How valid is the indivisible interpretation of quantum mechanics?

[javisot]

What concern do you have with entanglement and this correspondence?

Barandés' interpretation does not offer the local explanation that we would so much like for any form of entanglement. (Sarcasm mode on)

 

[pines-demon]

Barandés' interpretation does not offer the local explanation that we would so much like for any form of entanglement. (Sarcasm mode on)

More like Barandes interpretation does not want to define its properties in terms of usual entanglement terminology. I still don’t know if it super deterministic, hidden variables non local or something else (à la MWI). He wants to define it under Barandes made-up terms, by which nonlocality does not even exist to begin with.

 

[Fra]

I think at the heart of Barandes proposal and insight (and I fully agree with him even if there are missing pieces of his puzzle) is that Bell's definition and use of locality is not useful to understand natures causal interconnections! It is rather a reciepe for confusion to insist analysing things in those terms. This is why he introduces something he think is better - causal locality.

Of course to those that don't like Barandes ideas, this sounds like a way to just "reinvent established terms" and cause more confusion. But I think that is not the case. I think he he serious that analysing things using bells definitions is a problem, because the tools and paradigms we use to map things out, may cause artifacts, that are not artifacts of nature but from us trying to find explanations with the wrong ansatz, so that the missing pieces noone denies, ends up in the most akward positions.

/Fredrik

 

[iste]

Where does he claim the probabilities for trajectories exist? As far as I recall he has said that trajectories exists and is real, and probabilities for transitions - not trajectorires exist.

This is not a contradiction or problem for me though, because the ontology of the trajectories traced out by a subsystems configuration, is itself contextual to the subsystem itself (for me this is conceptually analogous to beeing contextual to an inside observer). The probability OTOH, is contextualy to the the WHOLE system (or conceptually analogougs to an external obseerver). If objectives probabilities of trajectories existed, that would as far as I see risk making it the precise type of HV theory that are ruled out by Bells theoreem.

/Fredrik

I might try to find exactly where in videos but pretty sure Barandes talks about the towers of probabilities for a stochastic process and that we don't have epistemic access to them and Albert I think tries to get Barandes to say they are not specified by the theory.

 

[Morbert]

I might try to find exactly where in videos but pretty sure Barandes talks about the towers of probabilities for a stochastic process and that we don't have epistemic access to them and Albert I think tries to get Barandes to say they are not specified by the theory.

I would like to see your source for this because it would be very surprising to me.

 

[gentzen]

I might try to find exactly where in videos but pretty sure Barandes talks about the towers of probabilities for a stochastic process and that we don't have epistemic access to them and Albert I think tries to get Barandes to say they are not specified by the theory.

I would like to see your source for this because it would be very surprising to me.

1:11:36

So the second point is there's some more structure. There's modal structure. There's probabilistic structure in this theory and there has to be because we need probabilities to sew things together. Here I'm very much motivated by my older work on the modal interpretations which suffered from a lack of the structure and from the Everett question mark theory which lacked the structure. A sort of connective tissue connecting one time to another that lets these wild fluctuations happen. We don't like that. Okay. So here's the second point. What I'm saying is that each system we want to consider whether it's open or closed in addition to having configurations forms a stochastic process. Now a stochastic process is a fairly old idea. Um stochastic processes go back the better part of a century. What is a stochastic process? In general, a stochastic process I can give you the mathematical definition, but I'm not going to do that. Won't be interesting to people. One way to think about a stochastic process is it's some kind of variable maybe representing the configuration of something. And this variable changes probabilistically with time. And this model comes with modal structure. Modal structure meaning like structure beyond just physical like ontological facts at locations in space like other structures. Modal structures include things like probabilities and laws and so forth. There's more modal structure here. These stochcastic processes come with probability distributions over multiple times. Right? So the the model comes with a probability distribution that the variable is some value x at a given time. It comes with a probability distribution that the variable is X at a certain time and Y at some other time. It comes with the probability that the variable is X at a given time, Y at a given time, Z at another time, and so on and so on. There's just this this tower of so-called multi-time joint probabilities just, you know, where where the system is. That's a lot of additional structure. Now often simplifications are made when stochcastic models are used in practice. We assume that a lot of that extra structure is somehow trivial or ignorable and we get a very simple kind of stochastic process called a Markov process. The characteristic feature of a Markov process is that you just have to know what's going on at one time and then you can you have probabilities that tell you what happened later and the ...

