Everybody sees the same elephant (says Carlo Rovelli)

vanesch

To summarize our differences: Rovelli (and you) accept the statement that "everybody sees the same elephant", while I think that what has been demonstrated is only:
"everybody sees everybody agree on the same elephant"

The difference is of course subtle, but in the last case, there's no reason that everybody really agrees upon the same elephant, but only has a consistent picture of its interactions with others, and is led to believe that they saw the same elephant. As such, everybody can see a different elephant, and different "others", and nevertheless be convinced, each one in his own world, that everybody agrees with him.

This is what is REALLY demonstrated (by those equations (6) and (8)). This is the internal consistency (which was already noted by von Neumann) of observation.

But I'm not basing myself on any specific Hilbert formulation of the issue to derive this, I base this only on the rejection of the notion of objective state of an observer. Only an objective state of an observer could contain the necessary information about what he really observed, and be accessible through interaction to another observer, and this objective state is denied to exist (not only because it would wreck havoc to the claim that, for another observer, the former observer is a quantum object, and hence DID NOT pick a particular outcome, but also, to avoid the problem with Bell's inequalities).

So if the state a|o1>|s1> + b|o2>|s2> from P's PoV and |s1> from O's PoV are BOTH correct descriptions then one can say two possible things:
these are only descriptions of OUR KNOWLEDGE. Indeed, P doesn't know yet what O did, so is still in "suspension".
Or we can say that S (and O) simply do not HAVE an objective state.

It was my understanding that R&S go for the latter, and they have to, because the former would only be a description of KNOWLEDGE, not of physical state, and would not stop us from ASSUMING that there is an objective physical state. This assumption would then wrek havoc for as well the "all quantum" view, as for Bell. Also the claim to completeness goes in this direction.

Now, is a |o1>|s1> + b |o2> |s2> is the best we can do, from P's PoV, not only to describe any knowledge, but of what could potentially be the physical state, then *there is no way for P to pick |o1>|s1> for sure*.
This happens only with probability |a|^2. There's no physical mechanism by which S can "inform" P about its observation of o1, because, from P's PoV, both possibilities are still present and they are NOT about ignorance, they are really physically present.

That's why I don't see how one can claim that P will obtain |o1>|s1> for sure, if S has picked |s1>.

What has been demonstrated, however (equations 6 and 8 you quoted way back in this thread), is that P will observe only o1 together with s1, OR o2 together with s2. So P will find "agreement" between his observation of O and of S, and hence deduce that "S agrees with him". But he doesn't know what S observed, really, because this information is not "physically present" BY DEFINITION. If it were physically present, the STATE for P would NOT be a |o1>|s1> + b |o2> |s2>, but rather |o1> |s1> with (Bayesian) probability |a|^2 and |o2>|s2> with (Bayesian) probability |b|^2, where the Bayesian probabilities do not descibe an objective state, but only P's knowledge.
The objective state would of course be |o1>|s1>, but P would simply be ignorant of it.
And we're back to Copenhagen.


I was only referring to Rovelli's "vector of answers to binary questions" on which he builds a quantum theory as describing the correct probabilities for the next answer. So the only thing that exists here, are "answers to binary questions".

Yes, but that's "by sheer magic" if postulated, because IF indeed {A, \alpha} is independent of {B,\beta} there's no reason for them to agree afterwards.
So, or they don't agree (but aren't aware of it), or they agree, and then the question is: given that {A,\alpha} and {B,\beta} didn't influence eachother, how come ? The question has then not been solved at all.

You could say:
From A's PoV, only {A,\alpha} and {A,B} happen. They agree. This is found by the quantum formalism (where B is treated as a quantum system, which did NOT get an explicit answer back then, but just entangled).

From B's PoV, only {B,\beta} and {A,B} happen. They also agree. This is also found by the quantum formalism, where this time, A is treated as a quantum system which did NOT get an explicit answer back then, but just entangled.

However, there's no way to make both magically equal without saying that {A,\alpha} REALLY resulted in one result or the other, and {B,\beta} REALLY resulted in one result or the other, and ALREADY AGREED BACK THEN. And then we're back home, with EPR correlations and all that.

Well, no. That's the point: it is only shown that C will OBSERVE A and B as having the same results.
A might observe C and B to agree also upon the same results (although they differ from the first set), and B might observe A and C to agree also upon the same result (which might still differ from the first and second set).

That is what is really demonstrated.

