On the operational refutation of Bell's theorem

[Heinera]

I think we have somehow lost sight of the operational definition of Bell's theorem in the latest threads. Even those of you who think that the proof of Bell's theorem is wrong can't rule out the possibility that his conclusion is correct, can you? I mean, I could come up with a horribly wrong proof of Fermat's last theorem, but in some sense I would still have gotten to the right conclusion, wouldn't I? (since we now have Andrew Wiles' correct proof).

So, the only way to prove Bell wrong is to come up with a counterexample. Here is how that must work:

We have three computers, called SOURCE, ALICE, and BOB. There are two communication channels, one from SOURCE to ALICE, and one from SOURCE to BOB. The channels are one way, i.e, you can only send data from SOURCE to ALICE, not the other way. The same goes for SOURCE to BOB.

At fixed intervals, the computer SOURCE should send a data package to ALICE and BOB. The data package can be anything. Upon reception, the computer ALICE will also receive one of two setting a=0 or a'=45 from the outside, that is, exogenously given. The computer ALICE should then produce an output of either +1 or -1. The same goes for BOB, with exogenously received settings either b=22.5 or b'=67.5.

We do this 10 000 times. Collect the results for ALICE and BOB. Match the 10 000 pairs for ALICE and BOB. Compute the following for each of the four combinations (a,b), (a,b'), (a',b), (a',b'):

[(Number of same results) - (Number of different results)]/(Total results for the combination)

We now have four correlations. If the exogenously given settings are generated at random (in any sensible sense of the word "random", e.g. by an RNG), Bell's theorem (or rather the CHSH theorem) says that at least one of them will be way off from what QM predicts (at least 0.2 off).

For anyone who wants to refute Bell's theorem, there is only one way forward: write computer programs that adhere to the above specifications and at the same time reproduce the four QM predictions for the correlations, within sensible error margins.

[FrediFizzx]

There needs to be four computer stations; add one for data analysis. This has already been done here.

EPRsims/Joy_local_CS_no0s3Ds0.pdf

CHSH result = 2.79567
Bam! Bell's junk physics theory is disproven.
.

[Heinera]

There needs to be four computer stations; add one for data analysis.
.

No, the data analysis is this:

[(Number of same results) - (Number of different results)]/(Total results for the combination)

You don't need an extra computer for that, do you?

[FrediFizzx]

There needs to be four computer stations; add one for data analysis.
.

No, the data analysis is this:

[(Number of same results) - (Number of different results)]/(Total results for the combination)

You don't need an extra computer for that, do you?

potayto, potahto. There are different procedures that can be done that don't change the physics.
.

[gill1109]

There needs to be four computer stations; add one for data analysis. This has already been done here.
EPRsims/Joy_local_CS_no0s3Ds0.pdf
CHSH result = 2.79567
Bam! Bell's junk physics theory is disproven.

Michel did not do what Heinera asked for. Nor did John Reed. Nor did you, Fred. Nor did anybody at all, ever, yet.

I wish you would try.

Nobody can change physics. That's not the question. The question is what can you do with a network of classical computers communicating classically according to a particular quite rigorous protocol. Can you imitate with classical computers and classical internet connections what they did in real life, adhering to that particular protocol, but with what they claimed were quantum sources, detectors and channels, in Delft, Boulder, Vienna, Munich...? Nobody has tried yet, as far as I know.

Of course, according to the quantum computing folk, you could do it with quantum computers connected by quantum internet, and that is how it was done in Delft, Boulder, Vienna, Munich. For instance, more recently still, Jian-Wei Pan is maybe doing it with the two measurement locations being in Shanghai and in Beijing, and the central location being the Chinese "Micius" satellite https://www.scientificamerican.com/article/qutrit-experiments-are-a-first-in-quantum-teleportation/

I hope to visit him in about a week.

I'll also take the fast train from Shanghai to Beijing. No quantum teleportation of human beings, yet. If you come to either place too, you can follow my course https://www.math.leidenuniv.nl/~gill/lecture_course.html there.