 

[gentzen]

I would like to see your source for this because it would be very surprising to me.

You are right to be surprised. It doesn't happen often that people understand criticisms and adapt accordingly:

The discussion did not just "feel" more substantive. What I have watched so far (~1h 43min) was factually a lot more substantive than many previous discussion of Barandes, not just the one with Maudlin.

What is different is that Barandes understood various critisisms of his approach in the meantime, understood that there were implicit assumptions that he didn't spell out, and now tries to explain those to Albert.

Oh no another 2h and half podcast on Barandes. Does he discuss entanglement? I don't want to listen it just to hear the same arguments introduced.

More like Barandes interpretation does not want to define its properties in terms of usual entanglement terminology. I still don’t know if it super deterministic, hidden variables non local or something else (à la MWI).

Looks like you are really finished with Barandes. But that podcast actually really clarifies what Barandes had in mind, but did not spell out.
In the end, he might land a bit too close to Fra/Fredrik, because he doesn't necessarily get laws of physics from his approach, and is actually happy with that.

 

[Morbert]

1:11:36

it seems like if you were to somehow take all of those multi-time joint probabilities to be laws, it would be a huge amount of information you'd have to specify. In principle, an infinite amount of information. And we don't want laws that are infinitely complicated.

The problem with these joint probabilities, denoting observed relative frequencies the omniscient observer happened to observe, is they are not reproducible. The observer could collect a new sample of trajectories, and the observed relative frequencies don't have to be the same. They don't have to be regular.
[https://www.physicsforums.com/threa...opic-theory-of-causality.1080105/post-7305662]

In particular, probabilities assigned to whole trajectories, as constructed from higher-order conditional probabilities in the sense of (10), are then left unspecified as well. The higher-order conditional probabilities of an indivisible stochastic process could, in principle, vary contingently from one set of instantiations or runs of the process to another.

What I was hoping for was these towers of probabilities giving rise to inaccessible probabilities for trajectories that are nevertheless regular or lawlike. I fully accept that there are in principle probabilities for trajectories.

 

[iste]

I would like to see your source for this because it would be very surprising to me.

On that same video there are statements of the sort all through about 1:17 to 1:46, and quite a clear concise summary from about 1:42 - 1:46. Barandes says these trajecyory probabilities exist but we have no epistemic access to them.

 

[Morbert]

On that same video there are statements of the sort all through about 1:17 to 1:46, and quite a clear concise summary from about 1:42 - 1:46. Barandes says these trajecyory probabilities exist but we have no epistemic access to them.

Would you agree that these are the same probabilities he talks about in 2.2 In his correspondence paper where he says "In particular, probabilities assigned to whole trajectories, as constructed from higher-order conditional probabilities in the sense of (10), are then left unspecified as well."

 

[iste]

What I was hoping for was these towers of probabilities giving rise to inaccessible probabilities for trajectories that are nevertheless regular or lawlike

Well, they could be lawlike, Barandes isn't specifying whether they are lawlike or not because he doesn't have any access to them.

Would you agree that these are the same probabilities he talks about in 2.2 In his correspondence paper where he says "In particular, probabilities assigned to whole trajectories, as constructed from higher-order conditional probabilities in the sense of (10), are then left unspecified as well."

Would you agree that these are the same probabilities he talks about in 2.2 In his correspondence paper where he says "In particular, probabilities assigned to whole trajectories, as constructed from higher-order conditional probabilities in the sense of (10), are then left unspecified as well."

