Has further work been done building on J Christian's model?

[gill1109]

The discussion in this thread has somehow veered to the topic of another. Namely, the thread http://www.sciphysicsforums.com/spfbb1/viewtopic.php?f=6&t=31 on Joy Christian's proposed experiment. But to answer a question by Michel, in this thread: yes, that experiment is good enough for me.

Please read Joy's short experimental paper http://arxiv.org/abs/0806.3078, or (just Section 4 of) the longer paper http://arxiv.org/abs/1211.0784.

Following Joy's explicit instructions, we will collect data from N runs. The data for each run consists of video films made by a battery of video cameras (or some other kinds of sensors) of two spinning hemispheres. We use them to compute the exact directions of the angular momentum s_k and minus s_k of each hemisphere in each run k; k=1, ..., N.

We now pick settings a and b. We *calculate* A_k(a) = sign(+s_k · a) and B_k(b) = sign (−s_k · b) for each run, and correlate them in the usual way: compute the average of the product. We call the result E(a,b).

Now we repeat this for other pairs of settings, same set of directions s_k if we like. Joy has said again and again explicitly that we are allowed to use the same set of runs.

I repeat: it was not my idea, it was Joy's idea, to use the same set of observed/computed spin directions s_k, k=1, ..., N, for each pair of setting directions a, b.

I quote from the other thread http://www.sciphysicsforums.com/spfbb1/viewtopic.php?f=6&t=31, the embedded quotes by Joy are taken verbatim from his experimental papers:

These sensors will determine the exact direction of the spin angular momentum s_k (or −s_k) for each shell in a given explosion, without disturbing them otherwise so that their total angular momentum would remain zero, at a designated distance from the center.

Once the actual directions of the angular momenta for a large ensemble of shells on both sides are fully recorded, the two computers are instructed to randomly choose a pair of reference directions, say a for one station and b for the other station.

The correlation function for the bomb fragments can then be calculated as E(a, b) = lim_N 1/N sum_k {sign(+s_k · a)} {sign (−s_k · b)}

I am making a bet about the value of the correlation function at just four points: E(a, b), E(a, b'), E(a', b), E(a', b').

We are going to do one experiment in order to settle this bet, and it has one "N", we don't take a limit.

I had insisted on a CHSH style experiment where each run is assigned (by random choice) to just one of the four correlations. But I am happy to relinquish that demand if Joy really is happy to calculate the four correlations on the same set of N runs (N values of s_k).

I explained elsewhere why I am certain I will win the bet when we follow Joy's original specification (same set of N exploding ball data for each pair of directions) but only with rather large probability if we follow more closely a standard CHSH protocol (different sets of perhaps different numbers of exploding balls, for each pair of directions).

Joy is certain he will win the bet. Joy is certain his maths is right and mine is wrong. I am certain his maths is wrong and mine is right. So we are both entirely happy with his specification of the experiment and with our ensuing bet. The outcome of the bet is determined as follows: do we see a value of the usual CHSH criterium more than half way beyond local realism (+2) towards the best that QM can do (+2 sqrt 2), yes or no? If yes, Joy wins, if no, I win.

I believe that Joy and I already agreed on the main lines of the experiment. We have almost finished recruiting a team of highly respected, neutral, independent adjudicators, who will advise us in drawing up a definitive protocol and definitive agreement on the bet: who will win, under which circumstances. When that's all done we proceed to do crowd-funding to fund the experiment. Joy believes he has an experimenter who can do it for him. The experiment will be performed and the adjudicators will adjudicate.

If anyone has advice to either of us how to fill in the details they are welcome to give advice on the other thread (the thread on Joy's experiment, http://www.sciphysicsforums.com/spfbb1/viewtopic.php?f=6&t=31). Please be constructive, scientific. This could be the experiment of the century. I believe that it's good either for a Nobel prize or for an igNobel prize, depending on who wins.

In other words, we have a win-win situation.

[minkwe]

Richard, tell us why E(a, b) = -a.b is not enough for you? Repeatedly pointing to Joy's paper is not a substitute for your own opinion why it is not enough. If you agree that it is enough then it is clear how to proceed: each run a large number of particles pairs are generated (not necessarily the same number). Alice and Bob randomly pick a and b. Then calculate E(a,b). We repeat as many runs as you like and each time we compare the corresponding E(a,b) with QM. If they match within statististical error Joy wins. No need to mention CHSH. Do you have a problem with that?

It is enought for me. I have no problem with that at all, and I already told you so.

I told you that I want even less: I just want to see the values of E(a, b) at four particular pairs of values of a and of b.

Does anyone have a problem with that?

"as many runs as you like" includes 4 doesn't it?

[minkwe]

There is something odd about Joy's experiment. We are going to collect N values s_k of the direction of spin of two hemispherical objects. We are then going to pick directions a and b, and compute 1/N sum_k sign(a . s_k) sign(b . -s_k). We'll do this for several pairs of directions and Joy has no objection at all that we use the same N spin directions s_k for each new pair of directions.

Now you are back talking about computing all the correlations from the same set of pairs again. Bah !? I give up. This is a waste of time.

Joy has no objection at all that we use the same N spin directions s_k for each new pair of directions.

Huh!?