[minkwe]

I think we have somehow lost sight of the operational definition of Bell's theorem in the latest threads. Even those of you who think that the proof of Bell's theorem is wrong can't rule out the possibility that his conclusion is correct, can you? I mean, I could come up with a horribly wrong proof of Fermat's last theorem, but in some sense I would still have gotten to the right conclusion, wouldn't I? (since we now have Andrew Wiles' correct proof).

So, the only way to prove Bell wrong is to come up with a counterexample. Here is how that must work:

We have three computers, called SOURCE, ALICE, and BOB. There are two communication channels, one from SOURCE to ALICE, and one from SOURCE to BOB. The channels are one way, i.e, you can only send data from SOURCE to ALICE, not the other way. The same goes for SOURCE to BOB.

At fixed intervals, the computer SOURCE should send a data package to ALICE and BOB. The data package can be anything. Upon reception, the computer ALICE will also receive one of two setting a=0 or a'=45 from the outside, that is, exogenously given. The computer ALICE should then produce an output of either +1 or -1. The same goes for BOB, with exogenously received settings either b=22.5 or b'=67.5.

We do this 10 000 times. Collect the results for ALICE and BOB. Match the 10 000 pairs for ALICE and BOB. Compute the following for each of the four combinations (a,b), (a,b'), (a',b), (a',b'):

[(Number of same results) - (Number of different results)]/(Total results for the combination)

We now have four correlations. If the exogenously given settings are generated at random (in any sensible sense of the word "random", e.g. by an RNG), Bell's theorem (or rather the CHSH theorem) says that at least one of them will be way off from what QM predicts (at least 0.2 off).

For anyone who wants to refute Bell's theorem, there is only one way forward: write computer programs that adhere to the above specifications and at the same time reproduce the four QM predictions for the correlations, within sensible error margins.

You are obviously behind the times on this. Have you been asleep all this while? What you describe has no resemblance to an actual EPR experiment and thus completely irrelevant to QM.

Why don't you first present step by step the detailed QM prediction for your described simulation. In other words, I challenge your claimed QM prediction for your described simulation.

Again, the problem here is lack of physical intuition from mathematicians who have never performed a real life experiment. Anybody can write any junk down and claim QM predicts blah blah for it. I'm calling BS on that. Demonstrate it. Don't use a QM prediction for something else as a bait and switch.

I want to see the detailed QM prediction for the simulation you described.

Secondly, please describe how Alice and Bob will be able to correlate anything without additional information . How should they know which value from Bob corresponds to which value from Alice.

[Heinera]

You are obviously behind the times on this. Have you been asleep all this while? What you describe has no resemblance to an actual EPR experiment and thus completely irrelevant to QM.

Who said it should resemble "actual EPR experiments"? Bell's theorem compares LHV theories to QM theory. No actual experiments needed.

Why don't you first present step by step the detailed QM prediction for your described simulation. In other words, I challenge your claimed QM prediction for your described simulation.

You may challenge the predictions of QM all you want. But these are the predictions Bell used when drawing his conclusion, which is why they are relevant to Bell's theorem. If you think those predictions are wrong, then Bell's theorem shouldn't concern you.

Secondly, please describe how Alice and Bob will be able to correlate anything without additional information . How should they know which value from Bob corresponds to which value from Alice.

The two computers ALICE and BOB shouldn't correlate anything. That is a task left to us analyzing the results. Since the 10 000 results will be in sequence for both ALICE and BOB, the correspondence is trivial.

What you are trying to do is bring up the good old loophole discussion again. But that is irrelevant, because no one is arguing that the loopholes don't exist. This computer setup is deliberately specified in a way that rules out any possibilities for introducing loopholes.

[FrediFizzx]

There needs to be four computer stations; add one for data analysis. This has already been done here.
EPRsims/Joy_local_CS_no0s3Ds0.pdf
CHSH result = 2.79567
Bam! Bell's junk physics theory is disproven.

Michel did not do what Heinera asked for. Nor did John Reed. Nor did you, Fred. Nor did anybody at all, ever, yet.

I wish you would try.

Nobody can change physics. That's not the question. The question is what can you do with a network of classical computers communicating classically according to a particular quite rigorous protocol. Can you imitate with classical computers and classical internet connections what they did in real life, adhering to that particular protocol, but with what they claimed were quantum sources, detectors and channels, in Delft, Boulder, Vienna, Munich...? Nobody has tried yet, as far as I know. …

Sorry to bust your bubble but the simulation I posted at the link above is how Nature works and is how QM can predict the individual +/-1 outcomes for A and B. Nature does not have to follow a "particular rigorous protocol" made up by you, Heine or anyone else.
.