Yes, agreed.

 

[Fra]

I might try to find exactly where in videos but pretty sure Barandes talks about the towers of probabilities for a stochastic process and that we don't have epistemic access to them and Albert I think tries to get Barandes to say they are not specified by the theory.

In the end, he might land a bit too close to Fra/Fredrik, because he doesn't necessarily get laws of physics from his approach, and is actually happy with that.

What I was hoping for was these towers of probabilities giving rise to inaccessible probabilities for trajectories that are nevertheless regular or lawlike. I fully accept that there are in principle probabilities for trajectories.

On that same video there are statements of the sort all through about 1:17 to 1:46, and quite a clear concise summary from about 1:42 - 1:46. Barandes says these trajecyory probabilities exist but we have no epistemic access to them.

Well, they could be lawlike, Barandes isn't specifying whether they are lawlike or not because he doesn't have any access to them.

Would you agree that these are the same probabilities he talks about in 2.2 In his correspondence paper where he says "In particular, probabilities assigned to whole trajectories, as constructed from higher-order conditional probabilities in the sense of (10), are then left unspecified as well."

I now listened to the key part where Albert keeps asking what is the fundamental law but isn't satisfied with Barandes saying "it's just stochastic", while Barandes struggles with understanding the category of answer Albert wants, which seem to reason from the traditional equation/system paradigm with initial/boundary conditions. As Barandes points out, the "law" is probably not best cast that way. But Albert don't seem convinced.

I see how the reaons is not convincing from Barandes presentation alone; to someone with a deep stance in the traditional paradigm. I follow Barandes logic, but its because I am already roughly on the same page to start with, I didn't need to be convinced. And my own arguments would necessarily go beyond the discussion and likely cause more confusion.

** In QM standard hilbert picture.This is system dynamics, and DYNAMICAL law encoded in the hamiltonian (or alternatively the equations of motion)

** In stochastic picture.There exists no traditional dynamical law. "Time evolution" is just a stochastic process in configuration space. No time evolution law is needed. But the process is constrained by Gamma,

Although different, I think gamma loosely encode what the hamiltonian encode.

So instead of asking why this hamiltonian, how to understand the internal structure of the hamiltonian, Barandes says its "just a stochastic processs", there is no fundamental "dynamical law". There is no NEED for more. But instead we need to ask, why the particular stochastic constraint Gamma; and what the particular configuration space. No of that is innocent. I think Barandes does not speak much about this because the same "fine tuning" exists i hilbert picture. So this is not a NEW problem, and its not the problem his correspondence is set out to solve, at least not at this stage.

For ME this is a key point, and i think conceptually contemplating emergent gamma is easier than emergent hamiltonian (although it is a different discussion) But I think Albert's objection is simply that he feels there must still be a dynamical law behind a stochastic model (just like there is in statistical mechanics). This is perhaps why he says to Barandes "you cant just say its a stochastic system". But I would say that to really appreciate the stochastic picture, a paradigm shift of thinking about physical theory is needed, as it boils down to the question, exactly WHY is the stochastic picutre better? And what is the difference to the simlpe statistical mechanics and entropic dynamics?

About the higher order probabilities, that one CAN contemplate for general stochastic processes (but that aren't needed in Barandes picture), are IMHO not relevant for normative probabilities. They makes no sense from my perspective at least (as normative in an agent picture), the history should be implicit in Gamma itself. I think the first order transitiions are sufficient, because they answer what i think is the only relevant question. P(future|now) - but, there is more information in here than the state of NOW, in the general case I think P itself evolves. But this is not true for a "stable quantum system". So this discussion gets us to the fringe of things.
Because the previous state history would refer to the systems own path, and from inference perspective, it is intutitive to me that these have no role to play because the system should have learn what cna be learned from its own histroy, or something is wrong. They may be allowed, but I see them as serving no purpose. This is just my hunch about this. I'm guessing there is a reason why Barandes leaves this murky stuff for future work. As it in itself, has no role to play for the stochastic quantum correspondnece as i see it.

