SciPhysicsForums.com

A guest blog thread I did at,

http://challengingbell.blogspot.com/201 ... f-joy.html

This proves that Joy Christian's classical local realistic model does in fact contradict Bell's theorem via the computer program GAViewer. Here is the GAViewer script code that is based on Joy's one page paper. Code is based much on previous work by Albert Jan Wonnink with some contributions from others.

Code: Select all

function getRandomLambda()
{
     if( rand()>0.5) {return 1;} else {return -1;}
}

function getRandomUnitVector() //uniform random unit vector:
   //http://mathworld.wolfram.com/SpherePointPicking.html
{
     v=randGaussStd()*e1+randGaussStd()*e2+randGaussStd()*e3;
     return normalize(v);
}

     batch test()
{
     set_window_title("Test of Joy Christian's arXiv:1103.1879 paper");
     N=20000;  //number of iterations (trials)
     I=e1^e2^e3;
     s=0;
     
     a=getRandomUnitVector();
     b=getRandomUnitVector();
     minus_cos_a_b=-1*(a.b);

     for(nn=0;nn<N;nn=nn+1) //perform the experiment N times
     {
          lambda=getRandomLambda(); //lambda is a fair coin resulting in +1 or -1
          mu=lambda * I;  //calculate the lambda dependent mu
          C=-I.a;  //C = {-a_j B_j}
          D=I.b;   //D = {b_k B_k}
          E=mu.a;  //E = {a_k B_k(L)}
          F=mu.b;  //F = {b_j B_j(L)}
          A=C E;  //eq. (1) of arXiv:1103.1879, A(a, L) = {-a_j B_j}{a_k B_k(L)}
          B=F D;  //eq. (2) of arXiv:1103.1879, B(b, L) = {b_j B_j(L)}{b_k B_k}
          q=0;
          if(lambda==1) {q=((-C) A B (-D));} else {q=((-D) B A (-C));} //eq. (6)
          s=s+q; //summation of all terms.
     }
     mean_mu_a_mu_b=s/N;
     print(mean_mu_a_mu_b); //print the result
     print(minus_cos_a_b);
     prompt();
}
//Typical result is:
//mean_mu_a_mu_b = 0.87 + 0.00*e2^e3 + 0.00*e3^e1 + 0.00*e1^e2
//minus_cos_a_b = 0.87
//The scalar parts match and others vanish!  Proving the result is -a.b.
//Thus Dr. Christian's arXiv:1103.1879 paper is a classical local realistic
//counter-example that in fact contradicts Bell's theorem.


GAViewer can be found here.

Hi Fred

I no longer have R on my computer and although I looked a little at GA, I have not played with it, and its a long time since I last read Joy's one-page paper, so could you say more about one aspect. if that is OK:

The GA program line if(lambda==1) {q=((-C) A B (-D));} else {q=((-D) B A (-C));} //eq. (6)
seems to use an "if" statement which appears, maybe naively, to me to be the only place that criticism could arise about non-locality creeping in covertly? There is no explicit "if" statement within Eq 6?

In the R software versions of last year [was it as long ago as last year?], if I am paraphrasing the flatlanders viewpoint correctly, they claimed that the software used non-local information to prune down the data points which is equivalent to obtaining a nice curve on a sheet of paper by trimming away the non-fitting parts with a pair of scissors. They claimed that was destroying valid, but non-fitting, data. It did not help me much when at one point I was trying and failing with my own program,to get a nice cos curve with 100% of generated pairs. After some inquiries it turned out that not all pairs generated in the 'official' programs were making it through to the results. So could you confirm that all generated pairs are being used in the results in your new GA program? The spherelanders view was that some of the pairs generated in R and Java etc. were invalid in a sphereworld. And I can go along with that and it does seem that to obtain valid data for the sphereworld you need to use non-local information. But that's OK as, say, for a sculptor to hack away at a cube to turn it into a sphere, the sculptor would need to look at the work of art in progress from all angles, as a whole, to check that a sphere is emerging. When, and only when, you have the sphere you can start to calculate results with the valid sphereland pairs of data. But is the advantage of this GA software that there is no need to remove any cubeland data pairs as the GA software only uses sphereland data pairs? So there is no need for you to prune data in your code as the GA framework has pre-set that for you?