[gill1109]

There needs to be four computer stations; add one for data analysis. This has already been done here.
EPRsims/Joy_local_CS_no0s3Ds0.pdf
CHSH result = 2.79567
Bam! Bell's junk physics theory is disproven.

Michel did not do what Heinera asked for. Nor did John Reed. Nor did you, Fred. Nor did anybody at all, ever, yet.

I wish you would try.

Nobody can change physics. That's not the question. The question is what can you do with a network of classical computers communicating classically according to a particular quite rigorous protocol. Can you imitate with classical computers and classical internet connections what they did in real life, adhering to that particular protocol, but with what they described [first I wrote "claimed", but there was no *claim* at all - RDG] as quantum sources, detectors and channels, in Delft, Boulder, Vienna, Munich...? Nobody has tried yet, as far as I know. …

Sorry to bust your bubble but the simulation I posted at the link above is how Nature works and is how QM can predict the individual +/-1 outcomes for A and B. Nature does not have to follow a "particular rigorous protocol" made up by you, Heine or anyone else.
.

Experimenters are free to *choose* to follow a certain protocol. Nature has no choice but to submit to the protocol. The protocol was made up (and followed) by experimenters in Delft, Boulder, Vienna, Munich ... Not by me, Heine, John Bell, or anyone else.

[gill1109]

You are obviously behind the times on this. Have you been asleep all this while? What you describe has no resemblance to an actual EPR experiment and thus completely irrelevant to QM.

Michel, it seems you have been asleep for about five years, if not more. What about the actual EPR experiments whose results were published in 2015 - the experiments themselves carried out earlier that same year in Delft, Boulder and Vienna? And just after them came Munich, and after that, a host of others?

Most of the rest of the scientific world thought those experiments were relevant to QM. The results certainly seem to match QM predictions. AFAIK, there is no known other physical theory which makes the same predictions. Do you have one? Please modify your computer programs to match the protocol of those experiments *and* reproduce the results of those experiments.

By the way there was an amusing announcement on the internet today which vanished again in no time.
https://gizmodo.com/google-says-its-achieved-quantum-supremacy-a-world-fir-1838299829
"A Google researcher’s paper claiming to have achieved quantum supremacy, a major early milestone in the field of quantum computing, appeared on a NASA website this week before being removed, the Financial Times reports."

First you see it, then you don't. Most likely, the Google researcher's paper was only published in a parallel universe.

See also https://www.technologyreview.com/f/614416/google-researchers-have-reportedly-achieved-quantum-supremacy/

Unfortunately, The Financial Times is behind a paywall. A month's free trial costs one Euro... But you have to remember to stop if yourself, before the month has ended...

[minkwe]

Who said it should resemble "actual EPR experiments"? Bell's theorem compares LHV theories to QM theory. No actual experiments needed.

QM said it should. What do you think QM is all about if not actual experiments? So show us the detailed calculations for the QM prediction for the simulation.
You said:

We now have four correlations. If the exogenously given settings are generated at random (in any sensible sense of the word "random", e.g. by an RNG), Bell's theorem (or rather the CHSH theorem) says that at least one of them will be way off from what QM predicts (at least 0.2 off).

"...way off from what QM predicts", predicts for what. In your very statement, you imply that QM predicts a different result for the simulation than what will be obtained. Now you don't want it to resemble actual experiments, yet you want to compare a QM prediction for an actual experiment, with a silly simulation you have prescribed, which does not resemble the experiment, and you think that makes any sense? Perhaps you don't realize how silly statements like these sound. We already went through a similar exercise about randomness elsewhere. As soon as you scratch the surface, you find the shiny claims are backed by nothing.

You may challenge the predictions of QM all you want.

Reading comprehension 101: I challenged your claim that "at least one of them will be way off from what QM predicts" . QM makes predictions about experiments which are valid. It is you that is taking the QM prediction and claiming that it applies in a different scenario without justification. I challenge your claim, so demonstrate the prediction for exactly the scenario you want us to simulate.