/Fredrik

 

[Morbert]

Although different, I think gamma loosely encode what the hamiltonian encode.

Γ indeed encodes the dynamics of the theory. It just doesn't select a unique probability measure on the space of trajectories. It's easy to argue that this is peculiar. It's much harder to argue that this is a deficiency.

 

[Fra]

It's easy to argue that this is peculiar. It's much harder to argue that this is a deficiency.

I would rather say that to me, that having "probability of histories" out of the way is not peculiar, it is a good sign! Not by a remote chance a defiency.

This is because that is imo a "non-physical" context. Which Barandes calls out of empirical reach. But for me that is not a practical issue, it really means its non-physical. And really have no place.

(My confidence/logic behind comes from the interacting subsustems picuture in agent based modelling paradigm; which is stochastic at its heart but has more stuff than Barandes speaks about)

/Fredrik

 

[iste]

Γ indeed encodes the dynamics of the theory. It just doesn't select a unique probability measure on the space of trajectories. It's easy to argue that this is peculiar. It's much harder to argue that this is a deficiency.

Its only not a deficiency if you think its trivial to construct a stochastic process that fulfils all the properties you want it to with regard to quantum behavior. Its like Bohm coming along and saying he has a hidden variable theory with deterministic trajectories and then just refusing to specify how the theory works. You have no idea what kind of consequences you are implying if you don't do that kind of thing. Part of the reason quantum foundations exists is that quantum theory seems to elude classical-like explanations, so to have unquestioned faith in an underlying stochastic description like that just seems ass-backward.

 

[Fra]

Its only not a deficiency if you think its trivial to construct a stochastic process that fulfils all the properties you want it to with regard to quantum behavior.

Its like Bohm coming along and saying he has a hidden variable theory with deterministic trajectories and then just refusing to specify how the theory works. You have no idea what kind of consequences you are implying if you don't do that kind of thing. Part of the reason quantum foundations exists is that quantum theory seems to elude classical-like explanations, so to have unquestioned faith in an underlying stochastic description like that just seems ass-backward.

As Barandes does not construct the stochastic process - he has provided a correspondence to QM, and the gamma is constructed from the correspondence using the hilbert/hamiltonian, the scepsism is expected and rationa, similar to Alberts thinking.

[This is and was my first objection to Barandes view as well that i noted in one of the first threads on this the very first time i say his papers. So Barandes work itself - is NOT what is behind my confidence. But I soon realized it's value in a correspondence that to recover the standard model, it is enough to recover more stable unistochastic processes, which can in turn be argued to be a limiting attractor of more general processes that are not unistochastic, which I think will be evolutionary and need to be constructed differently, and here the system dynamics paradigm is unsuitable to start with, Barandes correspondnece just fits nicely into these existing ideas as a "correspondence" checkpoint. If I did NOT have this in my bag, I would likely be more sceptical to Barandes, I admit]

Convincing someoneone that thinks differently is very difficult until the arguments are 100% mature, and you have a computable checkable model on the table. As we know too well this is why similar thinkers tend to aggregate spontanously in research fields where their fellow research will understand the incomplete and not yet mature ideas.

Can you pinpoint the top thing that you find most impalatable with the stochastic picture, perhaps we can discsuss that one thing at a time. If its a "probabilities" then I suggest looking beyond the mathematical definition of probability and considering what the purpose of a the measure is, and purpose for what context, and how is ot constructed from what input? I'd say that from the perspecitve of a subsystem in stochastic picture, the transiution probability defines the "dice" of the random walk. And as that takes place, one step at a time, there is not meaning in considering the "probability of future trajectories". It would make not difference to the stochastic process itself! Because the dice is remodelled after each step, anything else would not be optimal. And as I understan QM, it does represent in some way a not yet understood an optimal process. If we can understand this optimization when maybe a real construction of the stochastic process will be possible. And in doing that, having Barandes correspondence will serve as a valuable "checkpoint".

