Exploring the Paradoxical Einstein-Podolsky-Rosen Experiment

[JesseM]

Take a look at the photon COUNT as it goes from figure 4.23 to figure 4.26. Goes way way way down per unit area doesn't it?

Perhaps just because the detector where those hits are being registered is getting farther away?

I will bet anybody here a steak dinner that we've all got it backwards. The interference pattern will ALWAYS show up without coincidence counting. When D2 is moved to the imaging plane, a _subset_ of photons winds up being detected which form two gaussian patterns corresponding to the known 'which-path" information.

A subset of an interference pattern can't look like two gaussians! After all, at the minima of an interference pattern no photons are being detected, but there should be photons there in the sum of the two gaussians.

 

[peter0302]

Sorry, you're right I meant to say "D1 is Fix". So we're looking at, Figs. 4.5, and 4.18-4.21. You're also right that there's no way you can pull 4.21 out of 4.18 (I thought, erroneously, you could pull 4.26 out of 4.23). Can I retract my steak dinner bet? :)

HOWEVER, what I haven't changed my mind on is the original point.

Of course it is--why do you think that contradicts my statement? I interpret it to mean they are looking at the subset of photons arriving at D2 for which their entangled twin went to that one fixed position that D1 is at (even though there'd be plenty of hits at D2 where there was no hit at D1, but there would have been a hit if D1 was replaced by a wider CCD).

The "subset of photons arriving at D2 for which their entangled twin went to that one fixed position that D1 is at" would be ALL of the photons that strike D2. That is why putting a CCD there instead of a narrow-band detector should not change the result.

But why do you think that proves D1 isn't narrow? Figures 4.7 and 4.8 seem to show that D1 needs to be moved if they want to build up the pattern of photons at that location while keeping D2 fixed, which wouldn't be necessary if D1 was already wide enough to pick up all the photons coming through the lens.

D1 is narrow, but the focal point of a lens is a POINT, so it doesn't matter how narrow D1 is. And, again, I am NOT talking about figures 4.7 and 4.8. Those are a different mode of the experiment.

You're confused, it is D1 which is behind the lens and which is moved from the focal plane to the imaging plane--l

Yes, I was citing the wrong figures. Look at figures 4.5, and 4.18-4.21. They all show results at D2 when D1 is fixed in the x direction but moved from the focal point to the imaging plane. This setup is the critical one which I contend does not require the coincidence circuit.

 

[peter0302]

And I'd also add that this is precisely what Cramer is trying to do...

 

[JesseM]

Sorry, you're right I meant to say "D1 is Fix". So we're looking at, Figs. 4.5, and 4.18-4.21. You're also right that there's no way you can pull 4.21 out of 4.18 (I thought, erroneously, you could pull 4.26 out of 4.23). Can I retract my steak dinner bet? :)

Steak dinner? What steak dinner? ;)

The "subset of photons arriving at D2 for which their entangled twin went to that one fixed position that D1 is at" would be ALL of the photons that strike D2. That is why putting a CCD there instead of a narrow-band detector should not change the result.

But do you agree that wouldn't be true if D1 was narrow? After all, even when D1 is on the imaging plane, you can see from the graph that photons can arrive at a number of positions (the two sharp peaks at different locations), so if you fix D1 at one position, there can be cases where a photon is registered at D2 but the corresponding photon misses D1, since it goes to a different position in D1's plane. You seem to be assuming that D1 is wide enough that it will catch all incoming photons, but the post by Ben says that's incorrect, and the fact that they show a double-headed blue arrow in D1's plane when the position of D2 is fixed in fig. 4.7 suggests it's incorrect as well (if D1 was already catching all incoming photons, what need would there be to move it around?)

D1 is narrow, but the focal point of a lens is a POINT, so it doesn't matter how narrow D1 is. And, again, I am NOT talking about figures 4.7 and 4.8. Those are a different mode of the experiment.

In what relevant way is it different? Do you deny that in figure 4.7, D1 is exactly at the focal distance, and when its position in the horizontal plane is varied it picks up photons at a range of locations? I don't see why we should expect all the light to be focused at a single point anyway--in classical optics a lens will only focus light perfectly at a point if all the light rays are coming in perfectly parallel, but in the quantum experiment there's some uncertainty in the momenta of the photons.

Yes, I was citing the wrong figures. Look at figures 4.5, and 4.18-4.21. They all show results at D2 when D1 is fixed in the x direction but moved from the focal point to the imaging plane. This setup is the critical one which I contend does not require the coincidence circuit.

I think you're wrong that in fig. 4.5, every hit at D2 would have a corresponding hit at D1. It's clear from fig. 4.7 that even when D1 is at the focal distance, photons can hit it at a range of horizontal positions.

 

[Ken G]

What I don't understand is, why is there so much buzz about experiments like this-- do they not always confirm the predictions of quantum mechanics? So we can focus on the predictions of the theory, and look at why they come out as they do in the theory itself, that's where the insights are-- we only need the experiments to tell us we can do that, and I'm pretty much good on that already, frankly. It's not as if we're all expecting quantum mechanics to fail when it "seems too bizarre to be right", but lo and behold, the experiments say it's right. Speaking for myself, I always expect quantum mechanics to be correct in every situation that it makes a prediction! So I hear of experiments like this and just say "yup, right again", and I'm done with it. Indeed, we could come up with more and more bizarre thought experiments with counterintuitive results, and if the experiment is actually realizable (unlike abominations like the cat paradox), then the problem is always going to be with our intuition, of course. What's the big deal?

 

[akhmeteli]

You call me nuts, and I thus dismiss your post -- except for this particularly offensive comment. It's clear that you are interested in physics, and know just enough to be dangerous.If you want to learn, ask questions -- like "what is an observation? Can non-humans make observations? Try to learn enough to read Dirac -- learn about QM in practice -- from the hydrogen atom to basic radiation theory. Then revisit interpretations -- and make sure you study the Peierls -(Wigner) approach-- both are Nobel prize winners ; they postulate that the wave function collapse occurs as the neural networks in the brain provide the single answer from a measurement;that is, the wave function describes your state of knowledge.

Certainly, abrasive language can be offensive and distract from the discussion. So I'll try to keep to the point. So what's your opinion on the question you mentioned: "Can non-humans make observations?" If, for example, an experimental setup is fully automatic, and the results are stored by the computer on a hard disk, do you think these results can change when a human reviews them?
By the way, as far as I know, Peierls was not a Nobel prize winner. Certainly, this does not make him a less respected physicist. However, I am not aware of any experimental confirmation of his postulate that "the wave function collapse occurs as the neural networks in the brain provide the single answer from a measurement". However big an authority on quantum mechanics Peierls may be, I don't think I have any moral obligation to agree with him, not because I disrespect such an authority, but for the simple reason that such people as Einstein, de Broglie, Schroedinger, and others disagreed with such thinking. As for the Born rule, I like article http://arxiv.org/abs/quant-ph/0702135 , where an analysis of an exactly solvable model of spin measurement shows that this rule may emerge from thermodynamic irreversibility.

 

[vanesch]

I am maintaining that we have not the least experimental justification to require that "the "axioms of quantum mechanics" must apply to macro systems!

True, but the question is actually this one: is it *thinkable* (can we find a vision, a picture, a consistent toy world) in which it can be done ? Or are we SURE now that quantum mechanics doesn't apply to macro systems ?

In other words, is there an experiment that proves without any doubt that quantum mechanics CANNOT describe macro objects, or is there still a possibility that it is ?

I think (I might historically be wrong, I'm no expert, I only know the common myths ) that Schroedinger's observation tried to show that *evidently* quantum mechanics is not applicable - gives rise to absurdities, wrong results - when applied to a cat. I think he was wrong, in that things are more subtle and that decoherence shows a way to get out consistent views, all respecting quantum mechanics.

So the question is not: did an experiment show that a macro object DID do something 'non-classical' and purely quantum mechanical, but rather, was there an experiment that FALSIFIED a prediction of quantum mechanics concerning macroscopic objects.

And in as much as the first has not been done with things like cats (and probably never will be - although I have my ideas about that), *it hasn't been shown either that quantum theory gives DIFFERENT results from what is observed.

In other words, *we don't know* in how much quantum mechanics "really applies" to macroscopic objects. It's an open question.

Now, you point out that perhaps we are not requiring this, we are choosing it-- but if that's true, why are we choosing this is if it is not required? Where do we benefit from this choice if it is not forced on us by nature?

Because nature would be SIMPLER if quantum mechanics was just universally valid! We would have a unique, universal set of principles. Now, I will hasten myself to add that this is probably dreaming out loud, because probably our current theories are still approximations to future theories, which will be approximations to even more future theories etc...

However, we DON'T KNOW what is the scope of quantum theory as of now. We don't know how universal it is. Gravity is in any case a pain. So this might be an indication of a fundamental problem. But for all we know, we cannot be sure that certain principles of quantum theory, in current or modified form, are NOT valid on macroscopic scales. We have no indication either way.

Right, I see that we are on the same page-- we are playing with "toy worlds" here, but the issue on the table is which one best describes the real world in a given situation. As I have never seen anyone meaningfully apply quantum mechanics to the state of a cat as a whole, I claim that is a clear case of using the wrong "toy world".

But what gives priority to classical physics ? What if quantum mechanics (as decoherence seems to show) REDUCES to observable effects which are identical to classical physics ? Why should we then say that classical physics is right and quantum physics is wrong ? Calculationally, I agree, classical physics is way easier to deal with. But why should classical physics have priority over quantum physics conceptually - which rises the problem of the transition between both ?

As it turns out, we do not suffer much from this problem. Nevertheless, it is a real problem, and your solution will not handle it any better than mine. Indeed, even the theory of large nuclei does not follow the approach you are suggesting! The first step always looks something like "well we can't really solve quantum mechanics for this system, so here's what we do instead".

Yes, but that is
1) for practical purposes
2) not a contradiction.

Indeed, we know that from the moment that the entangled states are complex enough, that probably no observation will give any interference effects, and that from that moment on, we will get IDENTICAL results between a semi-classical approach and a full quantum approach. As the former is practically much easier to handle than the latter, we prefer of course to do the former. This is what happens in much of quantum chemistry too. From the moment that explicit interference has become "unobservable" (that means, hidden in very high order correlation functions which are never observed), you can switch to a semi-classical approach with probability distributions.

But again, it is not a proof of the *unapplicability* of quantum mechanics as a principle. On the contrary. It is where quantum theory becomes identical to classical theory.

Certainly you can start somewhere, and see where it gets you, that's an excellent way to do science. But the question is, where does this get you, in regard to a cat, or in regard to wavefunction collapse? Are we trying to motivate actual new observations, or are we trying to satisfy ourselves that we in some sense understand the outcomes of impossible ones? What gets us somewhere is the mindset that says we are coupling quantum systems to macro systems expressly because we can rely on the macro system to act classically, which our brains like and we can actually call it a "measurement". What other kinds of experiments can we do? Given that, where is the gain for us in treating our macro system quantum mechanically, and why did we need a macro system involved in the first place if we were just going to treat it quantum mechanically?

On the practical, applied side, I agree with you. But the point is, if you insist on the inapplicability of quantum mechanics to macrosystems, then you are going to look for a *transition* theory. The "real" theory that will describe what happens when a system switches from quantum theory to classical theory (which are then "asymptotic" approximations to a more complete framework).
However, if it turns out that quantum mechanics IS actually valid "all the way up", then you will have excluded a whole scope of possibilities, and you will be looking for an entirely wrong theory.

In other words, you have excluded too soon a theory, that was not really falsified.

And there's another reason to play with a toy world in which to take your theory totally seriously (far beyond its proven domain of applicability): you get a good feeling for the machinery of the theory. You get a good understanding of what exactly the axioms imply - whether this corresponds to the real world or not.

 

[peter0302]

But do you agree that wouldn't be true if D1 was narrow?

No, I don't agree. When D1 is placed at the focal point, all photons incident normal to the focal plane pass through the focal point. So, the detector at the focal point should register _every_ photon that passes through that lens as long as the photons are sufficiently perpendicular to the lens, which the diagrams certainly imply, and which should certainly be possible to do as a practical matter.