Assuming that they are equal is an extra assumption, which, IMO, goes against the statement that the observation result of B is not an objective result, coded into its state, for everybody.

That's the difference between "everybody observes the same elephant", and "everybody observes the others as observing the same elephant".


So do I. I think the derivation of the quantum formalism from the 3 postulates is nice, and coherent. But, as I said, this only pertains to the "answers" vector of a single observer, and nowhere the link is really made to the "answers vector" of another observer.
So it looks like me to be a theory about what an observer can know about the world.

marcus

Hi vanesch, there is a problem with word order in the last sentence of your post which makes it difficult to understand. Could be a simple typing error. Is this what you meant to say?:

If that is the desired meaning, then slightly more idiomatic ways to say it would be:

or

(two common idioms are "looks like" and "seems to be")

vanesch

Yes, sorry. Some of my posts are typed while I'm reading simultaneously about dinosaurs for the kid and discussing with my wife

BTW, I just finished almost "Three roads to QG", and it occured to me:
isn't the "solipsist" version of how I read RQM, just an extreme version of the holographic principle ?

marcus

I too must sometimes quickly finish typing a post while I am discussing something with my wife.

about your BTW question:
you have to get selfAdjoint to comment on that.

My tentative opinion is like this: the holographic principle seems to be about INFORMATION and maybe also about boundaries or horizons.

the idea of a boundary or horizon is connected to the idea of an observer. you sort of define an observer, in some contexts, by putting a plastic bag around the system and say that the observer is somewhere outside that plastic bag. In a strange way, in some contexts but probabably not in others, the observer IS the plastic bag, and what she or he observes is the information that reaches the plastic bag and maybe comes through it.

so I have to answer I DONT KNOW if what you are asking about is the same, but I do understand you that the concern is about information (as I gather also is the central focus in quantum mechanics whether Carlian or not)

vanesch

What I meant was the following. SA and you seem not to agree with me, but I think that the RQM formulation is entirely "observer" centered, and is about information an observer has, to himself, which is not available to eventual other observers.

Now, the (strong) holographic principle, as I understand it, comes down to saying that whatever we can know about anything, is entirely determined by the closed surface around us ; or even (strong version), that that is in fact the only thing that really is, and the "volume behind it" is just an illusion created by the information flowing in from the surface ; with suitable reformulation, all physics is the physics of the surface around us, observers.

Now, consider each "observer" in its own "plastic bag", thinking there's an entire universe behind it, with other observers walking around in it and all that, but he looks just at the physics of his plastic bag ; nevertheless, he'll see his "peer observers" on his plastic bag agree with him.

And so each observer lives in his own plastic bag, with his own "knowledge" and information flow coming off it, thinking there's an entire universe behind it.

This is then nothing but the "subjective worlds" I talked about - though now they become indeed "individually objective" (but surely with some form of solipsism to it).

selfAdjoint

Well is relativity solipsism? It seems to meet your definition; there is no absolute energy, length or time seen by everybody, rather energy, length and time are "oberver dependent". Each observer in his own rest frame sees a length, energy, and time that is "ordinary" for him, but no other observer who has a velocity relative to the first one will see those values. And this is not just illusion; as we well know from experiment, the different definitions in two "interacting" frames have real consequences for the observers in them.

So according to your definition hasn't physics been solipsist since 1905?

vanesch

Of course relativity (as everything) can be formulated in a solipsist version, but the fact that different coordinates are used by different observers is usually not seen in this way, in fact quite the opposite: relativity says something else: all events are THE SAME for everybody, and it are the local descriptions that are different (sounds familiar, no ). However, the concept of an event is ontologically postulated, and the same for everybody.

Now, he idea of describing events by different coordinate systems doesn't come from Einstein, but from Galileo ; Einstein just made it a bit spicier by adding time to the mixture. But does the use of different coordinate descriptions imply solipsism ? No! It only means that the description of the events by observer O is isomorphic to the description by the events by observer P. The isomorphism is a lorentz transformation of course.
That means that observer P has no difficulty to know how observer O would describe certain events. He gives it different coordinates, but he knows how one maps upon the other.
And then he realises that all these different, isomorphic, viewpoints have an underlying objective reality, namely the manifold of events.
Now, it can be that certain events, say event A, in P's past lightcone are not in O's past light cone, but nevertheless, P knows how O can describe this (even though O doesn't have any information about it yet at this moment). And if P and O do meet later, (when O and P finally have the information about the event A, they will indeed come to the same conclusion about the coordinates of event A: P will find that O indeed gave it the coordinates he though O would.

