The following is my email correspondence with Joy Christian. It shows how seriously he took my bait and how he quite agreed with my main point, namely that the “Quantum Crackpot Randi Challenge” should be earnestly attempted.

Somebody who just does not believe in what modern experiments observe, say relativity, is a common crackpot, nothing too bad – could be a nice guy. However, telling people incessantly to have the model that explains quantum experiments while carefully hiding that he actually does not accept those experiments, that is outright dishonest. Obnoxious his insistence about that other people’s stuff is all “simple minded”. Bell simple minded? At this point it is not anymore about a nice, honest crackpot, but about somebody who plays the fame game, somebody driven by megalomaniac attention seeking.

Since he forwarded my mail to another crackpot on the internet without asking my consent, I just go ahead and post this here. There are long parts of mathematically blown up trivialities (it all comes down to the formula I = +/-1 and no more!), thus I put in bold font all you need to read. For those too busy, here the summary in two lines:

Joy Christian: “You are of course right to worry that this is not what we observe in experiments.
Proper Scientists: “CRACKPOT!”

I also put a few comments like so “[Comment: blah blah]” and will at the very end conclude with some remarks.

15 May 2011 06:22

Dear Professor Christian!

Your pointing out that the codomain should be equatorial S2 in S3 is intriguing. However, the very core of your argument, i.e. locality, may benefit from a more convincing exposition; I think this is proven by your trying for a number of years now to convince people without full success. Are you interested in cooperating to make the core of your argument more accessible to a physics audience?
[Comment: I hope it is perfectly clear to all that I never seriously considered cooperating with that crackpot!]

My idea is the following: Locality means that the photon that arrives at Alice’s detector has all the necessary information locally with it. This local information determines either completely what Alice is going to measure, or at least the probabilities of what she can measure are determined in such a way that afterwards all expectation values between Alice and Bob will work out correctly. In other words, the photon can be thought of as having a hidden book or little computer with it, containing all the information necessary (either all infinite data on S2 or in form of a finite amount of parameters that together with the direction of Alice’s detector axis feed into a formula evaluated by the photon’s little hidden,classical (i.e. non-quantum) computer).

For a physicist, math is always suspect as long as there is no working model. If the photon can have all the information necessary locally with it in a classical computer, it must be possible to make a working model in a classical computer.

The modeled situation can be very simple: We only consider three different detector angles that can be chosen by Alice and Bob, namely 0, Pi/8, and 3Pi/8, and only consider two photon singlet EPR states (if alpha=beta, Alice will get +1 when Bob gets -1). This setup is well known,  can be understood by any physics student, and it violates Bell’s inequality severely (0.32 > 0.43 is clearly wrong by a large amount).

Any codomain topology can be simulated in virtual reality plainly by programming the correct relations between the data. Since the violation is numerically large and computer memory is very large nowadays, the problem of a finite resolution of the S2 or even S3 codomain should not be a problem.

I will write a simple program that draws a complete state (according to your model mu=+/-I) randomly and evaluates the necessary data that Alice’s and Bob’s photons need to take with them. The photons (i.e. their hidden variables) are send to two other computers via the internet, where other people (Alice and Bob) pick angles 0, Pi/8, or 3Pi/8 independently for every arriving photon as they wish or randomly chosen by their computer. Say ten thousand EPR pairs are send and evaluated in this way. The statistical correlations of the locally evaluated measurement results are as a last step compared and the Bell violation is displayed, which is accomplished, if desired, even on a forth computer to make it utterly convincing. All the programs are open source, everybody can check that the programs do not secretly establish an internet connection between Alice’s and Bob’s computers after the angles are chosen (in other words: there is no cheating via secret non-locality). If the Bell inequality is violated in this way, it will be a huge confirmation that is going to spread over the internet like a firestorm.

Indeed, this would be so convincing, if I could figure out how to actually extract the measurement results from your A(alpha,mu)=+mu(…) and B(beta,mu)=-mu(…), I would have done it already. However, I do not see how it can be done [Comment: This is a lie of course, the calculation is very simple.], and given the real valued quaternions and soon, there should not be any problem. In fact, I fear that if this cannot be done rather soon, it will be likely understood as a clear indication that your argument is wrong in the very core.

I would be very honored indeed if you were willing to cooperate on this so that we can prove locality in a totally convincing manner that everybody can accept even if they do not know Clifford algebra and that nobody can refute, as everybody would be able to just download it as an internet java application and run it with their friends, seeing the Bell inequality violated in seconds with today’s computer and connection speeds.

Looking forward to your clarifications and hopefully cooperation




Joy’s Answer:

Hello Sascha,

Thank you for your message. Simulating my model is not so easy. Several people have tried over the years without success. The problem is that the hidden variable "mu" of my model is quite different from the simple minded hidden variables usually considered. It randomizes the handedness of the entire Euclidean space in which the experiment is supposed to be taking place. I think the best hope for simulating my model is---not by trying to extract information from "mu"---but by implementing the correct algebraic logic of the model as described in my latest
paper (see the attached). What has to be done is program equations (1) and (2) of the attached paper in such a manner that the fixed and random bivector basis defined by equations (3) and (4) in it are duly respected. [Comment: This is precisely what I discuss below in the email of the 17th of May; this is where he claims he can win the Quantum Crackpot Randi Challenge.] Note that what is being randomized in equation (4) are the basis of an entire algebra, which in turn defines the sense of rotation for every single point of the 3-sphere. If you are able to program the mathematics described in equation (4) of the attached paper, then you might have a chance of success. I am coping this email to Albert Jan, who is also interested in simulating my model. See also his website


16 May 2011 10:18

Dear Joy,

Thank you sincerely for your reply. Please allow me to clarify the situation.