After all, even when D1 is on the imaging plane, you can see from the graph that photons can arrive at a number of positions (the two sharp peaks at different locations), so if you fix D1 at one position, there can be cases where a photon is registered at D2 but the corresponding photon misses D1, since it goes to a different position in D1's plane.

That's absolutely true when D1 is at the imaging plane. That si why the coincidence circuit is indeed required to see the gaussian pattern. But that's not true when D1 is at the focal _point_.

You seem to be assuming that D1 is wide enough that it will catch all incoming photons, but the post by Ben says that's incorrect, and the fact that they show a double-headed blue arrow in D1's plane when the position of D2 is fixed in fig. 4.7 suggests it's incorrect as well (if D1 was already catching all incoming photons, what need would there be to move it around?)

In Figure 4.7, they also show photons at are clearly not collimated and normal to the lens. So a detector fixed at the focal point would not pick up every photon that passes through the lens. In Figure 4.5, by contrast, the photons are clearly shown to be collimated, normal to the lens, and all passing through the focal point.

In what relevant way is it different? Do you deny that in figure 4.7, D1 is exactly at the focal distance, and when its position in the horizontal plane is varied it picks up photons at a range of locations? I don't see why we should expect all the light to be focused at a single point anyway--in classical optics a lens will only focus light perfectly at a point if all the light rays are coming in perfectly parallel, but in the quantum experiment there's some uncertainty in the momenta of the photons.

So I think we're starting to hit the issue. Perhaps you're right that the setup in 4.5 and 4.7 is the same, except that in 4.5, where D1 is fixed at the focal point, they're only dealing with a specific subset of photons which happen to be normal to the focal plane. Perhaps they're not actually generating collimated beams of entangled photons. So you're saying in either case the coincidence circuit is required to pick out the subset of photons for which position information is utterly impossible to obtain, thereby generating the interference pattern behind the slits.

However, I still think this could be done without a coincidence circuit. Shouldn't there be a way to collimate both beams of photons without a coincidence circuit? Then collimating the beams would accomplish the same thing, forcing all of the photons incident on the lens to strike D1, have position information destroyed, and therefore have the photons detected by D2 exhibit interference.

Incidentally, the way Cramer's destroying the which-path information is the problem in his experiment. He's using half-silvered mirrors, I believe. These change the phase of he photons. Just like in DCQE, what he'll wind up with is two out-of-phase interference patterns that combine to a perfect gaussian pattern. So indeed, he will see a gaussian pattern always unless he uses a coincidence circuit. But I think using a Heisenberg lens like Dopfer doesn't present the same problem.

I think you're wrong that in fig. 4.5, every hit at D2 would have a corresponding hit at D1. It's clear from fig. 4.7 that even when D1 is at the focal distance, photons can hit it at a range of horizontal positions.

Right, unless the photons are normal to the focal plane. So that seems to be the issue then.

 

[guguma]

I wrote what I did only after some serious consideration. Physics is about describing nature. If you do your homework, you will find this idea goes back at least to the Greeks. Newton's Laws are computational recipes, just like the Schrodinger Eq. The only difference between the two is that Newton's ideas have been around for much longer
than the Schrodinger eq. . The consequence of that is that we have had several centuries to understand the descriptive power of Newton. So we have built a common intuitive consensus that we understand Newton, which is a huge difference from before Newton.

What's wrong with computational recipes?
Ask yourself exactly how it is you understand Newton?

At least some of the time people go on and on about something that was settled 20 years ago -- virtual particles are a good example. Spend a little time checking out your thoughts about something -- do a Google. As a retired professor I say, as I did to me students, do your homework and more. Those that do learn more, and make informed posts, which generally elicits more good stuff. How much time, guguma, have you spent understanding the concept of measurement in QM -- yours is a view not commonly held? There's 80 years of history to consider.Not for a moment do I think that humans are necessary to keep the universe ticking, as you put it. Perhaps a drunk could consider a bottle of scotch in anthropormorphic terms, and I suspect that only folks going to AA might concur.

Is a car wreck a measurement? Could the act of peeling a banana be an anti-measurement? Could waves crashing on a beach be a measurement? I keep asking this question, but I have yet to get an answer.

Consciousness has big effects on many things. This forum would not exist without consciousness. I'm sure that you can think of other things. And, when was the last time you saw bottles of scotch playing tennis, or going to school?

I'm literally dying to know how my sofa makes measurements, how do my pots and pans, stored so that they are in contact, make measurements, how does my hair make measurements, how does the sun make measurements?

Then there's a second round of questions: what do these various things measure, how do they do it, and how can I know?. I've asked this question also, without any answers. I'm in hope that my questions will be answered, not ignored.
Regards,
Reilly Atkinson

Prof. Atkinson,

First of all I want to state that I am not taking an offensive side against you, nor trying to mock you or something. It is easy to look like saying offensive things on a forum, but I can tell you that I am not. I am just trying to express my opinion on the matter, and I am very well aware that I have to do much (and much and much) homework, thus I am not questioning your experience it must be vast compared to mine. I will try to express my whole opinion on this issue now, including consciousness, and what I only ask for you is to read it sincerely and discuss it with me rather than questioning my experience, because the best way I can do my homework is to discuss with people who have more experience on the matter and if my arguments are outright falsifiable that is wonderful I will be happy to be falsified so that I can take one step further and learn to think otherwise.

First of all I want to state my opinion on human consciousness:

I think consciousness is far too overrated in any area of academics, and the only academic endeavor I have come across which does not overrate human consciousness is biological sciences (especially genetics).

If we look into a human being as a whole, humans are no different than an input-output mechanisms. We take a certain input and respond with an output to it. This mechanism is evolved through a process of evolution and natural selection, slowly and step by step and in the end we have this very complex neural network (which both of us should agree is where we think our consciousness comes from) and a bunch of other networks responsible for the continuation of this neural networks functions (simply put keeping it alive). This neural network is especially very efficient in depth and object recognition, it wonderfully organizes a serious amount of EM wave input and separates between different objects. We can separate between two objects on top of each other by color, shape, size etc. We should also agree that this neural network should have a memory like structure to be able to do this analysis, I am very well aware that there is no concrete scientific explanation as to how memory is maintained but certain disorders show that people are able to lose this memory structure completely, or partially. We can deduce that if this neural network is given a certain input the first thing it does is to compare it to other inputs in its memory terminal and concludes its output based on this comparison. So when a curved and closed object especially with its outer boundary is with a different color and a constant radius is recognized by our eyes we compare it to our previous input and group it into the circles, balls, spheres section of our memory. Every other function of this mechanism fears, love, hate, happiness, joy, AND thinking and consciousness are pretty much the same thing. If nothing in the universe moved, I mean nothing, including ourselves and every tiny bit of our composition too (just imagine it), would we have a consciousness of time. I do not think so because time is actually a ratio of motions and we are standardizing it to a particular reference object to make things easier. So I did not find it very surprising (it is not that it came to my mind or anything) when I learned in special relativity that space and time are not inseparable and our calculations are in need to be fixed especially for fast moving objects.

Does a naturally blind person have any consciousness of color. Yes this person can see colorized objects in his dreams (this is proven I guess) but which color is which? How does his neural network paints images? and what sort of images does this person see? Of course he can see certain images and paint them but it will not be based on the input from the environment thus this person will have a different consciousness than a person who sees and who can associate and categorize his input.

Lets compare this to a cpu, todays cpus are terrible at image and depth recognition, but they are wonderful in calculations. A cpu is designed by humans, and humans are designed by nature thus both of them are designed by nature (through physical processes I do not at all mean "Intelligent Design"). So a cpu takes an input and processes it according to its hardcoded or softcoded functions and gives an output. A CPU is awesome in one function and a Human is awesome in another, actually humans are so awesome that they do not even recognize that they are input-output machines and start thinking that they have this "consciousness" which is so different than a neural network + memory + functions.

There is even a deeper natural selection in the universe than genetic natural selection, which is on the particle (field, string, whatever) scale. Why less antiparticles? the exact solution is unknown but the general solution is that there is a physical symmetry break in the nature which favors particle structure over the antiparticle structure. why H atoms are abundant but not Uranium, again physical process, why certain molecules combine while others not? physical process. Why a certain gene (or rna, or dna or bacteria) survives in the gene pool? physical processes, just like antiparticle, particle selection, nature has a tendency to protect these stable genes this goes on a while and we get a human being which is extremely complex but still composed of particles, it only has a certain stability associated with it and an input-output mechanism which makes this stability continue. Consiousness is a complex process of memorial + input-output functions but still it is an input-output process, there is nothing special to it.

Now let's go to QM:

I think we both agree that what QM told us (actually reminded us) is that "We cannot talk about or predict the behavior of a single physical process because you have to have a certain input to talk about a process. You see a moving ball and what you talk about is the moving ball and what input you got. you shoot photons through a double slit and you talk about the photons + the double slit + screen, or photons + the double slit + detectors behind the slits + the screen and it is no surprise that two of them has different outputs"

So it is the detector that changes the output thus the input you got. If you were absent in these experiments but made a computer draw and print the interference images for you and when you would look at these two different papers would you say that "It must be the computers' consciousness which made this happen; if I have been present t would be different due to my consciousness"? I do not think so.THIS IS THE IMPORTANT PART

What I am advocating is that measurement is an interaction itself, whether it produces an input for you to interpret or not. We say measurement collapses the wave function but the measurement is the interaction between the detector and the detected, in some cases it may be your hair, or eyes or in some cases it is a Geiger counter and a photon. Or a bottle of scotch and sunlight.

You prepare a quantum system and say it can provide only two eigenvalues, when you do something to it it provides only one of them, what happened? You did not know how your system exactly was (and it is not incompleteness of QM it is just impossible for you to know because to know you have to measure it thus interact it with something else thus disturb it and see the disturbed output) you interacted it with something else and that interaction made it collapse telling you that after this interaction its energy is X.

Conclusion:

1. Consciousness is nothing special
2. One can only deduce conclusions about two interacted physical systems
3. No one knows the exact state of a prepared system, one can only provide a subset of possible outcomes which can come out after an interaction (eigenvalues), and every interaction disturbs the system thus you only talk about the interacted systems.

 

[JesseM]

No, I don't agree. When D1 is placed at the focal point, all photons incident normal to the focal plane pass through the focal point.

This would be true for light rays in classical optics, but I'm not so sure it would be true in QM--maybe if you had measured/filtered all the photons to make sure they had parallel momentum beforehand. But in this setup I don't think there was anything done to them to ensure that they'd have parallel momentum, and if you look at fig. 4.3 on page 30 (which seems to show a setup using a lens to measure non-entangled particles going through a double-slit, but the idea of what the lens is supposed to do seems pretty similar to the actual experiment), the rays going to the imaging plane (red lines) are actually coming in at very different angles; in the actual experiment, the subset of photons going to the upper detector D1 that had the right momentum so their twins went through the double slit and registered at D2 would have to have hit the upper lens at the same sort of angle as seen in fig. 4.3, I think. Therefore there's no reason to think these photons would be focused in the focal plane, they aren't coming in parallel like the blue lines in fig. 4.3.

By the way, the link to the paper again is http://www.quantum.univie.ac.at/publications/thesis/bddiss.pdf , if anyone is trying to follow this discussion but lost track.

So, the detector at the focal point should register _every_ photon that passes through that lens as long as the photons are sufficiently perpendicular to the lens, which the diagrams certainly imply, and which should certainly be possible to do as a practical matter.

I don't think the diagrams imply that at all; fig. 4.3 seems to imply something quite different about the point of what the lens is supposed to do.

After all, even when D1 is on the imaging plane, you can see from the graph that photons can arrive at a number of positions (the two sharp peaks at different locations), so if you fix D1 at one position, there can be cases where a photon is registered at D2 but the corresponding photon misses D1, since it goes to a different position in D1's plane.