So it is perfectly all right, in relativity, to give ontological status to events, and even to the description by P of an event. It can be ontologically stated that event A, in P's description, takes on these coordinates.
That's an ontologically true statement, which is true as well for P as for O (O will give A other coordinates, but can derive that P will give it the coordinates in the statement, and hence agree in principle about it).

All ignorance about the truth value about certain such statements (because of lack of information) is just that: ignorance. But it doesn't lift the statement from its ontological status. It is not because you ignore that I'm drinking a beer, that it is not ontologically true that I'm drinking a beer, for instance. And we agree over that later.

But the RQM description isn't that. There is no isomorphism between the description of the situation by O and by P. O seeing P in a superposition is not related, through any transformation rule, to O seeing the system in a particular outcome ; and as you know, the "superposition" is not an ignorance description, because otherwise there would be no distinction between superpositions and mixtures, and there would be no quantum interference.

So there is no underlying ontology (a statement which is even explicitly made) - contrary to the manifold of events in relativity.

If there is a transformation rule of descriptions, then we can define "ontological state" as the equivalence class over this transformation rule, of all descriptions. This is what the lorentz transformations do in SR, or the general coordinate transformations in GR.
But there's no such rule in RQM.
(there is, in MWI: it are the unitary transformations).

selfAdjoint

Just a very quick comment on this. If we accepted a four-dimensional equivalence relationship instead of restricting ourselves to three dimensions, then we could define observation classes of O and S that were equivalent in giving the same resolutions in the future. the pair O/superposition S/singlet would be one element but there would also be one with O/singlet and S/superposition.

And on the knowledge-amplitude gap; nobody wants to introduce complex-valued knowledge, which is what you would have to do to satisfy the properties of amplitudes with a knowledge based explanation. Feynmann notably tried to interest his comperes in the idea of complex probability; I would rather begin with complex valued memory, memory with a phase. In any case I think the amplitudes have the ontological character of information or data.

Finally, I am rereading the S&R paper with the emphasis on seeing if their dvelopment of states allows them to be somehow preservd over time. Again my first impression was that they only existed in interactions (which are as ontologically real according to S&R as relativity interactions). That would require continuing self-interactions to support the state until the next outside interaction. I think of a boy tossing a ball back and forth as he walks down the street; we could pretend that it is not possible for him to just hold the ball, so if he wants to keep it by him he has to keep tossing it.

vanesch

I couldn't agree more. And know what these equivalence classes look like ?
Wavefunctions + time evolution a la MWI (+ a "token" for each observer in the story)!

Of course I cannot work out this equivalence class stuff rigorously, but my reasoning is that all the "I know it is" and "I know the other guy's in a superposition" kind of observer-world relationships are correctly described that way using ONE MWI description, hence this must be a correct description of the equivalence class.

The only difference is in fact in "at the end of times, all observers see the same elephants", versus "at the end of times, all observers see all other observers agree on the same elephants they saw" (which is the only statement we can really verify).

selfAdjoint

Of course if you make the superimposed states (mathematically represented by vectors in a Hilbert space) "real", then you certainly have no problem with reportablility. But don't you run into the basis problem? If we restrict ourselves to the spectra of the interaction operators it seems we avoid that, no?

vanesch

Yes, sure, because it is now put in by hand in RQM. The "answers to questions" now specify (as in von Neumann/Copenhagen) *intuitively* what are the pointer states, or the basis states in which we should "project" to obtain the probabilities for answers. Nothing wrong with that, but then one shouldn't point this out as a *problem* for another approach, if you need to go to irreducibly intuitive concepts yourself to answer the same question.

That said, I'd say - and I'm not 100% clear about this myself - the hope of an MWI/decoherence approach is that there is some natural solution to the basis problem when we split the universe in observer/restofuniverse.
If that doesn't work, we can still put it in by hand (by defining what exactly are the "states of the observer that are to be associated with definite awareness") - just as does RQM when positing the basis states which give the "answers to questions", or von Neumann/Copenhagen which specify the Hermitian operator corresponding to a "measurement apparatus" entirely intuitively. MWI would not be worse off this way than the alternatives (which do the same, without any hope for reducing this intuitive input), but surely would be uglier than when there would be a natural appearance of "basis" - as decoherence does seem to suggest, but I'm not yet clear about how much of this is "marketing" and how much is hard results from decoherence. I thought the situation was rather promising, but I'm not sure.

selfAdjoint

Well I take (I hope not too quickly) as a major concession from you that RQM is now seen as a contender with MWI, sharing some of the same kind of problems, rather than as simply wrong, which I took to be your earlier position.