Edit: I thinking reconstructing the stochastic process is hihgly non-trivial. The confidence for me is more rooted in where I see the most promising way forward towards the difficult to reach goal. Currently paradigms and models are IMO mainly plauged by unnaturallnes and alot of unexplained fine tuning, that moreover seem to explode as you try to enlarge the theory and unify with gravity. There is where my journey started, basically the failure of things everyone has been working on for decades told me there is something wrong with it, not at detail level, but at systematic and paradigm level.

/Fredrik

 

[Morbert]

Its only not a deficiency if you think its trivial to construct a stochastic process that fulfils all the properties you want it to with regard to quantum behavior. Its like Bohm coming along and saying he has a hidden variable theory with deterministic trajectories and then just refusing to specify how the theory works. You have no idea what kind of consequences you are implying if you don't do that kind of thing. Part of the reason quantum foundations exists is that quantum theory seems to elude classical-like explanations, so to have unquestioned faith in an underlying stochastic description like that just seems ass-backward.

No faith is involved. The correspondence is now the topic of multiple papers. It's there plain as day. If people don't think a quantum system can be understood as a general stochastic process, the reasoning is there in literature for them to attack.

You'll have to be more specific about the kind of consequences theories of generalized stochastic processes imply.

 

[Fra]

No faith is involved. The correspondence is now the topic of multiple papers. It's there plain as day. If people don't think a quantum system can be understood as a general stochastic process, the reasoning is there in literature for them to attack.

You'll have to be more specific about the kind of consequences theories of generalized stochastic processes imply.

I am guessing the presumed objection is not the correspondence itself.

But that the a TYPE of indivisible stochastic process is very different from the divisible old style random processes.

Thus the "faith" may be required - in order to work more in stochastic pictute - in wether this indivisible stochastic picture is better suited to allow for a better explanation of quantum weirdness to be found than hilbert/wavefunction picture. Admittedly just ny itself, gamma itself is not very intuitive, neither are the rules for splitting or composing.

Barandes has no first principle motivation. I have faith in that we will find one so i like it. Without such faith i may see it as an reformulation that changes one weirdness for another weirdness

/Fredrik

 

[iste]

Can you pinpoint the top thing that you find most impalatable with the stochastic picture

I don't find anything impalatable with a stochastic picture. I think its probably the most desirable kind of hidden variable description. But Barandes doesn't give us an actual stochastic process if it only carries one-time transition probabilities from a fixed initial time and nothing else of the other stuff required for a stochastic process; and the problem with that is that you have no theory, you can't assess, clarify, verify its properties. You don't really know what you're referring to or even if it is actually feasible on the way you want formally.

But that the a TYPE of indivisible stochastic process is very different from the divisible old style random processes

I don't think the "indivisible stochastic process" is a stochastic process. I think Batandes has probably made a mistake here in thinking that his correspondence is describing a correspondence between quantum theory and a stochastic process. But its not. It doesn't have properties of a stochastic process and you can't pretend it does without postulating probabilities that don't come out of the original correspondence. Without those probabilities its simply not a stochastic process, and the correspondence doesn't give you one.

No faith is involved. The correspondence is now the topic of multiple papers. It's there plain as day. If people don't think a quantum system can be understood as a general stochastic process, the reasoning is there in literature for them to attack.

You'll have to be more specific about the kind of consequences theories of generalized stochastic processes imply.

The correspondence is one between an operator formalism and transition probabilities that isn't really fundamentally any different from how the Born rule works.

And if the other probabilities that define the postulated underlying stochastic process are not described anywhere, then the point is all kinds of consequences are possible which you won't know unless youhave some kind of theory for that underlying process. In the Albert conversation, Barandes cites some guy Gillespie as having tried but not succeeded in making a non-Markovian quantum theory, and also I think Barandes says he said one could not formulate quantum theory as a kind of Markovian stochastic process.