That's absolutely true when D1 is at the imaging plane. That si why the coincidence circuit is indeed required to see the gaussian pattern. But that's not true when D1 is at the focal _point_.

Sorry, I meant to say "even when D1 is on the focal plane". And my mistake carried over to the graphs, the case where D1 is on the focal plane shows an interference pattern rather than two shark peaks (as depicted in the diagram of fig. 4.7 on page 38), so this shows that photons are arriving at a range of horizontal positions in this plane.

In Figure 4.7, they also show photons at are clearly not collimated and normal to the lens. So a detector fixed at the focal point would not pick up every photon that passes through the lens. In Figure 4.5, by contrast, the photons are clearly shown to be collimated, normal to the lens, and all passing through the focal point.

Ah, I see what you mean. But what is different in the setup that they're ensuring the photons are collimated? I would guess that the only way they ensure this is by placing D1 at the focal point and then ignoring all the hits at D2 that don't correspond to a hit at D1--coincidence counting, in other words. They didn't do anything different to the beam of photons beforehand to make sure they were all collimated, it's just that they only paid attention to the ones that ended up at the focal point, which they can retroactively say must have been coming in parallel to the plane. If I'm right about that, then this would mean you're wrong that all hits at D2 will also have a corresponding hit at D1--there'd be plenty of cases where D2 registered a hit, but it was thrown out because the twin didn't hit D1 due to it not coming in parallel to the focal plane.

So I think we're starting to hit the issue. Perhaps you're right that the setup in 4.5 and 4.7 is the same, except that in 4.5, where D1 is fixed at the focal point, they're only dealing with a specific subset of photons which happen to be normal to the focal plane. Perhaps they're not actually generating collimated beams of entangled photons. So you're saying in either case the coincidence circuit is required to pick out the subset of photons for which position information is utterly impossible to obtain, thereby generating the interference pattern behind the slits.

Yes! Although I wasn't saying that originally, but it's the conclusion I came to after seeing your above point about the photons shown coming in parallel in fig. 4.5, before reading the paragraph above...great minds think alike!

However, I still think this could be done without a coincidence circuit. Shouldn't there be a way to collimate both beams of photons without a coincidence circuit? Then collimating the beams would accomplish the same thing, forcing all of the photons incident on the lens to strike D1, have position information destroyed, and therefore have the photons detected by D2 exhibit interference.

I don't know if there'd be a way to collimate them except by some kind of filter which blocks photons that are coming in at the wrong angle--but blocking photons in one beam wouldn't block the corresponding photons in the other beam, so you'd still need coincidence-counting. I suppose if you filtered both beams, then since they are entangled by momentum, ideally any time one photon made it through the filter its entangled twin would as well? If that was possible then you'd have a point, in a setup like 4.5 it would seem like every time you had a hit at D2 you'd also get one at D1. But I'm not sure if there's any way to do this sort of filtering.

If it is possible, then it does seem like every hit at D2 should correspond to a hit at D1, and that therefore the total pattern of photons at D2 would show interference if the beam going to D2 is filtered in this way. I'm not sure if this is actually a problematic conclusion though, even if you then move D1 to the imaging plane and remove the filter from the beam at D1, I don't think there'd be any way to use the hits at D1 to determine the which-path info for the hits at D2--as you can see from the way the red curves representing the beam are depicted in fig. 4.8, and the red lines representing particles focused into two distinct points are shown in fig. 4.3, using the lens to determine the which-path info crucially depends on looking at a subset of photons that do not come in parallel to the plane of the lens. So if you take the subset of hits at D1 which also correspond to hits at D2, in the case where the D2 beam was filtered so the photons were collimated, I don't know if you'd still get those two distinct peaks in this subset at D1 which allow you to determine which slit the photons at D2 went through. Perhaps it's a position-momentum uncertainty issue--collimating the beam at D2 means you are confining them to a narrow range of momenta, which means the corresponding subset of hits at D1 must also be confined to the same narrow range, so that may destroy the possibility of measuring the photons at D1 in such a way as to gain precise position information about which slits the photons at D2 went through.

 

[JesseM]

Perhaps it's a position-momentum uncertainty issue--collimating the beam at D2 means you are confining them to a narrow range of momenta, which means the corresponding subset of hits at D1 must also be confined to the same narrow range, so that may destroy the possibility of measuring the photons at D1 in such a way as to gain precise position information about which slits the photons at D2 went through.

Incidentally, after doing a little searching on the subject, it seems the setup shown in fig. 4.3 of the thesis is known as a "Heisenberg microscope", and it's understood that such a lens can allow you to retroactively determine either the position or the momentum that incoming photons had prior to hitting the lens, depending on whether you measure them in the image plane or the focal plane--see p. 49-50 of the book , a thought-experiment involving such a microscope actually played an important role in the conceptual development of Heisenberg's uncertainty principle, helping to answer the question which introduces that article: "Are the uncertainty relations that Heisenberg discovered in 1927 just the result of the equations used, or are they really built into every measurement?"

 

[peter0302]

That could be the answer. A _collimated_ beam will _always_ produce an interference pattern because momentum will be (relatively) certain, so position will be uncertain.

A non-collimated, scattered beam, on the other hand, will be a totally random mix of position certain (two nice gaussian patterns) and momentum certain (interference pattern) and neither certain (blob inbetween) to create one, single gaussian pattern.

Have we resolved the question finally?

 

[JesseM]

That could be the answer. A _collimated_ beam will _always_ produce an interference pattern because momentum will be (relatively) certain, so position will be uncertain.

A non-collimated, scattered beam, on the other hand, will be a totally random mix of position certain (two nice gaussian patterns) and momentum certain (interference pattern) and neither certain (blob inbetween) to create one, single gaussian pattern.

Have we resolved the question finally?

It seems plausible that the total pattern of photons going through the double slit will form an interference pattern if the beam is collimated (since in this case we'd expect that if the upper detector D1 was placed at the focal point of the lens, all the photons in the upper beam corresponding to photons that made it through the collimation filter of the lower beam would have the right momentum to be focused onto the focal point), although I'm still not totally confident. The interesting case, and the one where I'm even less confident, is when the lower beam going through the double-slit and ending up at D2 is collimated, but the upper beam is not; if the total pattern of hits at D2 does show interference, what happens when you move the upper detector to the image plane at D1, and look at the subset of hits there that correspond to hits at D2? If there is indeed interference at D2, then it seems the corresponding hits at D1 can't show two distinct peaks without violating complementarity, but I don't have a good mental picture how the rays would avoid being focused into two distinct peaks at the image plane (maybe trying to think in terms of neat rays like in classical optics is not a good idea here).

Maybe the issue is that by collimating the lower beam, you are in effect measuring the momentum of all the photons that continue on to the double slit, and due to the position/momentum uncertainty relation this destroys the position entanglement of these photons with the photons on the upper beam; so in this case you can no longer be sure that if a photon on the lower beam went through a slit, the corresponding photon on the upper beam also went through one of the positions corresponding to a slit, and thus you could also no longer be sure these upper photons would be focused by the lens onto one of two spots on the image plane. I don't know what the pattern of this subset of photons on the upper beam would look like with the detector in the image plane--maybe it would just look identical to whatever the total pattern of photons on D1 is in the normal version of the experiment without collination shown in fig. 4.8 (in the 'normal' version only the subset of hits at D1 corresponding to detections at D2 looks like two discontinuous peaks, the total pattern of hits at D1 would presumably look different, perhaps a gaussian).

 

[peter0302]

I wish I had access to a source of entangled photons to try these things out. :) More importantly I wish Cramer would get on the ball and tell us how his experiment failed (which I'm sure it did).

 

[Ken G]

So what's your opinion on the question you mentioned: "Can non-humans make observations?" If, for example, an experimental setup is fully automatic, and the results are stored by the computer on a hard disk, do you think these results can change when a human reviews them?

I know your question is to reilly, but I can give you an answer that I'll bet is close to his as well, because it just involves keeping track of what is actually happening. First of all, the "experimental setup" you refer to did not spring up spontaneously, it was put together for a purpose. That purpose is not of incidental connection to the way we do science, it is the way we do science, so is integrally related to the equations that we used science to establish. Even the "results" you mention are conditioned to be results by us, the universe gets its "results" all the time, it needs no "experimental setup". So why did you attach your computer to such a setup? A computer that is hooked up to a random noise source is also getting "results" from the universe's point of view, if you will.

The results will not "change" when a human reviews them, because a "change" is a comparison between two things, and here we only have one. I do not think reilly is saying that the role of an intelligence is to change anything, but rather to have the one thing in the first place, whatever it is that we have that makes it something we would call a "result" and include it in our conception of reality.

However, I am not aware of any experimental confirmation of his postulate that "the wave function collapse occurs as the neural networks in the brain provide the single answer from a measurement".

My interpretation of this remark is that the "collapse" being referred to is not just the destruction of the superposition state of the subsystem, which is a physical effect that occurs any time our way of conceiving the state of the subsystem chooses to "average over" interactions with noise from a larger system we are not analyzing, but it also includes the determination of which result "actually occured". It is only a conscious mind that requires that final step, an "unlooked at" universe is perfectly content to function forever as an accumulation of mixed states, like dice that are rolled but never looked at. No experiment can tell the difference, until that experiment also involves the connection to an intelligence. Still, in my view, the quantum mechanics is over before this final stage of collapse is completed, it's a perfectly classical step. Indeed, the classical nature of this step is the whole reason for including it in our science-- the final stage of all science is classical, it's in the guts of science.

 

[Ken G]

True, but the question is actually this one: is it *thinkable* (can we find a vision, a picture, a consistent toy world) in which it can be done ? Or are we SURE now that quantum mechanics doesn't apply to macro systems ?

This must be the divergence in our views right here. You are essentially taking an "assumed true until proven false" approach to building toy worlds that are intended to mimic the real one, whereas I take an "assumed false until proved true" approach. I would say we have many examples in the history of physics where my approach would have saved some embarassment, and a lot of philosophical hand-wringing as well. On the other hand, your approach has led to new physics like neutrinos and positrons. So I think this shows the advantages of each-- when looking for new physics, by all means go ahead and assume your axioms extend across the frontier of what is known. But when building a philosophical world view, do the opposite, or you fall victim to the very same type of mysticism that science was essentially invented to replace.

The issue is in what is testable. It was "harmless" to postulate neutrinos, positrons, and now supersymmetry, based on "good" axioms to date. But concepts that you know are not testable by their very nature, like the wave function of a cat, lead you down the primrose path. Knowing going in that you can never test these notions, to hold them as true anyway makes one guilty of not looking for new science, but rather looking for that "warm fuzzy feeling" that is recognized as the illusion of control and understanding when the truth is pure mystery. The scientifically honest thing to do is recognize mystery when we encounter it, and not pretend that an approach that yields testable results in one area can be extended to an untestable realm with the pure intention of extinguishing that sense of mystery. So I say, if a cat has a wave function, prove it-- make a testable prediction. Failing that, the requirement is to say "I have no idea if the concept of wave function has any connection with a cat, and I will not build a philosophy around the idea that it does, simply to assuage my sense of order". There are more accessible ways to assuage our desire for order that do not even require an education in physics.

I think (I might historically be wrong, I'm no expert, I only know the common myths ) that Schroedinger's observation tried to show that *evidently* quantum mechanics is not applicable - gives rise to absurdities, wrong results - when applied to a cat. I think he was wrong, in that things are more subtle and that decoherence shows a way to get out consistent views, all respecting quantum mechanics.

I'm not sure why he did it either, but it's my sense that he was trying to expose flaws in the Copenhagen interpretation by using it to argue to an absurd result, the result that a cat could be in a superposition state. In other words, he took it as given that a cat could not-- which is why it is so ironic that his "paradox" is often expressed as saying that quantum mechanics shows a cat can be in a superposition state! It sounds like you and I can agree the paradox is irrelevant, because the superposition state (whether it can exist or not) cannot be created that way, due to the problem of decoherence. But what you are saying is that if a closed system containing a cat starts out in a pure state, quantum mechanics says it will remain in a pure state. I'm not denying that, I don't need quantum mechanics to be wrong-- I'm saying that even if you can get a closed system including a cat to be in a pure state (and I don't say you can), the cat, as a subsystem, will not be in a pure state. When you project a system onto a subsystem, you lose the pure state unless you can track all the coherences that connect the subsystem to the larger system-- and the fact that you can't do that is exactly what makes a cat a classical system.

