the requirement that all observers must have consistent perceptions, so they must be manipulating consistent information. That's what "real" is to an empiricist-- consistency of perception. Any "appearance" that is perfectly consistent is not only indistinguishable from what is real, it is what is real. Different people might be looking at different parts of the elephant though, that is no challenge to realism.

Now, the interpretation of that mixed state is something else again. It has no place in the actual calculation, it's just the story we tell ourselves to achieve cognitive resonance. And that story should not invent elaborate world views just to make the mathematics seem cleaner (unless we are rationalists). That is the sole reason why the mixed state outcome used in the actual calculation is interpreted as the electron going into just a single outbound state. The notion of "collapse" of the mixed state was never needed in that interpretation, but it was always there. The deBroglie-Bohm interpretation of QM is the one that preserves that classical interpretation most closely, but CI rejects that because of the requirement that the details of the initial state not only be unknown, but unknowable. In empiricism, that which is empirically unknowable simply doesn't exist, an issue that classical physics did not have to face.

To see all this, just imagine that classical physics had some theorem about the limits of observational precision-- you'd have all the same issues right away, you'd have CI, and deBB-B, and MWI, all purely classical.

I now feel like you never understood the distinction, and you're interpreting all probabilities as ignorance probabilities. (I'd already started getting the impression you were treating parmness as just another word for expressing ignorance )

No, I understand just fine that you interpret mixed states that way, as would any user of MWI. I'm saying why your criticisms of CI are largely around not understanding it-- it's the basic empiricism you don't get. In CI, anything we can say about reality is an expression of information and ignorance. Again: "there is no quantum world."

 

[Dmitry67]

That definition is just precisely what I said-- collapse is simply the detection of a definite state (when quantum formalism indicates a mixed state). That's it, that's a collapse. Nothing about "watching the transition from a superposition to a definite state", there is zero connotation in that definition that we have to "watch" the collapse, the collapse is an endpoint and the "transition" is simply an inference on our part.
Yes, collapse is most definitely part of classical mechanics, there was just no need for the term there because it seemed obvious. But a collapse in classical mechanics is what we've been talking about quite a lot in this thread-- you flip a coin, and don't look. The state of the coin is now mixed, entirely classically. Now look at the coin. The state of the coin is now "collapsed." There is no difference between CI and MWI that is not accessed in flipping a coin.

1 You are taking the second part of the definition, just because you don't like the first part.
2 If wavefunction collapse is a part of classical mechanics, then you are using some kind of alien logic.

 

[Ken G]

In fact, this is a very useful device to contrast CI, MWI, and deBB-B. Just imagine someone adds to the postulates of classical physics another postulate-- that generalized coordinates come with a precision limit that looks just like the HUP. No axiomatic back story to it, no unitarity or Hilbert spaces, just the empirical discovery that measurements have a fundamental inability to create simultaneous precision better than what is specified in the HUP. This would still be what we might call "classical physics", because it has all those postulates, but it just adds this one more.

The reason I say this would still be classical physics is that all the calculations would now be exactly the same. We would predict positions and momenta, subject to the uncertainties in the initial conditions, exactly as is now done. In fact, most of classical physics would be absolutely identical, because HUP-type precisions are generally never obtained in classical physics in the first place, it would be quite an irrelevant added postulate. Until you get to the interpretation, that is-- now we will have exact analogs to CI, MWI, and deBB-B cropping up. So let's look at how each would handle this situation, it will help demystify these interpretations when they are applied to QM.

deBB-B: Let's start here, because this is the closest to the common classical interpretation. The theorem about precision of measurement would be attributed to some limitation we have in probing reality, not to reality itself. So there would "really be" a true initial momentum and position, but we wouldn't have access to it, so we would forced to distribute over our own ignorance, and map that into ignorance of the final state, which we would also not regard as real but only a reflection of our ignorance. The fundamental truth would still be entirely deterministic, but our ability to predict would not. For most applications, this would make no difference, but in chaotic applications like statistical mechanics, it would lead to things like the second law of thermodynamics.

CI: In the strictly empiricist view, there is no difference between a fundamental limit on observational precision, and a fundamental truth about reality. So if we cannot know the initial state to some precision, then there is no initial state to better than that precision. Normally this wouldn't make any difference, for the precision would be limited in the outcome in an unimportant way. But in chaotic applications like statistical mechanics, it would make a huge difference, and force upon us the introduction of the concept of collapse. When the initial uncertainty is not resolvable by our instruments, but it maps into a final uncertainty that is resolvable, then our instruments will detect a final outcome that is not "in" the initial state. This would be purely classical. All the same, we would have the CI interpretation-- a "collapse" must have occurred that we can only track statistically.

