Coming Soon!

[Joy Christian]

This would prove that the functions A and B which Joy wants Fred to supply simply do not exist.

I do not believe that for a second. Despite what Gill and John Reed have said, I am convinced that the process embedded in Fred's simulation, as described by him above, is manifestly local. That tells me that, while it is not going to be easy because the process is very complex, local analytical functions A(a, h) and B(b, h) can be written down for Fred's simulation.
.

[gill1109]

Nope. You are doing the wrong process. That doesn't prove non-locality.

The simple proof of non-locality, for me, is that you violated Bell’s inequality (originally proven by Boole in the 1850’s)!

I would also prove it by random testing if only you (Fred) would give us a version which was written in R or Python, and allowed the user to specify the settings and set the seed. Then I would quickly find a sequence of inputs for Alice and Bob such that if I change Bob’s n’th input then Alice’s n’th output changes, even though the seed is the same and all the other inputs are the same; the first n - 1 outputs did not change. Alice’s n’th input did not change. But changing Bob’s n’th input changed Alice’s n’th output.

This would prove that the functions A and B which Joy wants Fred to supply simply do not exist.

Such an n can be found as follows. Find a seed and use some large N and random binary inputs so we get a big violation of CHSH. Now simply use the same seed and the same inputs, running the program just for the first n trials, and testing what happens if Alice’s or Bob’s last input is changed. This will certainly find an n which disproves locality. The search can be done automatically.

CHSH = 2.78024! Ok, I can try that but not fully understanding it. How do I change A or B's last input?
.

Write separate programs which produce two long lists of inputs for a CHSH type experiment.
Rewrite your program so that instead of generating settings itself, it simply reads settings from an external files; and reads a random seed from an external file.
Now, generate two long lists of inputs. Binary; completely at random. [I will check that statistically they do indeed look like independent fair coin tosses].
Give them to your program, together with a seed.
It comes out with a CHSH of about 2.8; N is very big.
Now give it just the first n = 30 (say) settings of each list of inputs, and the same seed as before
Change Alice's last input and redo
Instead, change Bob's last input and redo
Now you have *three* versions of the 30th pair of outputs
The original. And two versions after one setting was changed.
Did Alice's output change when her input was the same but Bob's was changed?
Did Bob's output change when his input was the same but Alice's was changed?

If the answer is "no" then increase n by 1 and repeat
[always same seed ....]

I guarantee you that you will find an n such that Alice's output depends on Bob's input or vice versa even though everything else up till now was the same (same seed, same inputs, same outputs... as in the original big N experiment)

[if my prediction fails I will publicly eat my hat and give you a lot of money. The hat will be a straw hat which I'll liquidize and cook in a nice wild mushroom bouillon]

[FrediFizzx]

@gill1109 No change for either A or B doing 30 trials. How many times do I have to increment trials by 1? You must have some guesstimate. This sounds like a wild goose chase.
.

[gill1109]

@gill1109 No change for either A or B doing 30 trials. How many times do I have to increment trials by 1? You must have some guesstimate. This sounds like a wild goose chase.

I am pretty sure that about 100 should be plenty. With overwhelming probability, 1000.

I think that about 1 in 10 trials needs to be “fixed” in order to get from 2 to 2 sqrt 2. There is some hard theory about this. I will look it up.

[FrediFizzx]

@gill1109 No change for either A or B doing 30 trials. How many times do I have to increment trials by 1? You must have some guesstimate. This sounds like a wild goose chase.

I am pretty sure that about 100 should be plenty. With overwhelming probability, 1000.

I think that about 1 in 10 trials needs to be “fixed” in order to get from 2 to 2 sqrt 2. There is some hard theory about this. I will look it up.

Ok thanks. I will need to somewhat automate the process.
.

[Joy Christian]

@gill1109 No change for either A or B doing 30 trials. How many times do I have to increment trials by 1? You must have some guesstimate. This sounds like a wild goose chase.