So my bottom line is, a cat is a classical system, and the reason we couple quantum systems to classical systems is that we know we can count on the classical systems to respond classically. The logic of the cat paradox is exactly backward-- we should be asking how the quantum system got turned into a mixed state by its interaction with the cat, not how the cat got turned into a superposition by the quantum system.

So the question is not: did an experiment show that a macro object DID do something 'non-classical' and purely quantum mechanical, but rather, was there an experiment that FALSIFIED a prediction of quantum mechanics concerning macroscopic objects.

That's just the "correspondence principle" requirement that quantum mechanics is already held to. It doesn't show that quantum mechanics works on classical systems, only that it doesn't demonstrably fail on classical systems. I would say this means that quantum mechanics is "not even wrong" when applied to macro systems-- it simply isn't usable.

In other words, *we don't know* in how much quantum mechanics "really applies" to macroscopic objects. It's an open question.

It is only an open question by virtue of being untestable. That's not a strength of a scientific theory.

Because nature would be SIMPLER if quantum mechanics was just universally valid!

I don't agree there, and I'll express my disagreement with an analogy. Imagine you are an ornithologist studying the migration of the birds from some remote island. There are two species of birds on the island, and every Winter they disappear, and return in the Spring. You use radio tracking devices to track one of the species, but you find the other species rejects the trackers and pecks them off every time. So you track the one species, and see where they go. Now, does Occam's razor say it is simpler to assume the other species does the same thing, or does it say the simplest result is simply to not ask the question where the other ones go because it would be a pointless question to ask? I say the latter, if a question cannot be answered, the simplest thing is not to ask it-- not to assume the answer is something that cannot be falsified.

But for all we know, we cannot be sure that certain principles of quantum theory, in current or modified form, are NOT valid on macroscopic scales. We have no indication either way.

True-- but we also expect that we never will. That's the problem-- such axioms are only helpful if they lead to testable new physics. If they don't, they become philosophical baggage that the "razor" should trim away.

But what gives priority to classical physics ?

The tutor of our brains does that. Classical physics defines the guts of science. If you look at the structure of quantum physics, you see that it is designed as a theory to reduce quantum behavior in a predictable way into classical behavior. That's why we "measure" quantum systems, rather than just leaving them alone. Classical physics, on the other hand, is not a description of how classical systems can be made to act quantum mechanically. So it is we who give the priority to classical physics.

What if quantum mechanics (as decoherence seems to show) REDUCES to observable effects which are identical to classical physics ?

Decoherence is cherry-picked from all the things that can happen physically to a quantum system, and it is picked expressly because it is the subset of actions that leads to classical behavior. We choose that, we focus on decoherence, and set up our experiments to achieve it-- all to get the unknown to behave like the known, all to get a quantum system to leave a footprint on a classical one-- the latter being what we can use science on. So quantum mechanics doesn't "reduce" to classical mechanics. we project it onto classical physics on purpose, and formulate all our equations to describe the result of that projection. So classical behavior was always built into what quantum mechanics is, right from the start. There is no such thing as quantum mechanics without classical physics, that's what operators are. As a purely formal theory, a mathematician would likely say that quantum mechanics is just one arbitrary mathematical structure, and a fairly trivial one at that.

I may have said this before, but I think this is really the crucial point. There is not a physical place where quantum physics gives way to classical physics, we decide where that transition occurs when we change our approach to tackling a problem. The transition occurs the moment we feel compelled to average over some aspects of the state of the system that we do not wish to track explicitly. We know from experience that we can do that with our measuring devices, so that's why we feel comfortable coupling them to quantum systems to learn about the latter. So quantum mechanics cannot "reduce" to classical physics, because the averaging process goes outside the quantum system, it is a super-theory if you will, not part of the unitary transformations of quantum mechanics. This is precisely why, in my opinion, wavefunction "collapse" causes such hand-wringing within the confines of quantum mechanics-- it is expressly a process that leaves those confines. We set it up to do that, and then somehow forgot we did it, like a detective mistaking his own fingerprints at the scene of a crime.

Calculationally, I agree, classical physics is way easier to deal with. But why should classical physics have priority over quantum physics conceptually - which rises the problem of the transition between both ?

It's not just ease, it's the entire structure of scientific thinking. It was all built by classical brains-- electrons might have a very different approach to science.

Indeed, we know that from the moment that the entangled states are complex enough, that probably no observation will give any interference effects, and that from that moment on, we will get IDENTICAL results between a semi-classical approach and a full quantum approach.

But there is no full quantum approach at this state-- the instant you decide to average over what you can't know, you are not doing quantum mechanics any more (in the formal sense of the mathematical structure of the unitary operators, etc.). That's my point, the quantum mechanics becomes classical when we say it does, when we lose patience with following its axioms and resort to a semi-classical picture. If we always do that before we come to macro systems (and it seems to me that's true), then we simply have no quantum mechanics to test at the macro level, and cannot be impressed it hasn't been falsified.

From the moment that explicit interference has become "unobservable" (that means, hidden in very high order correlation functions which are never observed), you can switch to a semi-classical approach with probability distributions.

This is the crucial point we agree on-- but my interpretation of this is that it proves why quantum mechanics doesn't work for macro systems. To "work" doesn't just mean "doesn't make wrong predictions", it has to mean "is useful".

But again, it is not a proof of the *unapplicability* of quantum mechanics as a principle. On the contrary. It is where quantum theory becomes identical to classical theory.

Does it retain its axiomatic structure there? I don't think so, it seems to me it has to lose its soul, and become a mechanized simulation of that very classical theory it is becoming identical to. The kind of reduction you refer to happens when we add mass-energy to a particle by accelerating it until it behaves as though it had a trajectory, but that's different from what I'm talking about-- I'm talking about adding mass to the particle in the form of lots of other particles, like a baseball, and then treating its trajectory. That's a very different animal, for a quantum mechanical treatment that could make correct predictions in some situations would be wrong in others, since a baseball is not a quantum.

But the point is, if you insist on the inapplicability of quantum mechanics to macrosystems, then you are going to look for a *transition* theory.

Exactly, that is a good way to establish my point-- we would indeed require a transition theory, and I claim we do require a transition theory-- a theory in the realm where you are unable to use quantum mechanics for practical reasons, but the classical treatment of stochastically averaging over the unknowns fails to achieve sufficient accuracy. I maintain that we have a "blind spot" in our science of real systems because we can't treat that domain, but it rarely comes up.

And there's another reason to play with a toy world in which to take your theory totally seriously (far beyond its proven domain of applicability): you get a good feeling for the machinery of the theory. You get a good understanding of what exactly the axioms imply - whether this corresponds to the real world or not.

That I have no objection to-- if anyone can start their analysis with "the following is not intended to be taken seriously as a macroscopic theory, it is merely a macroscopic analog used to better picture our quantum axioms" then I'm fine. I've seen some use the Shrodinger cat that way. But inevitably, people mistake the analogy for the "real thing", and that opens the philosophical floodgates.

 

[peter0302]

The interesting case, and the one where I'm even less confident, is when the lower beam going through the double-slit and ending up at D2 is collimated, but the upper beam is not; if the total pattern of hits at D2 does show interference, what happens when you move the upper detector to the image plane at D1, and look at the subset of hits there that correspond to hits at D2? If there is indeed interference at D2, then it seems the corresponding hits at D1 can't show two distinct peaks without violating complementarity, but I don't have a good mental picture how the rays would avoid being focused into two distinct peaks at the image plane (maybe trying to think in terms of neat rays like in classical optics is not a good idea here).

If you look at the subset of hits at D1, when those photons are not collimated, that correspond to hits at D2, when photons are collimated, what you should get is exactly the same result as the Afshar experiment. There should be an interference pattern at D2, but peaks at D1. However, then the debate will be, as it is in the Afshar experiment, whether we can be certain that we know which photons at D1 corresponded to which slit.

Maybe the issue is that by collimating the lower beam, you are in effect measuring the momentum of all the photons that continue on to the double slit, and due to the position/momentum uncertainty relation this destroys the position entanglement of these photons with the photons on the upper beam; so in this case you can no longer be sure that if a photon on the lower beam went through a slit, the corresponding photon on the upper beam also went through one of the positions corresponding to a slit, and thus you could also no longer be sure these upper photons would be focused by the lens onto one of two spots on the image plane. I don't know what the pattern of this subset of photons on the upper beam would look like with the detector in the image plane--maybe it would just look identical to whatever the total pattern of photons on D1 is in the normal version of the experiment without collination shown in fig. 4.8 (in the 'normal' version only the subset of hits at D1 corresponding to detections at D2 looks like two discontinuous peaks, the total pattern of hits at D1 would presumably look different, perhaps a gaussian).

I think that's exactly right. The HUP is pretty hard to defeat!

 

[akhmeteli]

I know your question is to reilly, but I can give you an answer that I'll bet is close to his as well, because it just involves keeping track of what is actually happening.

I do appreciate your answer.

First of all, the "experimental setup" you refer to did not spring up spontaneously, it was put together for a purpose. That purpose is not of incidental connection to the way we do science, it is the way we do science, so is integrally related to the equations that we used science to establish. Even the "results" you mention are conditioned to be results by us, the universe gets its "results" all the time, it needs no "experimental setup". So why did you attach your computer to such a setup? A computer that is hooked up to a random noise source is also getting "results" from the universe's point of view, if you will.

No doubt, the experimental setup is prepared by humans. However, even if all humans die out after that, the setup can still work for some time. Depending on the way the results are stored, they can be perceived by animals or by intelligent life emerging a million years after that. I don't think this example is too far-fetched: indeed, scientists do study remains of dinosaurs that died millions of years ago. So do you really think those remains were much different before humans bothered to look at them?

The results will not "change" when a human reviews them, because a "change" is a comparison between two things, and here we only have one.

I am not sure about that, unless you insist that "the Moon is not there when nobody's looking at it". I think the results are there, whether somebody looks at them or not. If, however, you do insist, your position is unshakeable, but I cannot agree with it, so further discussion would be more appropriate in a forum on philosophy. So I think we can reasonably talk about two things: the results stored by the computer before and after somebody reviews them.

I do not think reilly is saying that the role of an intelligence is to change anything, but rather to have the one thing in the first place, whatever it is that we have that makes it something we would call a "result" and include it in our conception of reality.My interpretation of this remark is that the "collapse" being referred to is not just the destruction of the superposition state of the subsystem, which is a physical effect that occurs any time our way of conceiving the state of the subsystem chooses to "average over" interactions with noise from a larger system we are not analyzing, but it also includes the determination of which result "actually occured".

Could you explain more clearly the meaning of the phrase "the determination of which result "actually occured"?

It is only a conscious mind that requires that final step, an "unlooked at" universe is perfectly content to function forever as an accumulation of mixed states, like dice that are rolled but never looked at.

I thought die rolling is to all intents and purposes a classical process, so if you ensure that the initial position and velocity are the same with a great accuracy for several tosses (I heard the position must be accurate to about one micron), the die will always fall on the same side.

No experiment can tell the difference, until that experiment also involves the connection to an intelligence.

I don't mind if an "experiment also involves the connection to an intelligence". However, I have not heard about any such experiments confirming the Peierls' postulate. On the other hand, if you mean that no experiment can confirm or falsify the postulate, then I am not sure such a postulate has anything to do with science

Still, in my view, the quantum mechanics is over before this final stage of collapse is completed, it's a perfectly classical step. Indeed, the classical nature of this step is the whole reason for including it in our science-- the final stage of all science is classical, it's in the guts of science.

No offence, but this sounds like a mantra. I don't see why I should agree with that. I believe that is what Bohr and Heisenberg taught (or preached :-) ), but equally great physicists did not buy it.