The two computers ALICE and BOB shouldn't correlate anything. That is a task left to us analyzing the results. Since the 10 000 results will be in sequence for both ALICE and BOB, the correspondence is trivial.

It doesn't matter who is correlating after the fact, you still need to know which outcome at Alice corresponds to which results at Bob. It may appear trivial silly example advanced by a mathematician but once you start thinking about real experiments, which you must if you invoke QM, it falls apart in your hands.

What you are trying to do is bring up the good old loophole discussion again. But that is irrelevant, because no one is arguing that the loopholes don't exist. This computer setup is deliberately specified in a way that rules out any possibilities for introducing loopholes.

Nope, this is not about loopholes, I'm bringing up the fact that you apparently have not quite understood the importance of matching results. Matching of results is an unavoidable requirement of real experiments. Since QM deals with real experiments, you can't just eliminate the matching, and claim the same QM predictions still apply. If you insist, find me an experiment for which there was no matching and the QM results was obtained.

BTW, for anyone who has been fooled by the Delft, Boulder and Vienna experiments, let me ask you these questions:
- Does the word "heralding" exist in those papers? If they do any kind of heralding, they are doing coincidence matching. I'm not fooled by the word switch.
- Do they consider every particle pair produced by their sources in the final correlation calculations? Do they discard any data? Come on guys!

Here is a quote from the "loophole-free" Delft paper:

Every few hundred milliseconds, the recorded events are transferred to the PC. During the experiment, about 2 megabyte of data is generated every second. To keep the size of the generated data-set manageable, blocks of about 100000 events are saved to the hard drive only if an entanglement heralding event (E) is present in that block.

They are already "heralding" (aka selecting) blocks but even the selected blocks contain 100000 events and yet at the end their correlations were calculated 245 events! Yet as soon as anyone mentions data matching, the cheerleaders of the Delft paper are quick to pounce with "loophole" claims.

So again, the point is that you have to compare like to like. You guys have a tendency of comparing apples and oranges. Your inability to see the difference between them does not obviate the requirement to always compare comparable correlations. When will you guys learn this?

[FrediFizzx]

… Nature has no choice but to submit to the protocol. …

I think we found your problem with all of this.
.

[Heinera]

QM said it should. What do you think QM is all about if not actual experiments? So show us the detailed calculations for the QM prediction for the simulation.

QM is a theory. Corroborated by lots and lots of experiments. But the experiments are not the theory.

The QM predictions for the correlations are Since I'm not your physics teacher, I wont't post the detailed calculations. A simple Google search turned up dozens of excellent expositions - try that instead.

[rest is snipped]

Let me ask you a very simple yes/no question: Do you think it is possible to program three computers the way I specified, that give four correlations ?

[FrediFizzx]

Let me ask you a very simple yes/no question: Do you think it is possible to program three computers the way I specified, that give four correlations ?

Yes, it is possible.



That says it all.

EPRsims/Joy_local_CS_no0s3Ds0.pdf
.

[gill1109]

Here is a quote from the "loophole-free" Delft paper:

Every few hundred milliseconds, the recorded events are transferred to the PC. During the experiment, about 2 megabyte of data is generated every second. To keep the size of the generated data-set manageable, blocks of about 100000 events are saved to the hard drive only if an entanglement heralding event (E) is present in that block.

They are already "heralding" (aka selecting) blocks but even the selected blocks contain 100000 events and yet at the end their correlations were calculated 245 events! Yet as soon as anyone mentions data matching, the cheerleaders of the Delft paper are quick to pounce with "loophole" claims.
So again, the point is that you have to compare like to like. You guys have a tendency of comparing apples and oranges. Your inability to see the difference between them does not obviate the requirement to always compare comparable correlations. When will you guys learn this?

The selection of events is done independently of the selection of settings and independently of the measurement outcomes at the measurement stations. Michel, you need to read those papers better. The Delft experiment is a three measurement station experiment. Alice, Bob and Casper each choose a setting and do a measurement with binary outcome. The experimenters look at the correlations between Alice and Bob's outcomes, given Alice and Bob's settings and Casper's outcome.