MWI: The initial state is already a mixed state, and so are the final states, and they remain so. Since there is no postulate for "collapse", there must not be collapse, and so the mixed states persist and we perceive only a "single world" of the many.

So it's exactly the same choices, and for much the same reasons, just purely classical.

 

[Hurkyl]

I don't think this is analogous at all. All your analogy has is the discovery of a problem with classical mechanics and nothing else. It's missing the essential feature of "Hey, we've figured out new physics, and it tested very well. But upon initial analysis, it doesn't mesh well with anything else we know about physics."

The choices, and especially the motivations in your scenario can't be analogous to the real things.

 

[Ken G]

I don't think this is analogous at all. All your analogy has is the discovery of a problem with classical mechanics and nothing else. It's missing the essential feature of "Hey, we've figured out new physics, and it tested very well. But upon initial analysis, it doesn't mesh well with anything else we know about physics."

On the contrary, the hypothetical theorem I refer to would certainly be viewed as new physics! I'm saying we have all the postulates of classical physics, but it is discovered, to our amazement, that no matter how hard we try, we can never measure x and p to better simultaneous precision than delta x * delta p = h, where the value of h comes as a complete surprise to us. Indeed, it is perfectly possible that it might have gone just like that, well before quantum mechanics, had h been a heck of a lot larger (or our instruments a heck of a lot more precise). That would very definitely be new physics, even without anything about quantized action being found yet, just this constraint on our knowledge of classical systems. I'm saying had that been the case, exactly the same 3 ways of interpreting it would have appeared, and for all the same reasons (empiricists would say there is a collapse, determinists would say there were hidden variables, rationalists would say the ergodic outcomes predicted by the postulates of classical mechanics must actually occur but we don't perceive anything beyond our little h-bin). It's pure historical accident (and the smallness of h) that quantization and wave-particle duality were discovered at the same time as the HUP, so we had QM instead of just a need to re-interpret classical physics.

The point is, this device makes it quite clear what CI, MWI, and deBB-B are saying, independent from the details of QM. Those details have no direct bearing on any of those interpretations, they are all about how empiricist, how deterministic, and how rationalistic is the taste of the interpreter when faced with the impossibility of wrapping up empiricist and deterministic postulates into a nice concise rationalistic structure.

 

[Ken G]

Let me summarize the main things I have argued:

1) there is no scientific difference between CI, MWI, and deBB-B, they all function within the scientific method in exactly the same ways. The differences are purely philosophical, but philosophical differences are nevertheless important to recognize.

2) philosophical insights can be helpful in guiding new theories that differ from QM, such that one of the interpretations might carry over successfully into the new theory.

3) the main philosophical differences center on empiricism (CI) versus rationalism (MWI), since empiricism asserts that we know by perceiving, so we know there is collapse because we perceive it, and rationalism asserts we know by modeling mathematically, so we know there is unitarity (or we think we know, the bugbear of rationalism) by the structure of the most concise and unified version of the postulates. For good measure, we can throw in deBB-B, with its willingness to tolerate some anti-empiricist elements (unperceivable hidden variables) and some anti-rationalist elements (clunky extra postulates) in favor of maintaining classical determinism.

4) If the title of the thread is the "problem" with MWI, then this problem is not that it is internally inconsistent, nor disagrees with observations, because it isn't and it doesn't, it is that it falls through if the postulates of QM are not precisely upheld. Postulates change, but observations don't.

 

[Samshorn]

... the "problem" with MWI... is that it falls through if the postulates of QM are not precisely upheld...

Doesn't this conflict with your comment in the previous post that a precise analog of MWI could be conceived even in the context of classical mechanics? Surely one can always hypothesize a "many worlds" interpretation of ANY theory. In other words, we can always imagine that the world of our experience is embedded in a larger structure consisting of infinitely many such worlds, which do not interact with each other (at least not into the future). In fact, we can imagine this in infinitely many ways. One can argue that QM is more suggestive of such an embedding than some other - past, and possibly future - theories, but it's always possible to construct an MWI of any theory. Obviously the details of any specific MWI would change if the details of the underlying experiential theory changed, but it would still be a MWI. I don't think MWI is falsifiable as an abstract concept.

Your statement above implies that you think the problem with MWI is that, although not presently falsified, it is overly exposed to falsification. That's exactly backwards, isn't it? The optimum theory is the one most exposed to falsification that has not yet been falsified. For example, special relativity is more exposed to falsification than Lorentz's ether interpretation, as a conceptual framework for physics, which is why special relativity is regarded as a stronger (better) framework. The spacetime interpretation has less flexibility than an ether interpretation, and yet is has never (yet) failed.