By the way, I'd like to mention the article http://arxiv.org/abs/quant-ph/0702135 again. It is extremely relevant, and its conclusions seem fascinating to me. It clearly suggests a connection between measurements and thermodynamical irreversibility. Therefore, for finite systems, the results of measurements are reversible, but this has no more practical importance than any other processes forbidden by the second law of thermodynamics. The conclusions of the article seem to suggest that we do not need conciousness-related mysticism to understand quantum measurements.

 

[Ken G]

No doubt, the experimental setup is prepared by humans. However, even if all humans die out after that, the setup can still work for some time. Depending on the way the results are stored, they can be perceived by animals or by intelligent life emerging a million years after that. I don't think this example is too far-fetched: indeed, scientists do study remains of dinosaurs that died millions of years ago. So do you really think those remains were much different before humans bothered to look at them?

That's very much the issue, and note this is a purely classical issue. It is sometimes classified as a quantum mechanical issue, but I agree with you that paleontologists face the exact same issue all the time, as do poker players. It has to do with something we take very much for granted but is really terribly subtle: probability theory. That is a model of how we treat what we don't know, and there's never any reason to think that once we know it, that information wasn't "there" all the time-- but there's also no reason to think that it was! Poker players understand this quite well-- if you don't call the hand, it makes no difference at all what the cards "really were", and no theory of reality requires that they be anything but an unactualized probability. The same for every dinosaur bone that is not dug up. You can choose to believe they are there, because it seems silly not to, but the real point is it makes no difference whatever what you believe, all testable aspects of a theory of reality function exactly the same either way. The two are indistinguishable, such is the nature of probability-- it is a theory about what you don't know as much as about what you do.

What brings this into better focus is when you ruminate on the question "what is the probability of X?" We tend to ask this question as if it had a definite meaning (what is the probability I will die tomorrow, what is the probability an electron shows up in a given place in a given trial, etc.), but it does not. The answer depends entirely on what information I put in, my "fingerprints" are all over the answer (ask the first question to an actuary to see that, or the second to a physicist studying entanglement). There usually is no absolute answer to that question, obviously, so why do we pretend that the probabilities that result from our scientific theories are any different? They aren't-- our fingerprints are all over those too. So that's the sense to which science does not exist without us-- we choose the parameters we are considering. All probabilities are a comparison, made by us-- we choose both the numerator and the denominator. That choice only ends when our brains perceive an answer-- when the cards are shown. And we use the results to test if the comparisons we were making were appropriately constrained by our science (or our poker strategy).

I am not sure about that, unless you insist that "the Moon is not there when nobody's looking at it".

It is not necessary to assume it goes away, it is merely necessary to point out that it makes no difference at all to science if it is there or not. No scientific theory can test that assertion, it is a question that cannot even be asked. This is indeed the problem with most of what I read about quantum entanglement, people endlessly debating what science tells us the answer to certain questions are, when what science is really telling us is we can't put those questions to science.

So it is with the Moon-- it is pure philosophy if the Moon is there or not. Now I realize that you might think it is more natural to assume the Moon is there, even if you don't know it, but that's not the point I'm making. I'm saying that if you have two things that can happen to the Moon, maybe it is or is not hit by an asteroid, then all science can ever tell you is your estimate of the chances that the Moon is still there. It makes no difference at all if it is or not, scientifically speaking, until it affects you in some way. You can't use it to test your prediction, you can't perceive the result, it can't make you a happy or miserable person, it just doesn't matter. That we think something "really happened", even though we don't know what, is just a handy picture for thinking about all this-- it cannot be tested in any way, so it isn't science. I would say it is scientifically in a "mixed state" and leave it at that, science has no more to say on the matter.

So I think we can reasonably talk about two things: the results stored by the computer before and after somebody reviews them.

What this means to me is, we will agree to enter into the "interpretation", or "picture", that the computer stores "real" results even if we do not know what they are. That is fine with me-- I use that picture of reality myself, as a matter of fact.

Could you explain more clearly the meaning of the phrase "the determination of which result "actually occured"?

The registering in an intelligence what happened. Isn't that what you would mean by that phrase too?

I thought die rolling is to all intents and purposes a classical process, so if you ensure that the initial position and velocity are the same with a great accuracy for several tosses (I heard the position must be accurate to about one micron), the die will always fall on the same side.

That was a bad assumption made by the post-Newtonians. In point of fact there was never any way to do that, long before quantum mechanics, for a "suitably shaken die". If the die is not suitably shaken, it is not functioning like a die. The point is, even classical systems involve probability concepts in their analysis-- always have and always will (consider the crucial role of "ergodicity" in thermodynamics, for example). Are we in a position to replace thermodynamics?

However, I have not heard about any such experiments confirming the Peierls' postulate. On the other hand, if you mean that no experiment can confirm or falsify the postulate, then I am not sure such a postulate has anything to do with science.

The point is, and perhaps reilly can corroborate, one does not seek an experiment to "falsify" Peierl's postulate, for the postulate is built right into how we do science. How will one set up an experiment to falsify that postulate, when the postulate is central to what we mean by an experiment? It is really an axiom, that is the point-- it is an inseparable part of science itself, and that's what it has to do with science.

No offence, but this sounds like a mantra. I don't see why I should agree with that. I believe that is what Bohr and Heisenberg taught (or preached :-) ), but equally great physicists did not buy it.

Then I would like to see you, or them, describe a means for doing science that does not include the "mantra": the final stage of all science is classical, it's in the guts of science. Will anyone please cite for me an example of an experimental result whose final stage was not classical? How can anyone claim this is something they "don't need to agree with"?

By the way, I'd like to mention the article http://arxiv.org/abs/quant-ph/0702135 again. It is extremely relevant, and its conclusions seem fascinating to me. It clearly suggests a connection between measurements and thermodynamical irreversibility.

I'll give it a look, but I expect it to provide complete verification of my position. You see, thermodynamics is the quintessential example of a classical theory of probability, where nothing is ever actualized beyond what the intelligence can discern! All thermodynamic concepts (temperature, pressure, etc.) are based on the idea that states never distinguished by any intelligence are to be treated as if they were indistinguishable elements of reality.

Therefore, for finite systems, the results of measurements are reversible, but this has no more practical importance than any other processes forbidden by the second law of thermodynamics. The conclusions of the article seem to suggest that we do not need conciousness-related mysticism to understand quantum measurements.

I completely agree with all of that-- the irreversibility comes from our analysis technique. The instant we "average over" what we cannot know, we obtain a probabilistic treatment, and probabilistic treatments are also quintessentially irreversible. None of that refutes the importance of consciousness in deciding "what counts as indistinguishable", i.e., what is the very meaning of "the probability of X".

 

[reilly]

Certainly, abrasive language can be offensive and distract from the discussion. So I'll try to keep to the point. So what's your opinion on the question you mentioned: "Can non-humans make observations?" If, for example, an experimental setup is fully automatic, and the results are stored by the computer on a hard disk, do you think these results can change when a human reviews them?
By the way, as far as I know, Peierls was not a Nobel prize winner. Certainly, this does not make him a less respected physicist. However, I am not aware of any experimental confirmation of his postulate that "the wave function collapse occurs as the neural networks in the brain provide the single answer from a measurement". However big an authority on quantum mechanics Peierls may be, I don't think I have any moral obligation to agree with him, not because I disrespect such an authority, but for the simple reason that such people as Einstein, de Broglie, Schroedinger, and others disagreed with such thinking. As for the Born rule, I like article http://arxiv.org/abs/quant-ph/0702135 , where an analysis of an exactly solvable model of spin measurement shows that this rule may emerge from thermodynamic irreversibility.

First, you are on the money; Sir Peierls did not win the Nobel Prize. As you suggest, he was a very key figure in the early days of modern QM.

Quite a few years ago, so-called artificial neural networks became one of the tools many of us used in market research and business statistics. Thus many started following the research in neurophysiology, which I did for ten-fifteen years or so. The notion that most of what happens in the brain is the result of pulses traveling through neural networks is a central tenant of the field -- this is elegantly discussed by Sir Francis Crick in his book The Astonishing Hypothesis: the Scientific Search for the Soul.

One day it struck me that that there is a physically understandable mechanism behind probability collapse. As in, right now, the best we can say is that one of three candidates will become the next US president, add a dark horse if you want. This knowledge is stored in your memory. Then, once the election is over, and you hear about it, your knowledge changes, and your brain has to do some readjusting. Among the things it will have to change is the probability structure of the election; that structure clearly can be said to collapse to from (p1,p2,p3,p4) to (0,1,0,0). In fact, it's pretty unlikely that you will consciously be aware of such a collapse, but there is no doubt that it happens.

You decide to do an interesting double slit experiment with photons or electrons. Randomly change the width of the slits; randomly change the distance between slits; use polaroid or mylar to slow down the particles, randomly with one or two slits as you wish. Do the experiment for a long time, and do the random thing however you want. You probably won't have a clue about the pattern you'll see on the detector screen. So the probability structure in your brain will very likely be (?). If you do not look at the screen until the experiment is done, and then wait ten minutes, before you open your eyes, you still have that (?) structure, and after you have ("pattern").

If you watched the screen for the entire experiment, your notions of the pattern will clearly converge stochastically to the final pattern. The probability structure -- not a great name here -- changes gradually, but still is consistent with collapse as a change in knowledge. I think that this knowledge approach makes repeated and continuous measurements -- every day vision for example -- easier to handle in QM.

So you can see that noway a human can change what's on a disc without 1. programming and executing some program or routine, or, 2. trashing the disc. Once you read the disc, you know.Once you read a mystery novel you know, at the end for sure, who donnit. When neither of us is participating in the forum, it's still there. The best game in town is to assume there is an objective reality. Seems, generally, to be a good working assumption.

I was delighted to discover Peierls' work on QM interpretation promoting the idea that the wave function and consequent probabilities refer to our knowledge. For me, at least, many issues I had with QM were solved with the knowledge interpretation. And let's be clear:My state and actions have, generally, little effect on the world, so whether my eyes are open or closed makes no difference to anyone or anything but me.

As far as I know, he did not discuss neural networks -- they were yet to become important when he was writing. Also, it's consistent with the standard statistical practice of many years and in many disciplines. There's collapse in any practical probabilistic system; once you know, things change in your head often as a consequence of what's outside your head.

Regards,
Reilly Atkinson

 

[akhmeteli]

That's very much the issue, and note this is a purely classical issue.

I guess I have no choice but to make the following conclusion from this statement: you were not really talking about any interpretation of quantum mechanics (as more or less everything you said was applicable to classical mechanics). Rather you were talking about interpretation of probabilities in general. As for me, the question that bothers me most is: "Did quantum mechanics radically change the notion of causality?"

So it is with the Moon-- it is pure philosophy if the Moon is there or not. Now I realize that you might think it is more natural to assume the Moon is there, even if you don't know it, but that's not the point I'm making. I'm saying that if you have two things that can happen to the Moon, maybe it is or is not hit by an asteroid, then all science can ever tell you is your estimate of the chances that the Moon is still there. It makes no difference at all if it is or not, scientifically speaking, until it affects you in some way. You can't use it to test your prediction, you can't perceive the result, it can't make you a happy or miserable person, it just doesn't matter. That we think something "really happened", even though we don't know what, is just a handy picture for thinking about all this-- it cannot be tested in any way, so it isn't science. I would say it is scientifically in a "mixed state" and leave it at that, science has no more to say on the matter.

What this means to me is, we will agree to enter into the "interpretation", or "picture", that the computer stores "real" results even if we do not know what they are. That is fine with me-- I use that picture of reality myself, as a matter of fact.

I fully agree: "it is pure philosophy if the Moon is there or not". You may give any answer to this question: "yes", "no", or "I prefer to sit on the fence". However hard I try, I cannot understand what your answer is. My guess is your answer is either "yes" or "I prefer to sit on the fence". However, I believe that if we are not able to agree on this philosophical question, we cannot agree on the "question in question": "Can non-humans make observations?" Because if you answer "yes" to the first question, I think "yes" to the second question is natural and logical. If you don't say "yes" to the first question, then we have philosophical differences, and I readily admit that I have no chance to convince you. And vice versa, you have no chance to convince me. Thus, it looks like the second question is purely philosophical as well.