Best wishes

Ben6993 wrote: Hi Fred

I no longer have R on my computer and although I looked a little at GA, I have not played with it, and its a long time since I last read Joy's one-page paper, so could you say more about one aspect. if that is OK:

The GA program line if(lambda==1) {q=((-C) A B (-D));} else {q=((-D) B A (-C));} //eq. (6)
seems to use an "if" statement which appears, maybe naively, to me to be the only place that criticism could arise about non-locality creeping in covertly? There is no explicit "if" statement within Eq 6?

Hi Ben,

The code script above is for the GAViewer program that you can find at the link above. It is a pretty easy program to download and run the script on. Just copy and paste the script to a text editor like Wordpad, etc. and then just use a .g extension when saving the file as a text file. Then after launching GAViewer, do File, Open g file. You should see it parsing on the command line. Then just type test() at the command line and wait a couple of minutes for the result.

What the "if else" line does is to correct for GAViewer being in a fixed right handed bivector basis. And is also Joy's main physics' postulate. When lambda = 1, the particle pairs are in the right handed basis; when lambda = -1 (else), they are in the left handed basis. What do you see when looking at a left handed system from a right handed only perspective? The geometric order is reversed. Very simple.

Ben6993 wrote:In the R software versions of last year [was it as long ago as last year?], if I am paraphrasing the flatlanders viewpoint correctly, they claimed that the software used non-local information to prune down the data points which is equivalent to obtaining a nice curve on a sheet of paper by trimming away the non-fitting parts with a pair of scissors. They claimed that was destroying valid, but non-fitting, data. It did not help me much when at one point I was trying and failing with my own program,to get a nice cos curve with 100% of generated pairs. After some inquiries it turned out that not all pairs generated in the 'official' programs were making it through to the results. So could you confirm that all generated pairs are being used in the results in your new GA program? The spherelanders view was that some of the pairs generated in R and Java etc. were invalid in a sphereworld. And I can go along with that and it does seem that to obtain valid data for the sphereworld you need to use non-local information. But that's OK as, say, for a sculptor to hack away at a cube to turn it into a sphere, the sculptor would need to look at the work of art in progress from all angles, as a whole, to check that a sphere is emerging. When, and only when, you have the sphere you can start to calculate results with the valid sphereland pairs of data. But is the advantage of this GA software that there is no need to remove any cubeland data pairs as the GA software only uses sphereland data pairs? So there is no need for you to prune data in your code as the GA framework has pre-set that for you?

This is not an event by event simulation using the +/- 1 outcomes. This is an event by event numerical validation of Joy's correlation calculation in his one page paper. IOW, that the result of his equations (1) and (2) is -a.b. It is pretty easy to see analytically that the result is -a.b but now we have proof via a computer program that Joy's classical local realistic model does in fact contradict Bell's theorem. There is nothing non-local in that one page paper. In fact, a loophole free experiment will also validate Joy's model.

Hi Fred

Thanks for the explanations.

The new software may not be event by event simulation using +/- 1 outcomes but it is event by event and does use a SpherePointPicking routine which therefore has some similarities with the R and Java [and other] simulations.
I note that there were until recently two Cubist comments on Albert Jan's website. One of them seems easy to dismiss; it was disdain about using G = a.b in the calculations as it was inserting the desired answer into the calculation. That is a very strange comment as the software uses a.b as the correct answer only so as to display the software output as a residual from the desired answer. I cannot see anything wrong with using G = a.b to obtain a residual which is asymptotic to zero? A very strange comment unless I am missing something?