I would say the problem with MWI (well, one of the problems) is not that it's overly falsifiable, but that it's NOT falsifiable at all. (This isn't the most serious problem with MWI, but it's still serious.)

 

[Ken G]

Doesn't this conflict with your comment in the previous post that a precise analog of MWI could be conceived even in the context of classical mechanics?

Not really, because the same would be said of classical-MWI: it falls through if the postulates of classical mechanics are not also precisely upheld. In other words, I'm not saying that there is something special about quantum postulates, I'm saying there is something special about postulates in the first place. The empiricist view is that all theories are approximate, so none of their postulates should be assumed to hold precisely. One simply should never confuse the map for the territory. Since MWI is never falsifiable, it is only the motivation for it that is falsifiable (the aesthetic structure of the postulates).

Surely one can always hypothesize a "many worlds" interpretation of ANY theory.

Yes, that's completely true, although a "many worlds" interpretation would be sterile unless there is some kind of limit on perception that does not appear in the postulates. The motivation for it is always extreme faith in the postulates, whatever postulates are used, whenever one encounters a limitation in perception that does not appear in them. Perhaps the simplest way to sum up MWI-esque interpretations is the rule "add no postulates simply to acquiesce to what is perceived when a simpler set of postulates is not falsified." That's also the spirit of Tegmark's MUH.

One can argue that QM is more suggestive of such an embedding than some other - past, and possibly future - theories, but it's always possible to construct an MWI of any theory. Obviously the details of any specific MWI would change if the details of the underlying experiential theory changed, but it would still be a MWI. I don't think MWI is falsifiable as an abstract concept.

I agree. It's very hard to say just what it is exactly about QM, rather than any other theory based on postulates, that supports an MWI approach, when no other theory ever did (though it could be argued that Parmenides, 2500 years ago, did in a sense suggest a similar theory, but it was pre-scientific). We certainly encountered lack of complete information in theories before, like in thermodynamics. I think it was the discovery of a fundamental limit on information, the HUP, that is the real source of MWI, hence my simplified version. Rationalist theories won't build in lack of information, they'll just see it as the fault of the perceiver.

Your statement above implies that you think the problem with MWI is that, although not presently falsified, it is overly exposed to falsification. That's exactly backwards, isn't it?

I don't think MWI is falsifiable, I think the postulates that inspire it are. I agree with you that one can always make an MWI interpretation, but will one? If the motivation of MWI is some simple set of postulates, then any compromising of those postulates by observations will defeat the purpose of the MWI interpretation (which was to preserve the postulates in as pristine a form as possible). Now, it should be mentioned that even if the postulates are not exactly true, one might still wish to preserve them in a pristine form as a kind of approximate theory, but rationalism isn't too comfortable with approximate theories either! The issue is not if MWI is "one way to think about QM", any more than it is one way to think about thermodynamics. The issue is whether or not it is true.

The optimum theory is the one most exposed to falsification that has not yet been falsified. For example, special relativity is more exposed to falsification than Lorentz's ether interpretation, as a conceptual framework for physics, which is why special relativity is regarded as a stronger (better) framework. The spacetime interpretation has less flexibility than an ether interpretation, and yet is has never (yet) failed.

I agree that exposure to falsification is indeed a high goal of a theory. I wasn't aware that SR was more easily falsified than LET though, because they both make all the same predictions. I felt SR was favored more on Occam's razor-- if you don't need an aether, why have one? But if SR was proved wrong, and there is an "aether frame", then Lorentz' version would be wrong too. Though you might say he was "closer" in that case-- which is what I mean by different interpretations only really being different in regard to new theories they inspire.

I would say the problem with MWI (well, one of the problems) is not that it's overly falsifiable, but that it's NOT falsifiable at all. (This isn't the most serious problem with MWI, but it's still serious.)

I think we largely agree, but I would point out that neither CI nor deBB-B are falsifiable either, as they are all just inspired by their own philosophical angles (or angels?). Any valid interpretation is no more falsifiable than the theory it interprets. Even though I've been hard on the "world view" inspired by MWI, which frankly seems naive to me despite its sophistication relative to other even more naive world views, I do think it is good to understand all the interpretations, for one never knows where a useful insight might come from.

 

[Fra]

I think this discussion has several facets and issues, and it's hard to discuss them all at once. But somehow the core of the last few posts regards the understanding of the collapse or information update.

I personally think it's slightly confusing to talk about the information updates in "classical mechanics". Because with the exception of thermodynamics classical mechanics really is not an information theory at all.

Maybe it's less confusing to say that the collapse exists in "classical probability" rather than "classical mechanics".