The registering in an intelligence what happened. Isn't that what you would mean by that phrase too?"

I just did not understand the phrase at all. I suspected it could mean that a human makes a (to some extent arbitrary) decision on what the result is. That is how I understood the word "determine". Maybe this is just my problem with English, which is not my mother tongue.

That was a bad assumption made by the post-Newtonians. In point of fact there was never any way to do that, long before quantum mechanics, for a "suitably shaken die". If the die is not suitably shaken, it is not functioning like a die.

I don't think the assumption is bad. I think we just differ on definitions. I did not discuss a "suitably shaken die", I discussed a die as a well-known material object. Whether it fulfills the function of a die, I did not care. If you believe that it is not possible to predict a result of a throw for a die with accurately defined kinematic parameters (and known material properties of the die and of the surface), please advise.

The point is, even classical systems involve probability concepts in their analysis-- always have and always will (consider the crucial role of "ergodicity" in thermodynamics, for example). Are we in a position to replace thermodynamics?

As I wrote elsewhere in physicsforums, I have no problems with probabilities. I have problems with their status in the Kopenhagen interpretation. If you say that their status in quantum theory is the same as in classical physics (and it seems that you do say that), I have no problems with that. Again, I don't have problems with thermodynamics. What I want to emphasize is thermodynamics fulfills its function splendidly even if it is a superstructure upon classical microscopic dynamics, not quantum microscopic dynamics. While there is no irreversibility in classical mechanics (or quantum mechanics, by the way), there is practical irreversibility in classical statistical physics.

The point is, and perhaps reilly can corroborate, one does not seek an experiment to "falsify" Peierl's postulate, for the postulate is built right into how we do science. How will one set up an experiment to falsify that postulate, when the postulate is central to what we mean by an experiment? It is really an axiom, that is the point-- it is an inseparable part of science itself, and that's what it has to do with science.?

If this is an axiom, I am fully entitled to reject such an axiom. I do not agree, furthermore, I don't know why I should agree that it is an inseparable part of science itself. Indeed, if you think about it, you might agree that this "axiom" states pretty much the same as the statement "non-humans cannot make observations" (and I argued that this is a purely philosophical question). Indeed, if we assume the opposite, i.e. "non-humans can make observations" (like a computer storing results of an experiment), then the collapse of the wavefunction occurs prior to a review of the results by the human, i.e. outside of the human brain. Of course, you have every right to sell your philosophy, but I am under no obligation to buy.

Then I would like to see you, or them, describe a means for doing science that does not include the "mantra": the final stage of all science is classical, it's in the guts of science. Will anyone please cite for me an example of an experimental result whose final stage was not classical? How can anyone claim this is something they "don't need to agree with"?

I am not sure anybody can cite an example of an experimental result whose final stage is classical, for the simple reason that classical mechanics is wrong and quantum mechanics is right. Nothing can be precisely classical. For example, a voltmeter pointer has only approximate classical position. If you tell me that the position of a voltmeter pointer is not a "final stage", then we just return to the same question: "can non-humans make observations?", and we cannot agree due to philosophical differences. Actually, I do not agree with Bohr and Heisenberg that quantum mechanics requires classical mechanics for its interpretation, and again, this is not brazen irreverence, as I have no choice but to be irreverent either towards Bohr or towards Einstein. I choose to side with Einstein, you have every right to side with Bohr or anybody else.

I'll give it a look, but I expect it to provide complete verification of my position. You see, thermodynamics is the quintessential example of a classical theory of probability, where nothing is ever actualized beyond what the intelligence can discern! All thermodynamic concepts (temperature, pressure, etc.) are based on the idea that states never distinguished by any intelligence are to be treated as if they were indistinguishable elements of reality.
I completely agree with all of that-- the irreversibility comes from our analysis technique. The instant we "average over" what we cannot know, we obtain a probabilistic treatment, and probabilistic treatments are also quintessentially irreversible. None of that refutes the importance of consciousness in deciding "what counts as indistinguishable", i.e., what is the very meaning of "the probability of X".

Again, I have no problems with thermodynamics, probabilities, or Bayesian approach. I just think that they cannot be the last word, and at the most fundamental level nature is strictly causal. I am not trying to impose my beliefs on you, I'm just trying to explain that my point of view may be equally viable.

 

[akhmeteli]

The notion that most of what happens in the brain is the result of pulses traveling through neural networks is a central tenant of the field -- this is elegantly discussed by Sir Francis Crick in his book The Astonishing Hypothesis: the Scientific Search for the Soul.

I am not sure though that this is relevant to collapse of wavefunction. Furthermore, you seem to agree that a human cannot change what is stored by the computer (without reprogramming etc.) That seems to suggest that the collapse occurs without any human brain.

One day it struck me that that there is a physically understandable mechanism behind probability collapse. As in, right now, the best we can say is that one of three candidates will become the next US president, add a dark horse if you want. This knowledge is stored in your memory. Then, once the election is over, and you hear about it, your knowledge changes, and your brain has to do some readjusting. Among the things it will have to change is the probability structure of the election; that structure clearly can be said to collapse to from (p1,p2,p3,p4) to (0,1,0,0). In fact, it's pretty unlikely that you will consciously be aware of such a collapse, but there is no doubt that it happens.

You decide to do an interesting double slit experiment with photons or electrons. Randomly change the width of the slits; randomly change the distance between slits; use polaroid or mylar to slow down the particles, randomly with one or two slits as you wish. Do the experiment for a long time, and do the random thing however you want. You probably won't have a clue about the pattern you'll see on the detector screen. So the probability structure in your brain will very likely be (?). If you do not look at the screen until the experiment is done, and then wait ten minutes, before you open your eyes, you still have that (?) structure, and after you have ("pattern").

If you watched the screen for the entire experiment, your notions of the pattern will clearly converge stochastically to the final pattern. The probability structure -- not a great name here -- changes gradually, but still is consistent with collapse as a change in knowledge. I think that this knowledge approach makes repeated and continuous measurements -- every day vision for example -- easier to handle in QM.

So you can see that noway a human can change what's on a disc without 1. programming and executing some program or routine, or, 2. trashing the disc. Once you read the disc, you know.Once you read a mystery novel you know, at the end for sure, who donnit. When neither of us is participating in the forum, it's still there. The best game in town is to assume there is an objective reality. Seems, generally, to be a good working assumption.

I was delighted to discover Peierls' work on QM interpretation promoting the idea that the wave function and consequent probabilities refer to our knowledge. For me, at least, many issues I had with QM were solved with the knowledge interpretation. And let's be clear:My state and actions have, generally, little effect on the world, so whether my eyes are open or closed makes no difference to anyone or anything but me.

As far as I know, he did not discuss neural networks -- they were yet to become important when he was writing. Also, it's consistent with the standard statistical practice of many years and in many disciplines. There's collapse in any practical probabilistic system; once you know, things change in your head often as a consequence of what's outside your head.

You see, all you're saying can be said about a classical, not quantum system. Does this mean that the status of causality is the same in classical and quantum mechanics? That is, is there a causal structure underlying probabilities? If yes, I have no problem with that. If no, you need some arguments that are not equally applicable to classical mechanics.

 

[Ken G]

I guess I have no choice but to make the following conclusion from this statement: you were not really talking about any interpretation of quantum mechanics (as more or less everything you said was applicable to classical mechanics).

That's basically right-- but note the real implication is that what people call "interpretations of quantum mechanics" are really no such things, they are interpretations of the meaning of probability. To be a true interpretation of quantum mechanics, you must be interpreting the meaning of a pure state, but the way most people use the term, they are interpreting the meaning of a collapsed wave function, i.e., a substate of a pure state that also includes a dramatic and untraceable degree of decoherence. In short, a quasi-classical system!

As for me, the question that bothers me most is: "Did quantum mechanics radically change the notion of causality?"

I agree, this is a key question. I say, it did not, yet many people think that it did. I don't know why, I think it's because they imagine that pre-quantum physics was purely deterministic and did not involve probabilities. It is as though they have forgotten thermodynamics and weather forecasting.

I fully agree: "it is pure philosophy if the Moon is there or not". You may give any answer to this question: "yes", "no", or "I prefer to sit on the fence". However hard I try, I cannot understand what your answer is. My guess is your answer is either "yes" or "I prefer to sit on the fence".

My answer would be "I find it very useful to adopt the picture that at any time the Moon is either there or it isn't, but I recognize that no science requires this-- science only requires I can identify a probability that the Moon is there, a probability that may go completely unactualized until I look."

However, I believe that if we are not able to agree on this philosophical question, we cannot agree on the "question in question": "Can non-humans make observations?"

I think decoherence occurs naturally, but decoherence only yields probabilities, not actualities. So it depends on what you mean by "observation"-- most people mean the demonstration of an actualization, and that does require an intelligence, because that is where the actualization "lives". However, it is an important principle of physics that all the actualizations in these intelligent minds must be consistent. No one has any idea why this is, and there are certainly gray areas, but it does seem to hold well-- and it spawns the concept of "objective reality". But saying that actualizations must be consistent is not saying the actualizations don't require minds, it just says that-- the actualizations must be consistent when intelligence further actualizes the higher-order correlations.

Thus, it looks like the second question is purely philosophical as well.

The problem is with the definitions. We cannot say which statements are right or wrong until we can clearly define the words, and here the tricky words are "exist" (by which I mean "a probability that has been actualized") and "observation" (by which I mean, "the demonstration that actualization has occured", and expressly not "the decohering of a substate wave function to get it to behave classically", though that is certainly a key element of observation). I maintain that taking these definitions, an intelligence is required to have an observation, because there is no way to demonstrate that an "actualization" occured, and no way to distinguish it from a "probability", without one.

In other words, you can program a computer to generate a random number (pseudorandom is good enough for me), and it can send that result to another processor, which sends its result to another, and so forth. I would say the issue of at what point this constitutes an "intelligence" is exactly the point at which one can treat that original random number as "actualized". Prior to that point, you can still treat it as a probability distribution, and just propagate that probability distribution through all the subsequent processing, generating new probability distributions. At the point where you can say "treating this as a probability no longer gives useful predictions", then you have an intelligence. So it's not that "intelligence actualizes observations", it is "the actualization of observation is the definition of intelligence".

Whether that intelligence counts as "human" is an even murkier topic, that will probably have to wait until we have the first clue of how to differentiate "human intelligence" from other forms. I'm not sure we know what should qualify as human intelligence, any definition would seem to either leave out too many humans, or not leave out enough non-humans.

I just did not understand the phrase at all. I suspected it could mean that a human makes a (to some extent arbitrary) decision on what the result is. That is how I understood the word "determine". Maybe this is just my problem with English, which is not my mother tongue.

This certainly underscores the problem of definition. We probably need to do a lot more work around what is meant by these words, or we can have purely semantic differences disguised as real disagreements.

I don't think the assumption is bad. I think we just differ on definitions. I did not discuss a "suitably shaken die", I discussed a die as a well-known material object. Whether it fulfills the function of a die, I did not care. If you believe that it is not possible to predict a result of a throw for a die with accurately defined kinematic parameters (and known material properties of the die and of the surface), please advise.

I certainly believe it is not possible to predict the throw of a die no matter how well you prescribe its initial conditions, if the throw "mixes in" enough of the details of the environment. For example, you could specify the initial velocity and angular momentum of the die, but only to the precision of your instrument, and you have to propagate whatever uncertainty exists initially through a lot of exponentially magnifying factors. You can specify the amount of sound in the room, but that's not good enough-- you need to know the amplitude and phase of every vibration that could affect the die. You need to know not just the windspeed, but every eddy current in the air. You need to know not just the dimensions of the rolling area, and the material properties of the surface, but how it varies with position and whether or not "material properties" are defined suitably precisely in the first place. In short, the prediction is doing physics, whereas the "reality" is only the roll of the die itself. And to do physics, we make idealizations and approximations at every stage. At some point, we just throw up our hands and say "forget tracking these details, we're just going to average over what we don't know and make some kind of ergodicity assumption"-- and poof, we have no more than a probability, even in principle, at that moment. Since this is an inevitable component of Newtonian mechanics sooner or later in any complex system, we could never have said that the universe was deterministic, it would simply not be scientifically demonstrable long before quantum mechanics.