The other Cubist comment was that +/- 1 outcomes were not being used. This is some implied idea, at least in my interpretation, that the scissors and paste removal of bad pairs, or not 'good' indexed pairs, in the R programs has been replaced in your new software by a non-straightforward correlation routine which is equivalent to snipping a bit off each pair's outcomes. I.e. instead of using only unit outcomes but discarding non-'good' ones, all outcomes are tailored to non-unit values and all can be kept. So this second comment boils down to accusing the correlation routine as being used to tailor the outcomes. But this is just a very old argument about the one-page paper.

Ben6993 wrote:...it was disdain about using G = a.b in the calculations as it was inserting the desired answer into the calculation. That is a very strange comment as the software uses a.b as the correct answer only so as to display the software output as a residual from the desired answer. I cannot see anything wrong with using G = a.b to obtain a residual which is asymptotic to zero? A very strange comment unless I am missing something?

You are not missing anything, Ben, Gill's comment is idiotic. But then again, I find all the comments by that disdainer idiotic. I have decided to ignore his ramblings.

In any case, Fred has now updated the code and removed the G = a.b part. He has also posted a Python translation of the GAViewer code, done by Michel Fodje.

Yes, for some unknow reason I didn't realize that a and b were the same value for every iteration in the loop. So there was no advantage for adding a.b to the sum of final result. Now, if a and b are put in the loop so that they have different values for every iteration, then it would be necessary to add a.b so that one could see that -a.b is getting cancelled out perfectly. I have tried it and it does in fact work that way also.

FrediFizzx wrote:This is not an event by event simulation using the +/- 1 outcomes. This is an event by event numerical validation of Joy's correlation calculation in his one page paper. IOW, that the result of his equations (1) and (2) is -a.b. It is pretty easy to see analytically that the result is -a.b but now we have proof via a computer program that Joy's classical local realistic model does in fact contradict Bell's theorem. There is nothing non-local in that one page paper. In fact, a loophole free experiment will also validate Joy's model.

Nobody doubts that one can compute somehow -a.b. A local realistic model has to do more, it has to compute +/-1 outcomes A(a,l) and B(b,l). Else, it is a nice numerical computation of -a.b which nobody needs, because it has nothing to do with Bell's inequality.

Schmelzer wrote:Nobody doubts that one can compute somehow -a.b. A local realistic model has to do more, it has to compute +/-1 outcomes A(a,l) and B(b,l). Else, it is a nice numerical computation of -a.b which nobody needs, because it has nothing to do with Bell's inequality.

Many +/- 1 outcome simulations have been done for Joy's model. Let's see another example of a local realistic model that produces -a.b. I suspect it is not as trivial as you are claiming. Joy's model is in fact an alternative explanation for quantum correlations in a local realistic way.

FrediFizzx wrote:Many +/- 1 outcome simulations have been done for Joy's model.

I have not even seen yet a +/- 1 outcome formula for A(a,l) and B(b,l) which one could simulate. If you have seen such a formula, tell me where.

FrediFizzx wrote:Let's see another example of a local realistic model that produces -a.b. I suspect it is not as trivial as you are claiming.

If I'm free to use arbitrary spaces for values of A and B, and, additionally, free to define the product operation between them, no problem. A(a,l) = a in R^3, B(b,l) = -b in R^3, => A(a,l)B(b,l) = -<a,b> in R. And Bell has not even tried to prove a theorem claiming something different.

He has proven a theorem for A(a,l) in {-1,1}, B(b,l) in {-1,1}, and AB being simply the product of these numbers.

FrediFizzx wrote:Many +/- 1 outcome simulations have been done for Joy's model.

Schmelzer wrote:I have not even seen yet a +/- 1 outcome formula for A(a,l) and B(b,l) which one could simulate. If you have seen such a formula, tell me where.

http://rpubs.com/jjc/

FrediFizzx wrote:Let's see another example of a local realistic model that produces -a.b. I suspect it is not as trivial as you are claiming.