The Munich experiment, published in 2016, works the same way.

[minkwe]

The selection of events is done independently of the selection of settings and independently of the measurement outcomes at the measurement stations. Michel, you need to read those papers better.

No, you need to read my post better before you think you have addressed my point. BTW, I'm surprised that as a statistician, you believe when they tell you that the selection is independent of the measurement settings. I have evidence that it is not. Others have published evidence that it is not. Perhaps you mean that they did not explicitly use the measurement settings or outcomes to select the data. But that is hardly the same as saying the selection is independent of the measurement settings.

Remember this is in the context of Heine proposing some simulation in which we have exactly 10000 data points for Alice and Bob without any need for selection/matching. And when I pointed out to him that his fantasy simulation has no resemblance to actual experiments, he brought up loopholes, and you suggested loopholes are no longer an issue. Now you imply I'm referring to selection based on settings or outcomes but you know very well that I have produced a simulation which achieved much better agreement with QM than the Delft experiment without any use of settings or outcomes for matching.

Of course the silliest implication of all of this is the believe by mysterians that there is something mysterious about nature but as soon as the objects involved are computers the mystery stops. My point again is the following:

- You guys claim the QM prediction for a specific scenario will be way off from the outcome of a simulation of the same scenario. But then when you describe the scenario for which QM predictions were made, they have no resemblance to the the scenario you are asking Bell opponents to simulate, and when this is pointed out, you then make silly statements like "Who said it should resemble "actual EPR experiments"? Bell's theorem compares LHV theories to QM theory.". It is a classic bait and switch tactic.

[minkwe]

QM is a theory. Corroborated by lots and lots of experiments. But the experiments are not the theory.

QM is a theory about experiments. QM makes predictions for real performable experiments, not impossible fantasies.

The QM predictions for the correlations are Since I'm not your physics teacher, I wont't post the detailed calculations.

The QM predictions for what? Your computer simulation or the EPR experiment. You seem to think it is automatically both, but I'm asking you to show the QM calculations for your simulation because there is no other basis to compare that the QM results for an experiment, with the outcome of a simulation that bears no resemblance to the experiment.

A simple Google search turned up dozens of excellent expositions - try that instead.

Okay then provide the search terms you used, let us verify that you did not use the words "EPR" or "experiment" in it.

Let me ask you a very simple yes/no question: Do you think it is possible to program three computers the way I specified, that give four correlations ?

Provided you are not assuming the matching is not required, the answer is Yes. But you should know this already. What the simulation means is the real question. If you are trying to compare outcomes of simulations with QM predictions of actual experiments, you better be sure that you have understood and represented every component and feature present in the real experiments, in the simulation. In my experience, I've found a severe lack of depth in the numerous fake challenges of Bell proponents online, requesting simulations. They often ignore the most crucial facts of actual experiments, they key being data matching.

[Heinera]

Provided you are not assuming the matching is not required, the answer is Yes.

There is no matching issue in this simulation. The matching is trivial. Any matching issue would be a well known loophole. I thought you claimed that loopholes had nothing to do with things, and that the problem with Bell's theorem was it's neglect of the fact that each particle can only be measured once, with one setting only? Which is how it is measured in this simulation.

[minkwe]

Provided you are not assuming the matching is not required, the answer is Yes.

There is no matching issue in this simulation. The matching is trivial. Any matching issue would be a well known loophole.

Well then what you are asking bears no resemblance to actual EPR experiments for which the QM predictions were made. So please provide the QM predictions for the scenario you describe or point to an experiment did not do matching.

I thought you claimed that loopholes had nothing to do with things, and that the problem with Bell's theorem was it's neglect of the fact that each particle can only be measured once, with one setting only? Which is how it is measured in this simulation.

No, you are selectively misremembering what I said. Please pay attention: viewtopic.php?f=6&t=395#p9535 and read the full sentences. The question of whether Bell's inequalities apply to the EPR experiments and what the proper upper bound for those experiments should be, is a separate question from what exactly it is the QM predictions are about. This is not rocket science.

[Heinera]

Since this simulation is intended to demonstrate a mathematical point, there is no matching issue. So let me ask the question again: Do you think it is possible to program three computers the way I specified, that give four correlations ?