You can calculate both A(a, lambda) and A(a', lambda) if you want, but cannot add A(a, lambda) and A(a', lambda), or subtract them, in the computation of the correlation. The four correlations must be calculated separately, as specified in equation (16) of the first paper:

For the record, let me repeat that equation (16) of my attached
experimental paper describes exactly how the expectation values
E(a, b), E(a', b), E(a, b'), and E(a', b') are to be computed in my
proposed experiment. Four separate sums are to be calculated as
follows

E(a, b) = 1/N Sum_j A_j B_j ,

E(a, b') = 1/N Sum_j A_j B'_j ,

E(a', b) = 1/N Sum_j A'_j B_j ,

and

E(a', b') = 1/N Sum_j A'_j B'_j .

It is a matter of indifference whether N here is chosen to be the same
or different for each of the four alternatives.

The experimental procedure described in my paper is unambiguous.

[gill1109]

Richard, tell us why E(a, b) = -a.b is not enough for you? Repeatedly pointing to Joy's paper is not a substitute for your own opinion why it is not enough. If you agree that it is enough then it is clear how to proceed: each run a large number of particles pairs are generated (not necessarily the same number). Alice and Bob randomly pick a and b. Then calculate E(a,b). We repeat as many runs as you like and each time we compare the corresponding E(a,b) with QM. If they match within statististical error Joy wins. No need to mention CHSH. Do you have a problem with that?

It is enought for me. I have no problem with that at all, and I already told you so.

I told you that I want even less: I just want to see the values of E(a, b) at four particular pairs of values of a and of b.

Does anyone have a problem with that?

"as many runs as you like" includes 4 doesn't it?

Thanks for your nice remark!

We repeat with as many runs as Joy likes, but it must be at least as many runs as I find minimally acceptable. Joy and I have already been discussing reasonable numbers. I already told him there would be a minimal N demand by me. He accepted that.

My minimal acceptable N depends on the size of the stake. I don't mind losing 100 Euro's with a chance of 1 in 100, but I do mind losing 1000 Euro's with a chance of 1 in 100.

We will do the experiment with finite N, but we want to decide between Joy's prediction at N = infty of 2.828, and my prediction at N = infty of 2.000 or less. With N=4 statistical error is so large that we cannot decide anything. We need enough runs, such that if either of us is right, there is large probability that we'll come to the right conclusion. Very roughly speaking the error is about of size 1 / sqrt N so with N = 10 000 the standard error in the CHSH quantity will be about 0.01. Half the distance between 2 and 2 sqrt 2 is about 0.4 which is in the order of 40 standard deviations. So N = 10 000 would be fine for me, and no doubt fine for Joy too, but he already said that he was thinking of a much larger number too. Fine by me. It makes the experiment more expensive, and the analysis more time consuming, while it won't make much difference to the answer.

This discussion belongs, by the way, in the thread on Joy's experiment:

http://www.sciphysicsforums.com/spfbb1/viewtopic.php?f=6&t=31

[gill1109]

There is something odd about Joy's experiment.

Now you are back talking about computing all the correlations from the same set of pairs again. Bah !? I give up. This is a waste of time.

Please tell Joy what you think about this! He wants the correlations calculated "separately" but he allows us to use the same set of N observed spin directions s_k.

He told me many times, that I am not allowed to put the different values of A_k(a) = sign( a . s_k) (as a varies) and of B_k(b) = sign( b . - s_k) (as b varies) into one spread-sheet - but I am not going to do that. I'm not allowed to look at the sums and differences between, say A_k(a) and A_k(a') - but I am not going to do that either.

We are just going to look at the following sums and difference: E(a, b) + E(a, b') + E(a', b) - E(a', b'); where E(a, b) = 1/N sum A_k(a) B_k(b), etc. Each correlation will be calculated separately. I am happy with that. Joy says we can use the same s_k, k = 1, ..., N, throughout. Wonderful! I'm very happy with that.

I wrote in this forum, about what Joy wrote in Joy's experimental paper. I quoted from Joy's experimental paper. Please read what he wrote. If you think that what Joy wrote there is nuts, please tell him that, but in a polite way of course. Or tell him by private email in a less polite way. (In Dutch there is a saying "a soft doctor makes stinking wounds". In English there is "sometimes it's cruel to be kind" (Shakespeare: "Hamlet").)

It sure should be important to Joy, since he and I are negotiating the terms of a bet concerning the outcome of his experiment; the experiment described in detail in his paper(s).

The links to Joy's papers: http://arxiv.org/abs/0806.3078, or (Section 4 of) the longer paper http://arxiv.org/abs/1211.0784.

This discussion, too, belongs in the thread on Joy's experiment:

http://www.sciphysicsforums.com/spfbb1/viewtopic.php?f=6&t=31

[minkwe]

Please tell Joy what you think about this!

Huh? I just reminded you what Joy himself says about this.

[gill1109]

Please tell Joy what you think about this!

Huh? I just reminded you what Joy himself says about this.

Read what Joy says in Joy's paper. That's what counts. (Though you may notice also his words, written in this forum and on other discussion fora, "it is a matter of indifference to me").

Joy and I are holding a bet about the experiment which he describes in his paper. If you think he should revise those descriptions, then you can tell him, and he should revise the papers and post the revisions to arXiv. Our bet will need to be re-negotiated. He would be moving the goal-posts, right? However the bet is not yet definitively on, so there is room for re-negotiation.

Notice that there's another thread on this forum for discussing the experiment. This thread is about whether people have built further on Joy's work.