Schmelzer wrote:If I'm free to use arbitrary spaces for values of A and B, and, additionally, free to define the product operation between them, no problem. A(a,l) = a in R^3, B(b,l) = -b in R^3, => A(a,l)B(b,l) = -<a,b> in R.

.

LOL! That is not a physical model at all. And it is supposed to be <A(a,l)B(b,l)> = -a,b, not A(a,l)B(b,l) = -<a,b>.

Schmelzer wrote:And Bell has not even tried to prove a theorem claiming something different. He has proven a theorem for A(a,l) in {-1,1}, B(b,l) in {-1,1}, and AB being simply the product of these numbers.

That is fine if all you are doing is mathematics; we are doing physics.

FrediFizzx wrote:http://rpubs.com/jjc/

And which of these things contains the formula for A(a,l) in {-1,1}, B(b,l) in {-1,1}, and AB being simply the product of these numbers? Or I'm supposed to take a look at all of them to search myself?

Ok, the first one I will try. Ups. Some "good" which depends on a and b, and a lot of following things which depend on "good". LOL. This is a nice example for A(a,b,l), B(a,b,l).

FrediFizzx wrote:

Schmelzer wrote:If I'm free to use arbitrary spaces for values of A and B, and, additionally, free to define the product operation between them, no problem. A(a,l) = a in R^3, B(b,l) = -b in R^3, => A(a,l)B(b,l) = -<a,b> in R.

.

LOL! That is not a physical model at all. And it is supposed to be <A(a,l)B(b,l)> = -a,b, not A(a,l)B(b,l) = -<a,b>.

I do not object if you choose different denotations. And, of course, I do not even claim that this would be a "physical model". It has, of course, nothing to do with the experiments, which measure results A(a) in {-1,1}, B(b) in {-1,1}, and multiply these numbers.

FrediFizzx wrote:

Schmelzer wrote:He has proven a theorem for A(a,l) in {-1,1}, B(b,l) in {-1,1}, and AB being simply the product of these numbers.

That is fine if all you are doing is mathematics; we are doing physics.

AFAIK the actual measurement results obtained in the experiments are values in {-1,1}. Not?

Schmelzer wrote:

FrediFizzx wrote:http://rpubs.com/jjc/

And which of these things contains the formula for A(a,l) in {-1,1}, B(b,l) in {-1,1}, and AB being simply the product of these numbers? Or I'm supposed to take a look at all of them to search myself?

Ok, the first one I will try. Ups. Some "good" which depends on a and b, and a lot of following things which depend on "good". LOL. This is a nice example for A(a,b,l), B(a,b,l).

You have to read and understand the paper http://arxiv.org/abs/1405.2355 mentioned in the introduction to "un-confuse" yourself about that simulation.

FrediFizzx wrote:You have to read and understand the paper http://arxiv.org/abs/1405.2355 mentioned in the introduction to "un-confuse" yourself about that simulation.

So, I have to "understand" (I translate this as to accept without doubt) a paper which claims that some local process is equivalent to an obviously nonlocal one, and then to accept the simulation of the nonlocal process as a simulation of a local one?

Sorry, but this is nothing I'm ready to accept. The proof of Bell's theorem is much easier to understand than this equivalence proof. So, I Bell's proof. And I do not understand why one uses a suspect equivalence proof and simulates only a nonlocal process instead of simulating the local process directly?

I would suggest that you read the paper and ask questions. Then you might understand what the "good" part of the simulation does. Right now you don't have a clue.

FrediFizzx wrote:I would suggest that you read the paper and ask questions. Then you might understand what the "good" part of the simulation does. Right now you don't have a clue.

My guess is enforcing eq. (30).

And the question remains unchanged. Why not simply using a computation of the e_sigma, s_sigma which do not depend in any way on a and b, as the condition (30) obviously does? The article seems to claim that this should be possible, given that this condition is claimed to be a somehow rewritten (19). If it is possible, do it.