As I wrote elsewhere in physicsforums, I have no problems with probabilities. I have problems with their status in the Kopenhagen interpretation. If you say that their status in quantum theory is the same as in classical physics (and it seems that you do say that), I have no problems with that.

That is indeed what I would say-- what do you mean by their "status in the Kopenhagen interpretation?"

What I want to emphasize is thermodynamics fulfills its function splendidly even if it is a superstructure upon classical microscopic dynamics, not quantum microscopic dynamics.

Yes, it is a superstructure, and it is also an integral part of Newtonian physics. That's why the latter was never fully deterministic, it was more like asymptotically deterministic in a way that was purely philosophical. Did anyone think that Newton's laws turned thermodynamics into some kind of placekeeper until fully detailed predictions could be made?

While there is no irreversibility in classical mechanics (or quantum mechanics, by the way), there is practical irreversibility in classical statistical physics.

Excellent point, and crucial to understanding why determinism was always a red herring-- it was contradictory to the notion of irreversibility, which is a hugely important physical concept.

If this is an axiom, I am fully entitled to reject such an axiom.

You can claim to reject it, but if you actually apply it, then the claim is irrelevant. To actually reject the axiom, you have to find a way to do without it-- you have to find a way to do science that does not look like confronting an intelligence with experimental results that actualize the results and compare them to probabilistic predictions over a sequence of trials. I am completely at a loss as to how you imagine you can do science in some other way.

Indeed, if we assume the opposite, i.e. "non-humans can make observations" (like a computer storing results of an experiment), then the collapse of the wavefunction occurs prior to a review of the results by the human, i.e. outside of the human brain. Of course, you have every right to sell your philosophy, but I am under no obligation to buy.

But I'm not selling a philosophy, I'm challenging you to use your computer to do science without an intelligence! Remember, if the computer encounters a probability distribution of outcomes, then it generates a probability distribution of recordings and analysis. There is simply no way you can demonstrate it has done anything different, without invoking an intelligence to actualize the distinction. You can imagine that the result was actualized, and indeed we all do, but science is completely moot on the issue-- it doesn't need to take a stance and therefore should not (why would science take a stance on a matter it is moot about?). But science must take a stance when an intelligence actualizes the result, because that is a truth that must be contended with, indeed it is the very point of doing science to generate that truth. Science is an endeavor of an intelligence, surely that at least is noncontroversial.

I am not sure anybody can cite an example of an experimental result whose final stage is classical, for the simple reason that classical mechanics is wrong and quantum mechanics is right.

That is not what I mean by "classical"-- I simply mean a system we choose classical mechanics to describe (more definitions in need of clarification). It has nothing to do with what is "right" and "wrong", those are unsophisticated notions in human endeavors like physics-- it has to do with what we choose to do to solve a problem.

Nothing can be precisely classical.

Of course not-- what is precisely quantum mechanical? Nothing real is "precisely" anything.

I just think that they cannot be the last word, and at the most fundamental level nature is strictly causal. I am not trying to impose my beliefs on you, I'm just trying to explain that my point of view may be equally viable.

I cannot sway you from your beliefs, and don't want to-- but those exceed what science can tell us, and I am trying to keep careful track of where that line is.

 

[JesseM]

But I'm not selling a philosophy, I'm challenging you to use your computer to do science without an intelligence! Remember, if the computer encounters a probability distribution of outcomes, then it generates a probability distribution of recordings and analysis. There is simply no way you can demonstrate it has done anything different, without invoking an intelligence to actualize the distinction. You can imagine that the result was actualized, and indeed we all do, but science is completely moot on the issue-- it doesn't need to take a stance and therefore should not (why would science take a stance on a matter it is moot about?). But science must take a stance when an intelligence actualizes the result, because that is a truth that must be contended with, indeed it is the very point of doing science to generate that truth. Science is an endeavor of an intelligence, surely that at least is noncontroversial.

But by the same argument, why should you believe that any intelligence other than yourself can "actualize the result" in this way? From your point of view, you have no way to falsify the notion that "if someone other than me encounters a probability of outcomes, then they generate a probability distribution of thoughts and memories and analysis." If we are only concerned with laws of physics as a recipe for making predictions, not as things that can give us a model of how objective reality might look independent of our observations, then it seems to me that this solipsist point of view would be perfectly reasonable, perhaps more reasonable than the view that other intelligences can actualize results but computers can't.

 

[akhmeteli]

That's basically right-- but note the real implication is that what people call "interpretations of quantum mechanics" are really no such things, they are interpretations of the meaning of probability. To be a true interpretation of quantum mechanics, you must be interpreting the meaning of a pure state, but the way most people use the term, they are interpreting the meaning of a collapsed wave function, i.e., a substate of a pure state that also includes a dramatic and untraceable degree of decoherence. In short, a quasi-classical system!

I am not sure this is true for the Bohmian interpretation; I guess it makes an attempt to describe a pure state. I am not trying to decide here whether this interpretation is good or bad.

I think decoherence occurs naturally, but decoherence only yields probabilities, not actualities. So it depends on what you mean by "observation"-- most people mean the demonstration of an actualization, and that does require an intelligence, because that is where the actualization "lives". However, it is an important principle of physics that all the actualizations in these intelligent minds must be consistent. No one has any idea why this is, and there are certainly gray areas, but it does seem to hold well-- and it spawns the concept of "objective reality". But saying that actualizations must be consistent is not saying the actualizations don't require minds, it just says that-- the actualizations must be consistent when intelligence further actualizes the higher-order correlations.

Again, we either agree that objective reality exists, or we don't. If we don't, then we can only debate philosophical problems. My take is it would be unproductive and inappropriate for this forum.

The problem is with the definitions. We cannot say which statements are right or wrong until we can clearly define the words, and here the tricky words are "exist" (by which I mean "a probability that has been actualized") and "observation" (by which I mean, "the demonstration that actualization has occured", and expressly not "the decohering of a substate wave function to get it to behave classically", though that is certainly a key element of observation). I maintain that taking these definitions, an intelligence is required to have an observation, because there is no way to demonstrate that an "actualization" occured, and no way to distinguish it from a "probability", without one.

In other words, you can program a computer to generate a random number (pseudorandom is good enough for me), and it can send that result to another processor, which sends its result to another, and so forth. I would say the issue of at what point this constitutes an "intelligence" is exactly the point at which one can treat that original random number as "actualized". Prior to that point, you can still treat it as a probability distribution, and just propagate that probability distribution through all the subsequent processing, generating new probability distributions. At the point where you can say "treating this as a probability no longer gives useful predictions", then you have an intelligence. So it's not that "intelligence actualizes observations", it is "the actualization of observation is the definition of intelligence".

Whether that intelligence counts as "human" is an even murkier topic, that will probably have to wait until we have the first clue of how to differentiate "human intelligence" from other forms. I'm not sure we know what should qualify as human intelligence, any definition would seem to either leave out too many humans, or not leave out enough non-humans.

It is not very productive to debate definitions, but it seems to me you offer definitions that are far beyond the usual meaning of everyday English words, such as "exists". If, however, you insist on such definitions, there is little left to discuss, as the questions we consider turn into tautologies.

I certainly believe it is not possible to predict the throw of a die no matter how well you prescribe its initial conditions, if the throw "mixes in" enough of the details of the environment. For example, you could specify the initial velocity and angular momentum of the die, but only to the precision of your instrument, and you have to propagate whatever uncertainty exists initially through a lot of exponentially magnifying factors. You can specify the amount of sound in the room, but that's not good enough-- you need to know the amplitude and phase of every vibration that could affect the die. You need to know not just the windspeed, but every eddy current in the air. You need to know not just the dimensions of the rolling area, and the material properties of the surface, but how it varies with position and whether or not "material properties" are defined suitably precisely in the first place. In short, the prediction is doing physics, whereas the "reality" is only the roll of the die itself. And to do physics, we make idealizations and approximations at every stage. At some point, we just throw up our hands and say "forget tracking these details, we're just going to average over what we don't know and make some kind of ergodicity assumption"-- and poof, we have no more than a probability, even in principle, at that moment. Since this is an inevitable component of Newtonian mechanics sooner or later in any complex system, we could never have said that the universe was deterministic, it would simply not be scientifically demonstrable long before quantum mechanics.

Well, this seems to be a quantitative question. I think you'll agree that for a simpler system than a rotating die and for lesser times it is possible to accurately predict the result, and I'll agree that for more complex systems and larger times it can be impossible to predict the result. As for this specific system, let's agree to disagree.

That is indeed what I would say-- what do you mean by their "status in the Kopenhagen interpretation?"

As far as I understand, the Copenhagen interpretation (although, or maybe because I lived in that city for four years, I cannot write its name correctly:-) ) postulates that we cannot have anything more precise than probabilities. If in classical mechanics two similar experiments produce different results, you can standardize them and reduce the difference, or at least indicate the source of the difference (and this is how I understand determinism, not as a possibility to predict everything). Nothing of the kind in the Copenhagen interpretation.

Excellent point, and crucial to understanding why determinism was always a red herring-- it was contradictory to the notion of irreversibility, which is a hugely important physical concept.

I wrote above how I understand determinism. I agree that there is a contradiction, but I believe determinism is more fundamental than irreversibility, which is only a practical convenience: you cannot circumvent the Poincare recurrence theorem, but the recurrence times are mindbogglingly huge for large systems, let alone the environment effects.

You can claim to reject it, but if you actually apply it, then the claim is irrelevant. To actually reject the axiom, you have to find a way to do without it-- you have to find a way to do science that does not look like confronting an intelligence with experimental results that actualize the results and compare them to probabilistic predictions over a sequence of trials. I am completely at a loss as to how you imagine you can do science in some other way.

I am afraid you've lost me here. You called the Peierls postulate an axiom, I said I reject such an axiom. This postulate states that the wavefunction collapse occurs in the brain. Why do I need it to do science? The bulk of the science does not even need a notion of wavefunction, never mind the Peierls postulate.

But I'm not selling a philosophy, I'm challenging you to use your computer to do science without an intelligence! Remember, if the computer encounters a probability distribution of outcomes, then it generates a probability distribution of recordings and analysis. There is simply no way you can demonstrate it has done anything different, without invoking an intelligence to actualize the distinction. You can imagine that the result was actualized, and indeed we all do, but science is completely moot on the issue-- it doesn't need to take a stance and therefore should not (why would science take a stance on a matter it is moot about?). But science must take a stance when an intelligence actualizes the result, because that is a truth that must be contended with, indeed it is the very point of doing science to generate that truth. Science is an endeavor of an intelligence, surely that at least is noncontroversial.

I see no reasons to accept your challenge "to use your computer to do science without an intelligence" as I did not claim it is possible. I was saying that "observation" without intelligence may be possible, and therefore wavefunction collapse does not need to occur in a human brain.

That is not what I mean by "classical"-- I simply mean a system we choose classical mechanics to describe (more definitions in need of clarification). It has nothing to do with what is "right" and "wrong", those are unsophisticated notions in human endeavors like physics-- it has to do with what we choose to do to solve a problem..

As for what you choose to describe a system, you may choose classical mechanics, or you can choose quantum mechanics.

Of course not-- what is precisely quantum mechanical? Nothing real is "precisely" anything.

Of course, there are no absolutes, however quantum mechanics is more precise than classical mechanics.

I cannot sway you from your beliefs, and don't want to-- but those exceed what science can tell us, and I am trying to keep careful track of where that line is.