The key datum we need to clear everything up, the true nature of the amplitudes, remains beyond us, and we wind up putting some feature or other of their phenomena "in by hand". The differences are all in which feature we pick and which we try to get rid of. For you and MWI it's projection; you want the wavefunction to be real and evolve unitarily. For RQM, projection is OK because they take the information view of the amplitudes (with all the accepted problems of doing that, which they try to finesse), but they try to eliminate nonlocality.

vanesch

If you go back, you'll see that I never claimed RQM to be wrong, I claimed from the start it to be the "single-observer" version of MWI, in the sense that everything done in RQM is what a single observer in MWI would agree upon being "in his branch".
So I saw from the start RQM as the "what does a single observer see" view on MWI. As such, it is left in the middle what could be "ontologically" true, and the total wavefunction is simply not "something RQM wants to talk about". Not because it has, or it hasn't, any meaning, but because we decided to limit ourselves to what is seen by a single observer (and hence describe this single, subjective world of that observer without caring about any objective reality or not).

As such, this is interesting, but doesn't address several issues MWI tries to address. Indeed, RQM has already introduced a "preferred basis" (namely the one of the vector with the yes/no answers), and has already postulated a probability rule. Moreover, RQM does NOT try to give any description of an ontological reality, it only describes a subjective observer world.

I only protested against the wordings which try to introduce "objective reality as agreement between observers" and at the same time "the rejection of the objective existance of an ontology" and all that verbal game, which vehicle the impression that RQM is onto something new, which might *solve* longstanding issues. It doesn't. If you take on the PoV that RQM is a "new way of seeing things" without giving an ontology to things like "the state of an observer" and so on, you create more troubles than you solve, but they are put under the carpet by a subtle redefinition of concepts like objective and subjective, and by dancing around with the concept of the state of a system. So I protested against THIS view on things.

RQM is perfectly fine as the description of the subjective world of an observer - including the appearance, in this subjective world, of other observers and their observed (within this subjective world) agreements.

I tried to outline that by puzzling all these different subjective views together, we can arrive at an ontological, objective description ; which is exactly the one given by MWI (although, there, the path is taken in the opposite direction: we start from an objective system, and we try to deduce (many) subjective worlds from it).
All the statements made in RQM make perfectly sense from this PoV (but are in fact tacitly already assumed).

But the real problem, namely, the preference of one over several branches, (which is refuted in MWI, hence the "many", and which need projection elsewhere, or needs a classical token, as in Bohm), is not more addressed by RQM as it is by Copenhagen or by MWI (which tries to live with it).
The (correct) point RQM makes is that you are locked up in your own subjective world, and within that subjective world, all is fine (but we knew that already from MWI).

selfAdjoint

"Refuted?" Surely over the top Patrick? Nobody is proving anything here. "Disputed" would be better.

vanesch

I meant "rejected". MWI explicitly rejects the idea that "one branch emerges objectively", while this is exactly what Copenhagen-style interpretations do (= projection), as well as Bohmians (which keep both: they KEEP all the branches in their ontology, but one branch gets a "token", namely a strong non-zero value when we take the in-product with the position-states corresponding to the "classical position variables" - while all other branches give essentially 0 in this in-product ; there's no preferred basis problem here, because in Bohmian mechanics, the position basis is in any case preferred).

What MWI does, however, is to say that *from the point of view of an observer (= state)*, there is of course a branch emerging, but that's *subjectively*. It is the branch which contains said observer state.
What RQM does, is saying that *relative to an observer*, projection occurs or not depending on whether this observer has information.

Now, if you translate "observer has information" into "specific observer state containing the information", then it is really a game of words to say that the viewpoints are really different.

The only difference between MWI and RQM is in some word games around "subjective" and "objective". RQM starts from observers with information (= the MWI observer STATES with information), and accepts the fact that, depending on the information, the description of another system can be different from one observer to another. MWI does the same, and tells you that, according to the branch, relative to an observer state, the factor in that branch corresponding to another system can be different (the *relative state* of the other system depends on the state of the observer to which it is relative).
MWI and RQM both agree (and use the same formal demonstration) to show that the information state of another observer will be in agreement with the information state of the original observer and the information he gathered elsewhere.