Schmelzer wrote:

FrediFizzx wrote:I would suggest that you read the paper and ask questions. Then you might understand what the "good" part of the simulation does. Right now you don't have a clue.

My guess is enforcing eq. (30).

And the question remains unchanged. Why not simply using a computation of the e_sigma, s_sigma which do not depend in any way on a and b, as the condition (30) obviously does? The article seems to claim that this should be possible, given that this condition is claimed to be a somehow rewritten (19). If it is possible, do it.

I'm going to do this explanation for the benefit of any lurkers that may be reading as I figure you are so severely locked into de Broglie-Bohm non-localness that you aren't going to ever change your mind.

Yes, it is possible and here is a PDF of a Mathematica version of the simulation that has no a, b connection for "good". But then we don't get the sexy surface plot. LOL!

What Joy's complete states theory is saying is that only the "good" states can be detected. The other states don't ever exist in the first place so impossible to detect. You can read more about it on the blog or in the book. Ok lurkers, now is the chance to ask your questions.
.

FrediFizzx wrote:What Joy's complete states theory is saying is that only the "good" states can be detected. The other states don't ever exist in the first place so impossible to detect. You can read more about it on the blog or in the book. Ok lurkers, now is the chance to ask your questions.

In other words, Fred, one cannot see what is not there in the first place: http://rpubs.com/jjc/16567.

In this simulation, the initial states (v, g) are "prepared" by Nature within the physical space S^3. Those are all and the only states that actually exist.

The local-realistic correlations are then calculated between A(a; v, g) = sign(+v.a) = +/-1 and B(b; v, g) = sign(-v.b) = +/-1 in the standard way to give E(a, b) = -a.b.

FrediFizzx wrote:Yes, it is possible and here is a PDF of a Mathematica version of the simulation that has no a, b connection for "good". But then we don't get the sexy surface plot. LOL!
What Joy's complete states theory is saying is that only the "good" states can be detected.

That means, one first prepares the state, by eLeft, eRight, lambda.

Then, one has to wait until Alice and Bob make their decisions what to measure, to compute
aliceDet[[i]] = test[aliceAngle, eLeft, λ];
bobDet[[i]] = test[bobAngle, eRight, λ].

And only if this gives nice results, the experiment counts. A classical detector efficiency loophole. That detector inefficiency destroys the whole point I have described in the simple card game variant http://ilja-schmelzer.de/realism/game.php in the following words:
"An application in our game would require that they work without failure. Indeed, if the detector fails, I would have to say something like "Sorry, I have lost the cards of this round, so, we cannot count this round." This is, of course, something you cannot accept. It would allow us to prepare the cards as 1: red, 2: black, 3: invalid. For this preparation, you can never win, because there are no cards with the same color."

FrediFizzx wrote:The other states don't ever exist in the first place so impossible to detect.

Whatever the excuse for the failure - if the states don't exist or exist can be computed (or at least is computed in this particular simulation) only if a and b are known.

Schmelzer wrote:That means, one first prepares the state...

Wrong!

"One" does not prepare anything. Nature prepares the states (v, g).

E(a, b) = -a.b follows: viewtopic.php?f=6&t=164&p=4143#p4141

Joy Christian wrote:

Schmelzer wrote:That means, one first prepares the state...

Wrong! "One" does not prepare anything. Nature prepares the states (v, g).
E(a, b) = -a.b follows: viewtopic.php?f=6&t=164&p=4143#p4141

The last time I have heard about this, in quantum experiments people use experimental procedures named "state preparation". This is, typically, a measurement, and only in the case when this preparation measurement gives a nice answer, the state is considered to be prepared. So, I would not object if you say that "Nature prepares". The human experimenter, in this case, only accepts some choices of Nature, and rejects others.

But this state preparation procedure is presumed to be successfully finished before Alice and Bob make their decisions what to measure. Thus, Nature is, in this theory, also obliged to prepare the state without knowing a or b - else, we simply have the superdetermination loophole.