You see, there are a lot of things that exceed what science can tell us. For example, science cannot tell us that the Earth rotates around the Sun. You can use the Earth as the system of reference. The resulting dynamics will be more complex, but it will be still correct. Science cannot tell us that the speed of light is the same in all inertial systems of reference - the results of the special relativity coincide with those that Lorentz obtained using his contraction. The science does not tell us that there is objective reality. I prefer to accept some of these things. If you don't want to accept any such things, its your choice based on philosophy. I am not ready to discuss philosophy.

 

[Ken G]

But by the same argument, why should you believe that any intelligence other than yourself can "actualize the result" in this way? From your point of view, you have no way to falsify the notion that "if someone other than me encounters a probability of outcomes, then they generate a probability distribution of thoughts and memories and analysis."

That's quite correct, when I was talking about "actualizing probabilities by an intelligence", I meant "for that intelligence". There is no requirement that someone else's intelligence actualize my reality, you're right I could never demonstrate that nor would I even care to, my consciousness actualizes my reality. I merely avow based on the symmetry of the situation that if my conscious intelligence can actualize my reality for me, then so can yours for you. The only scientific constraint on it is that when an intelligence actualizes higher-order correlations (we "compare notes" on reality), the actualizations must be consistent, which allows us to imagine an objective picture of reality (to some extent). That's really amazing, but it says no more than it says-- the correlations between actualizations must be total to within our ability to test (with perhaps some transliteration prior to comparison, as with time dilation). Aspects that cannot be tested suffer from no such requirement of objectivity. This is all science needs, the rest is philosophy.

It seems correct that a scientific prediction is a mapping between a probability distribution of inputs onto a probality distribution of outputs. Our intelligence decides what counts as the possible array of inputs, and what counts as an array of outputs, and that determines the probabilities we obtain. We then use our intelligence to test that mapping, and none of this requires actualization. One could imagine a mind capable of saying "the individual trial yielded x heads and 1-x tails", and could average many trials to see if <x>=1/2. Our minds don't happen to interact with reality that way, we say "it's heads" or "tails"-- no one knows why, but science must deal with it, so that is why we have "interpretations of quantum mechanics", but they are really "interpretations of actualizations", and they are entirely classical (in the sense of pertaining solely to classical objects like measuring apparatus).

If we are only concerned with laws of physics as a recipe for making predictions, not as things that can give us a model of how objective reality might look independent of our observations, then it seems to me that this solipsist point of view would be perfectly reasonable, perhaps more reasonable than the view that other intelligences can actualize results but computers can't.

Exactly, and that is why it is indeed the scientific perspective. Adding a bunch of untestable pictures to that may appease our psychology, but it is an extraneous component of science (that we all like to do, certainly).

 

[Ken G]

I am not sure this is true for the Bohmian interpretation; I guess it makes an attempt to describe a pure state.

Yes, my comments are really focused on "MWI" type thinking.

Again, we either agree that objective reality exists, or we don't. If we don't, then we can only debate philosophical problems. My take is it would be unproductive and inappropriate for this forum.

I have no desire to debate whether or not objective reality "exists", the issue is how the concept is used in science. It is of course just that-- a concept, and one that science uses to great advantage, but the issue is, just what does science actually need, and what are we adding just to make ourselves more comfortable? The latter is what is inappropriate for the forum.

It is not very productive to debate definitions, but it seems to me you offer definitions that are far beyond the usual meaning of everyday English words, such as "exists".

One does not "debate" definitions, but the clarification of them is completely crucial. One can only criticize a scientific definition on the grounds that it does not conform to the axioms of science. I have said what I mean by my words, as relying on vague popular meanings is ineffectual.

If, however, you insist on such definitions, there is little left to discuss, as the questions we consider turn into tautologies.

Again I cannot concur, there is everything to discuss-- there is the discussion of what are the ramifications of these definitions! That is not "tautological". If you are using different definitions, I invite you to supply them, so we can consider the ramifications of your definitions-- I only require that they be expressed in a scientifically demonstrable way, as I believe I have done.

Well, this seems to be a quantitative question. I think you'll agree that for a simpler system than a rotating die and for lesser times it is possible to accurately predict the result, and I'll agree that for more complex systems and larger times it can be impossible to predict the result.

Yes, that suffices-- as long as classical mechanics is incomplete without probability theory, it may not be called a purely deterministic description of reality, which is all I'm saying. Quantum mechanics is not fundamentally new in that regard, contrary to how it gets advertised.

If in classical mechanics two similar experiments produce different results, you can standardize them and reduce the difference, or at least indicate the source of the difference (and this is how I understand determinism, not as a possibility to predict everything). Nothing of the kind in the Copenhagen interpretation.

But isn't that just a question of precision? Even a classical measurement of angular momentum was never know to be capable of being precise at the level of h. It was just a guess that it could, pure imagination-- and that's why I don't think of classical physics as establishing a deterministic reality, it had not been demonstrated to be such, even if we adopt your more careful definition of determinism (which one might call asymptotic determinism if you please it).

I agree that there is a contradiction, but I believe determinism is more fundamental than irreversibility, which is only a practical convenience: you cannot circumvent the Poincare recurrence theorem, but the recurrence times are mindbogglingly huge for large systems, let alone the environment effects.

To me, all of science is "a practical convenience", one cannot say irreversibility is and determinism isn't because one has no scientific prescription for demonstrating a difference between a convenience and a principle. Scientific theory is a bunch of convenient concepts, how could it be anything else?

I am afraid you've lost me here. You called the Peierls postulate an axiom, I said I reject such an axiom. This postulate states that the wavefunction collapse occurs in the brain. Why do I need it to do science? The bulk of the science does not even need a notion of wavefunction, never mind the Peierls postulate.

The axiom I was referring to was that the last step of science is classical because it involves a confrontation with our own brains, and our brains function classically. If we could match a superposition in our brain with the superposition of an electron, we would never need that final classical step, but we can't and we do. It's just how we do science, yourself included, so that's why I claim it is a required axiom-- for any scientist. I interpret Peierl's "collapse of the wavefunction" idea entirely in those terms, in the context of quantum mechanics.

I was saying that "observation" without intelligence may be possible, and therefore wavefunction collapse does not need to occur in a human brain.

Then please provide the scientific meaning you attach to the word "observation". Remember, I've been clear that the destruction of coherences is something I consider to be a natural process, but does not by itself constitute an observation, when the latter is defined as "a means to put a scientific theory to the test".

As for what you choose to describe a system, you may choose classical mechanics, or you can choose quantum mechanics.

Precisely, my point is that many people use those terms as if they were different types of reality, like reality at a different scale-- that's what you were doing when you claimed there was no such thing as a classical system. In fact these are just choices made by an intelligence to describe reality.

Of course, there are no absolutes, however quantum mechanics is more precise than classical mechanics.

Only when applied to simpler systems that can be prepared more precisely, yes. But the increased precision is a reflection of the system, not the theory.

I prefer to accept some of these things. If you don't want to accept any such things, its your choice based on philosophy. I am not ready to discuss philosophy.

Neither am I-- which is why none of my statements have espoused any philosophy. The discussion was never about philosophy, it was always about where science ends and philosophy begins.

 

[akhmeteli]

I have no desire to debate whether or not objective reality "exists", the issue is how the concept is used in science. It is of course just that-- a concept, and one that science uses to great advantage, but the issue is, just what does science actually need, and what are we adding just to make ourselves more comfortable? The latter is what is inappropriate for the forum.

Comfort is extremely important for physicists. Otherwise, as I said, there is no difference whether the Earth rotates around the Sun or it is the other way round. You praised the Occam's razor - this is also a matter of comfort.

One does not "debate" definitions, but the clarification of them is completely crucial. One can only criticize a scientific definition on the grounds that it does not conform to the axioms of science. I have said what I mean by my words, as relying on vague popular meanings is ineffectual.

We use English for discussion, so when definitions of everyday words defy their standard meaning, it makes life a bit difficult.

Again I cannot concur, there is everything to discuss-- there is the discussion of what are the ramifications of these definitions! That is not "tautological". If you are using different definitions, I invite you to supply them, so we can consider the ramifications of your definitions-- I only require that they be expressed in a scientifically demonstrable way, as I believe I have done.

Let us agree to disagree on this. I am really not ready to define every English word, sorry.

But isn't that just a question of precision? Even a classical measurement of angular momentum was never know to be capable of being precise at the level of h. It was just a guess that it could, pure imagination-- and that's why I don't think of classical physics as establishing a deterministic reality, it had not been demonstrated to be such, even if we adopt your more careful definition of determinism (which one might call asymptotic determinism if you please it).

Seems like we disagree on whether classical physics is deterministic. No hard feelings:-) You see, we are talking about different things. Whether classical mechanics is precise or not, it's a matter for experimental tests. But whether classical mechanics is deterministic or not, it's a matter of its structure. Let me give you an example. Whether a specific quantum field theory is renormalizable or not does not depend on whether it describes experiments well, it can be established without experiments, although the theory can be so complex that it may take years to establish it (and that's what happened with the Standard Model).

To me, all of science is "a practical convenience", one cannot say irreversibility is and determinism isn't because one has no scientific prescription for demonstrating a difference between a convenience and a principle. Scientific theory is a bunch of convenient concepts, how could it be anything else?).

So comfort is inappropriate, but convenience is OK? Sorry, just teasing:-) Again, we just disagree on whether classical mechanics is deterministic.

The axiom I was referring to was that the last step of science is classical because it involves a confrontation with our own brains, and our brains function classically. If we could match a superposition in our brain with the superposition of an electron, we would never need that final classical step, but we can't and we do. It's just how we do science, yourself included, so that's why I claim it is a required axiom-- for any scientist. I interpret Peierl's "collapse of the wavefunction" idea entirely in those terms, in the context of quantum mechanics.

It is not obvious for me that our brains function classically. Even if Bohr said they do, that's not enough for me, I'm awfully sorry. I am not even sure anybody really understands how brains function. Another disagreement, I am afraid, now on the axiom:-)

Then please provide the scientific meaning you attach to the word "observation". Remember, I've been clear that the destruction of coherences is something I consider to be a natural process, but does not by itself constitute an observation, when the latter is defined as "a means to put a scientific theory to the test".

Again, I cannot give you a definition of observation. Not only because I am too lazy (although this is certainly a reason:-) ), but also because we are discussing a primary notion, which is difficult to define by other, "more primary" notions. And again, your definition goes against the usual meaning of this English word. On the other hand, it looks like we agree here on the essence of the matter, which for me is: decoherence takes place outside of human brain. For me this also means that the results of measurements are independent of human brain. You, however, are not disposed to recognize objective reality and, consequently, that results of measurements are independent of human brain. This is a philosophical difference.

Precisely, my point is that many people use those terms as if they were different types of reality, like reality at a different scale-- that's what you were doing when you claimed there was no such thing as a classical system. In fact these are just choices made by an intelligence to describe reality.

I am not sure that was what I was doing as classical or quantum mechainc are not realities, they are descriptions of reality, I agree with your emphasis. But quantum description is more precise, that's why I said that there is no such thing as a classical system.

Only when applied to simpler systems that can be prepared more precisely, yes. But the increased precision is a reflection of the system, not the theory.

Sorry, I have to disagree with the first phrase. Semiconductors, superconductors, metals, black body radiation, you name it, are not "simpler systems", they need little preparation, but their classical description just sucks. As for the second phrase, quantum mechanics is more precise than classical mechanics for any system.

Neither am I-- which is why none of my statements have espoused any philosophy. The discussion was never about philosophy, it was always about where science ends and philosophy begins.

Sorry to disappoint you, but your statements are not free from philosophy (neither are mine). When you praise the Occam's razor - it's philosophy, when you declare that the Peierls postulate is an axiom that science cannot do without, that's philosophy. Actually, I suspect that our disagreements may be much less significant than they look, as it seems to me that a wavefunction for you is just a shorthand for probabilities, and I have nothing against Bayesian approach, where probabilities are pretty much subjective estimates.However, for me a wavefunction is something that obeys the equations of quantum mechanics, so I am not sure it's just a record of probabilities, and again we are just talking about different things.