But this is interpreted in a subtly different way: MWI shows that the *relative state* of another observer in the same branch as the original observer will be such that both are in agreement, but allows different versions of the COUPLES of observers (agreement within each couple).
RQM uses the same formal demonstration, but now focusses on ONE SINGLE COUPLE, and concludes that other observers agree, upon interaction, with the original observer.

And then the big theatre trick comes in: the different, mutually agreeing couples correspond to different subjective realities in MWI (all part of one objective reality), and the fact that there is agreement (within one couple of observer states) in RQM is called "objective reality".
So what were different subjective realities in MWI, is now, because of MUTUAL AGREEMENT WITHIN, called, "objective reality", and this is where a logical error is made in the exposition of RQM (but not in RQM itself, only in the exposition).

It is shown that there exists an A1 and an A2, and a B1 and a B2 and a C1 and a C2. It is then shown that Ai agrees with Bi agrees with Ci agrees with Ai. And this "common agreement" is called "objective reality".
However, it should have been SHOWN that there is ONLY ONE set possible, because we now have that A1 agrees with B1 agrees with C1 agrees with A1, but we also have that A2 agrees with B2 agrees with C2 agrees with A2 (which is evident in MWI and "hidden" in RQM). The demonstration that RQM uses to show that Ai agrees with Bi agrees with Ci agrees with Ai, under unitary evolution, concludes exactly that: that there are different "sets of agreement".
But the trick to call the "common agreement" of Ai with Bi with Ci with Ai "the objective reality O" (without an index i), is the logical error which allows one to "objectify" suddenly Ai's result, in agreement with O, with Bi's result, in agreement with O, with Ci's result, in agreement with O.

To give a simplistic illustration of the erroneous reasoning used here, consider the following:
we have 3 sets of numbers, A, B and C.
Now, we can show that if b is in B, then b/2 is in A, and if c is in C, then c/3 is in B. Moreover, we can show that if c is in C, then c/6 is in A.
Conclusion: we now know the element of A, B and C: namely a, 2a and 6a respectively. So there's now one element in each.

Counter example: A = {1,2,3}, B = {2,4,6}, C = {6,12,18}
The properties hold, but there's no unique element in A, B and C.

Nevertheless, this reasoning is exactly what is done in the exposition of RQM to introduce "objective reality". From the demonstration of the "agreement of observations" (the if b is in B, then b/2 is in A etc...), it is concluded that there is an objective (and hence unique I presume) reality of agreement.

selfAdjoint

I don't think the RQM "observers" at all correspond to MWI states. Maybe I am differing from the authors of the Relational EPR paper but I rely on their statement that only the interaction is "real". State-vectors, kets, even density matrices (I think), all those are a kind of knowledge or conceptual thing. a "shape that the interaction conforms to" in less human-centered language. And the resultant real value observed has to be preserved for later comparison, which I think can only be done, if this picture is to be sustained, by successive self-interactions (cf. my analogy of the boy tossing the ball from hand to hand because he was unable to just hold it). None of this in my view corresponds to what I take to be the MWI reification of the state-vector. And hence the discussion of results later in RQM as I understand it is no stage trick but a completely supported physical interaction. Correct me if I am wrong.

TopcatBV

Hi,
I read your posts and in my opinion RQM is a MWI variant, thanks to vanesch stubbornness.
Though, I'd like if you can repeat the experiment presented in Rovelli's paper and explain again what observer C will see (in both theories) about the two detectors (A&B) measurement of two entangled particles (a+b=0) generated by a single source. I understood that, in both theories, a-A and b-B interactions will transfer the a-b entanglement to A and B.
Say the C observer will see this superposition, 50% randomly a+b- or a-b+, before A sees B's measurement and B sees A's measurement.
A will "measure" B and there is 50% chance to get a+b- or a-b+
B will "measure" A and there is 50% chance to get a+b- or a-b+

In QM everything is fine, but I don't think I understood how the classical world is arising.

Can be said that:
-A sees B as a quantum object until he gets information of what B measured. B turns in a classical object (relative to A) when A has the information of what B measured?
and
-For C, A and B are quantum objects until C gets the information from them. After C sees A and B, they (A and B) transform (relative to C) in classical object having one particular value?

Thanks

Hope I made some intelligible sentences