I am pretty sure that about 100 should be plenty. With overwhelming probability, 1000.

I think that about 1 in 10 trials needs to be “fixed” in order to get from 2 to 2 sqrt 2. There is some hard theory about this. I will look it up.

Ok thanks. I will need to somewhat automate the process.

If you are going to revise the code, then please also write the process for A and the process for B separately. This will be a cosmetic change, but an important cosmetic change.
.

[FrediFizzx]

@Joy Well, the automation for the test won't have anything to do with locality except for testing for it. But I am stuck on how to automate the run 30 then run 30 with a changed then run 30 with b changed.
.

[gill1109]

@Joy Well, the automation for the test won't have anything to do with locality except for testing for it. But I am stuck on how to automate the run 30 then run 30 with a changed then run 30 with b changed.
.

What about taking the inputs from run 29 and adding four possible continuations with one more trial? This gives four possible “trial 30” sets of an input pair and an output pair. Now you look to see if there are two pairs with one input in common and the other output differing.

[jreed]

@gill1109 No change for either A or B doing 30 trials. How many times do I have to increment trials by 1? You must have some guesstimate. This sounds like a wild goose chase.

I am pretty sure that about 100 should be plenty. With overwhelming probability, 1000.

I think that about 1 in 10 trials needs to be “fixed” in order to get from 2 to 2 sqrt 2. There is some hard theory about this. I will look it up.

Ok thanks. I will need to somewhat automate the process.

If you are going to revise the code, then please also write the process for A and the process for B separately. This will be a cosmetic change, but an important cosmetic change.
.

That would certainly be an important change. When you get it written, please post it. I will be very interested to see how you set up Intersection without using data from both A and B.

[Joy Christian]

.
Using data from both A and B is allowed while matching trial numbers, as explained by Fred in his step 2 here: viewtopic.php?f=6&t=481&start=120#p13595

It is the rest of the steps that I want him to write out completely separately for Alice and Bob, because they describe purely local processes of Alice and Bob, independently of each other.
.

[FrediFizzx]

.
Using data from both A and B is allowed while matching trial numbers, as explained by Fred in his step 2 here: viewtopic.php?f=6&t=481&start=120#p13595

It is the rest of the steps that I want him to write out completely separately for Alice and Bob, because they describe purely local processes of Alice and Bob, independently of each other.
.

Yeah, all I am doing there is getting the bulk of the events matched up trial number-wise. The experimenters use clocks to tag events and match them up. We have the luxury of tagging events with the actual trial numbers. OutA1 goes to listAa1 and outB1 goes to listBb1, completely local. OutA1 and outB1 don't match. ListAa1 and listBb1 match each other trial number-wise.
.

[FrediFizzx]

I'm actually giving up on trying to automate the process. I don't have the expertise to figure it out. And I certainly don't have the patience to do it 100 times or more manually. Perhaps John can figure out how to automate Gill's process for locality. I would rather spend my time studying gravitational theory with torsion.
.

[jreed]

I'll take a try at it. I am familiar with the version that calculates CHSH, and I think I can do it with that one.

[FrediFizzx]

I'll take a try at it. I am familiar with the version that calculates CHSH, and I think I can do it with that one.

Yeah, that is the version Gill is talking about. Good luck with it. It is probably not going to matter anyways since matching events is probably a non-local process. But it has to be done anywise or you won't know anything.
.

[Joy Christian]

I'll take a try at it. I am familiar with the version that calculates CHSH, and I think I can do it with that one.

Yeah, that is the version Gill is talking about. Good luck with it. It is probably not going to matter anyways since matching events is probably a non-local process. But it has to be done anywise or you won't know anything.

Matching events in your step 2 is not a nonlocal process. To be sure, it requires data from both A and B, but that is not nonlocal. Alice and Bob can share the hidden variables, and the trial numbers are part of the hidden variables. They must match. If anything after your step 2 requires sharing data from both A and B (which I don't think it does), then your model is nonlocal.
.