Generally an information update is when prior information is facing the problem of howto account for a new information. One can speak of this in general terms regardless of wether the "information" is encoded in classical commuting structures, or in sets of non-commuting information.

In both cases do we have an prior state that AFTER incorporation of the new information, results in a new prior state. This is generically a "collapse".

In classical logic, there are still several thinking of what the best information update is like. We have bayesian updates, we have max ent updates and other more simple updates. In this process the idea is that you need to WEIGHT your prior state against the new state, and decide that for a given confidence in the new information how much do you allow it to deform your prior?

A simple form of information update is simly to always throw away the oldest datapoints and then recalculate the prior. Other method uses measures to calculate the "probability" that the new information is "right" and then allows it to influence the prior in the same way. This means the information update process possesses inertia.

It's simply not rational to let say a single data point, flip decades of solid experience. It probably takes another decade of deviations to significantly revise the prior here.

So there are two points here... one is to understand that the collapse in the above sense is not specific to SUPERpositions, it's there also with classical mixtures.

The difference is rather in the details of how the information update works. In Newtons mechanics, I think it's confusing to talk about information updates in between marbles etc. Because there exists no decision process in how it's usually understood. In QM OTOH, the "information updates" as seen from an external observer are sort of encoded in the hamiltonians. This is why my analysis unavoidably suggests that the interesting parts of this discussion overlaps with the understaning of interactions and unification.

Here there are of course significant difference between classical and quantum inferences.

This is exactly why in classical statistics (not mechanics) ie. thermodynamics, we actually understand time evolution as related to entropic flows. We need no hamiltonians to do statistical mechanics. The primary thing it he entropy measures. But then we only predict "trivial" decay & diffusion type flows.

When we look at quantum structures where information are composed of non-commuting parts, then this decay type of flow translates into very non-trivial time evolutions (read schrödinger equation).

Of course this thinking is trying to push current physics much deeper. It's the perspective that I see as the core of the discussion but when we avoid it and just try to not change the framework the entire discussion becomes strange.

In a way the unitary evolution of QM, is always known to mathematicallt look almost like diffusion equation, but the similariy is more coincidental and no one every properly undertstood it beyond that. That is what I'm trying to do, and one implication is that just like in classicla mechanics there is a flow just from uncertainty, the SE equation can be understood the same way, except it's significantly more complicated becase where not dealing with simple probability spaces for encoding information.

/Fredrik

 

[Chronos]

Ken, I think you hit the nail on its quantum head. MWI is not falsifiable by any test of which I am aware - which makes it sound more like theology than tautology to me.

 

[Hurkyl]

On the contrary, the hypothetical theorem I refer to would certainly be viewed as new physics! I'm saying we have all the postulates of classical physics, but it is discovered, to our amazement, that no matter how hard we try, we can never measure x and p to better simultaneous precision than delta x * delta p = h, where the value of h comes as a complete surprise to us.

While it would certainly be a place to look for new physics, it's clearly not new physics yet -- it's just an observation. It doesn't make any quantitative predictions that can be tested, there isn't any theory to explain anything, there is just the observation that engineers are consistently having difficulty designing measuring devices that should be theoretically possible according to Newton's laws, and the postulate they will never be able to get past a certain barrier.


QM, on the other hand, has a theory to explain things.

Indeed, it is perfectly possible that it might have gone just like that, well before quantum mechanics, had h been a heck of a lot larger (or our instruments a heck of a lot more precise).

This doesn't sound plausible at all.

There have been a lot of results over the past century that detail, quantify, and empirically confirming a myriad of ways in which nature most definitely does not act as if it was "classical, but with uncertainty and statistics!"

I'm saying had that been the case, exactly the same 3 ways of interpreting it would have appeared, and for all the same reasons

I don't see how the reasons could possibly be the same.

The starting point of MWI, for example, was Everett's research based on the observation that mixtures appear as an emergent property of large quantum systems, thus explicitly providing a potential mechanism by which the quantum mechanical description of large systems could approximate the classical description.


Your hypothetical situation simply doesn't have any of those relevant factors. You need let the Einsteins, Schrödingers, Hilberts, and Plancks of the world have a decade or so to study the observation before you could get to a point where the question of interpretation could possibly become a serious one.

(empiricists would say there is a collapse, determinists would say there were hidden variables, rationalists would say the ergodic outcomes predicted by the postulates of classical mechanics must actually occur but we don't perceive anything beyond our little h-bin).

I disagree on all three counts.

The good empiricists would say "we need to do more experiments" to get more information to help them understand what's going on.

The good rationalists would say "we need to look for new theories that can explain these new results, and still reproduce what we already know!"

(and, of course, they would ideally operate synergistically, rather than totally independently, and many would be both)