If you have these kinds of failures and demonstrations, and I'm sure there are other examples of people doing yhat, then clearly just postulating that probabilities exist doesn't really hold any water that you have a credible interpretation - people have every right to say that: without formal demonstration, it's not even clear that what is being proposed about an underlying stochastic process is possible. You can say the stochastic process is "not lawlike" but then surely one should still demonstrate that a "non-lawlike" stochastic process can reproduce not only correct quantum behavior but behavior of the kind you want. In the most successful stochastic formulation of quantum theory, trajectories for many particle wavefunctions cannot evolve in an independent way. If you don't know what the underlying stochastic behavior is like or some kind of formal results, then how do you know it won't also require similar non-local behavior in order to reproduce quantum behavior with non-local correlations? If it does, then that is a complication because it requires a space-time foliation which then has further metaphysical complications. And sure, in the most successful stochastic formulation (at the very least the latest), measurements at the ensemble level will still have the normal non-signalling properties of normal quantum theory, but it is with the introduction of trajectories that the non-locality of the stochastic process becomes overt. You simply don't know if similar applies to Barandes theory without at least constraining some properties of the underlying stochastic behavior and formally demonstrating it.
[/QUOTE]

 

[Fra]

I don't think the "indivisible stochastic process" is a stochastic process. I think Batandes has probably made a mistake here in thinking that his correspondence is describing a correspondence between quantum theory and a stochastic process. But its not. It doesn't have properties of a stochastic process and you can't pretend it does without postulating probabilities that don't come out of the original correspondence. Without those probabilities its simply not a stochastic process, and the correspondence doesn't give you one.

I see what you mean.

I would say the conceptuial difference between indivisibile stochastic and what you think of as regular stochastic process lies in the the difference between an objectively descrpitive and normative probabilities; where normative is contextual to the system. Meaning, each subsystem, which each configuration spaced has its own normative probablity. And when the total system of "interacting subsystems" - meaning we have in a way a system of interating stochastic processes, the overall total stochastic process cant be described with objective descriptions of the path. This is again conceptually the same reason why it evades the premise of bells anzats. The real paths of the configurations are not "objective hidden stochastic paths" in the bell sense.

This is where i think there is a bit of a paradigm change in thinking required to accept this. The conventional picture is an objective stocahstic process where the observer is just ignorance of the hidden variables. This si the bell ansatz. Ie you have some sort fo objective "dice".

In this view there is some sort of "system level" stochastic. But in Barandes view I would say the system is composed of "stochastically evolving" subsystems, each system has its own "dice" and it's when these INTERACT you get quantum weirdness. Quantum weirdness would not happen in a "regular stochastic process".

Causal law is replaced by stochastic normative processes. This why the stochastic process of one subsystem, is never influenced by another subsystem, because the normative probabilities are independent, except that they MAY be correlated. So in entanglement, the two "independent stochastical processes" are correlated without pathologies. It. their configurations are not correlated, only the normative transition probabilities.

Why this is "natural" - then consider the world of consisting of "interacting stochatic parts" like many parallell stochastic processes (that are tuned as defined by gamma). As opposed to a single total system level single stochastic process.

Here is a conceptual link to single processing vs paralell processing; which is also conceptaully analogous to system dynamics vs Agent based model interactions.

The indivisible stochastic process makes not sense if you try to understand it as system dynamics. But it makes more sense if you associate it to the agent/subsystems interacting with fellow subsystems, and each agent is a priori independent, except that they may be historicallty tuned.

/Fredrik

 

[Morbert]

@iste Your last post was hard to follow. All I can gather is you seem to be insisting theories of stochastic processes must include unique probability measures on spaces of trajectories: Unique relative frequencies. But you never say why. You only allude to consequences without example.