[FrediFizzx]

I'll take a try at it. I am familiar with the version that calculates CHSH, and I think I can do it with that one.

Yeah, that is the version Gill is talking about. Good luck with it. It is probably not going to matter anyways since matching events is probably a non-local process. But it has to be done anywise or you won't know anything.

Matching events in your step 2 is not a nonlocal process. To be sure, it requires data from both A and B, but that is not nonlocal. Alice and Bob can share the hidden variables, and the trial numbers are part of the hidden variables. They must match. If anything after your step 2 requires sharing data from both A and B (which I don't think it does), then your model is nonlocal.
.

Thanks Joy. Everything after step 2 is definitely local as it is very easy to see.
.

[gill1109]

I'll take a try at it. I am familiar with the version that calculates CHSH, and I think I can do it with that one.

Yeah, that is the version Gill is talking about. Good luck with it. It is probably not going to matter anyways since matching events is probably a non-local process. But it has to be done anywise or you won't know anything.

Matching events in your step 2 is not a nonlocal process. To be sure, it requires data from both A and B, but that is not nonlocal. Alice and Bob can share the hidden variables, and the trial numbers are part of the hidden variables. They must match. If anything after your step 2 requires sharing data from both A and B (which I don't think it does), then your model is nonlocal.
.

Exactly. If you change outcomes at B by using information correlated with the *setting* at A then you are violating locality.

Fred breaks a Bell inequality so somewhere he is not satisfying the conditions under which it can be derived! (Unless Bell’s inequality is wrong, of course … which would have hat-eating consequences for me).

[Joy Christian]

I'll take a try at it. I am familiar with the version that calculates CHSH, and I think I can do it with that one.

Yeah, that is the version Gill is talking about. Good luck with it. It is probably not going to matter anyways since matching events is probably a non-local process. But it has to be done anywise or you won't know anything.

Matching events in your step 2 is not a nonlocal process. To be sure, it requires data from both A and B, but that is not nonlocal. Alice and Bob can share the hidden variables, and the trial numbers are part of the hidden variables. They must match. If anything after your step 2 requires sharing data from both A and B (which I don't think it does), then your model is nonlocal.
.

Exactly. If you change outcomes at B by using information correlated with the *setting* at A then you are violating locality.

Fred breaks a Bell inequality so somewhere he is not satisfying the conditions under which it can be derived! (Unless Bell’s inequality is wrong, of course … which would have hat-eating consequences for me)

There are essentially two things you have not understood despite all the arguments with me for years. The first thing concerns some issues in the quaternionic 3-sphere model, but that is forgivable because the geometry and topology of the 3-sphere are not easy to understand.

But the second thing is the one you claim to be an expert in and yet you have not understood it despite years of arguments with me and many others: Bell inequality and singlet correlations are two separate things, entirely unconnected from one another. While absolutely nothing can exceed the bounds of Bell inequality, that does not prevent the existence of a local-realistic simulation of the singlet correlations. Fred's result was not easy to achieve, but some of us are not at all surprised that Fred has achieved what was supposed to be impossible to achieve.
.

[FrediFizzx]

Fred breaks a Bell inequality so somewhere he is not satisfying the conditions under which it can be derived! (Unless Bell’s inequality is wrong, of course … which would have hat-eating consequences for me).

Joy already broke Bell's theory years ago so you should have already ate the hat. When are you going to learn? Of course I didn't break the inequality because NOTHING can break it so your locality argument using the inequalities is false.
.

[FrediFizzx]

I relabeled and reconfigured the Mathematic code so might be easier to follow. Mainly grouped the A and B stuff separately in the matching section. Here is 200,000 trials at 1 degree resolution.



EPRsims/newCS-18-forum.pdf
EPRsims/newCS-18-forum.nb

Enjoy!
.