1) One may prescribe any topology desired to a simulated domain. [Other’s “simple minded hidden variables” in “Euclidean space in which the experiment is supposed to be taking place” are irrelevant].
2) Local realism implies that a classical computer model must (must!) be possible.
3) Large Bell violation is achieved already by merely considering only the three angles 0, Pi/8, and 3Pi/8. The two choices in eq. (4) of arXiv:1103:1879 [b_j b_k = – d_jk +/– e_jkl b_l] randomizes the basis somewhat. Lets construct an entirely new basis for every EPR pair and send the information along with the photons. Still, such is all trivial compared to what can be stored and calculated with an average desktop computer in mere milliseconds. [What do you mean by “If you are able to program the mathematics described in equation (4) …”? Surely you agree that this real quaternion equation can be put into mathematica for example by any undergraduate.]

Given these three aspects and given that several people failed for years to come up with even the simplest locally realistic Bell violation, your claim of locality is highly suspect. Do not get me wrong; I am not trying to attack you, but merely wonder whether you fully appreciate that these facts put the ball squarely into your side of the court.

Hand waving about profoundness over other’s “simple minded” variables invites suspicions about skirting around the issue – regardless of whether you like it or not – this is the impression easily taken away. For example: “What has to be done is program equations (1) and (2) of the attached paper in such a manner that the fixed and random bivector basis defined by equations (3) and (4) in it are duly respected.” Let me translate this into what it kind of sounds like: “My claim can be seen as clearly correct once somebody comes up with my key missing details in the key equations in such a manner that everything works out (which however is precisely what cannot happen in case the claim is wrong!).

[Comment: The following is the program that somebody internet savvy should actually write as a java applet or whatever and put out there as part of the Quantum Crackpot Randi Challenge! I personally only work with research software that is not so internet friendly.]
So let me suggest again: I can (almost any student can) write a program that fully simulates the experiment with the singlet EPR photon pairs and the three angles. It would have two modes, “QM”and “Classical”. In QM mode, the violation of Bell’s inequality is assured by non-locality (i.e. communication send from Alice to Bob if Alice measures first and vice versa). This communication is switched off in Classical mode, where instead “simpleminded” hidden variables are distributed to Alice and Bob, which, as you agree, cannot violate Bell’s inequality. The correct QM statistics would not result. This would be a very simple program as far as computer programs go. 1000 to 10000 EPR pairs can be simulated in a few seconds; good statistics is no problem.

The implicitly assumed topology, namely Euclidian R-cubed, is reflected in the structure of the hidden variables and the way Alice and Bob pick the three angles. Say, for example, that I wanted to introduce that the outcome of Alice rotating her crystal by 360 degrees (instead of twice around by a full 4Pi) must count as a separate axis because of the double covering [SO(3)versus SU(2) issue] resolved by Fermions. No problem: a week maximum and mode “Double Cover” is open source in the internet!

In other words: You claim that the only differences are well defined like for example the codomain topology being S3 instead of “naïve” R-cubed. If so, simply take the program and modify a few relations in order for your point being proven. If you cannot, it will be understood as a clear indication of that your central claim of locality is wrong.
[Comment: This is the point of my posting the “Quantum Crackpot Randi Challenge” and why you are all sincerely invited to put the program as explained above onto the internet. From now on, anybody with some crackpot “local QM” theory is cordially invited to either modify that program so that Bell’s inequality is violated or to shut the hell up! (Just like the traditional Randi Challenge!)]

You would be only able to counter this severe attack by explaining clearly why it is that there should be any problem in changing the implicitly assumed topology in the hidden classical data structure that previously simulates a simple physical experiment perfectly. For example: clearly show how the six directions (instead of three because of double cover) plus two handedness choices lead to an infinite amount of data on all of S3 in such a finely fractal mapping that no finite resolution via digital computer memory can carry it along and no floating point precision would be able to calculate it locally.

You spend a number of years on this now. If it is possible at all, it must be possible with that simple example! I fit is possible at all, we should be able to make it work in a few months maximum for that particular example. You know very well, if even just this one single simple example works to refute non-locality, we can reserve plane tickets to Sweden. Lets get started.



Joy’s Answer:

Hi Sascha,

I like your passion, and I agree with most of what you are saying. My latest paper is only two months old, and no one has tried to simulate the model described in it so far. The previous model with "mu" was tried by several people, but I cannot be sure whether they did things correctly, because I myself know nothing about computers. I cannot write even one line of code.

The model is quite simple. Alice chooses a direction, say a, and calculates a fixed bivector using the fixed basis defined by equation (3). Then she waits until she receives the hidden variable lambda, and then calculates the random bivector using the random basis defined by equation (4). She then takes the product of the fixed bivector and the random bivector, as defined in equation (1). The product gives her a number, +/-1. Bob follows exactly the same procedure for another direction, say b. For a large sample of random bivectors the results of the products obtained by Alice and Bob are compared, and the correlations are calculated.That is all there is to it. I claim that if this procedure (i.e., this algorithm) is followed correctly, you will see violations of BI locally.


17 May 2011 14:51 [Comment: Here I basically just show that his mathematics is trivial, blown up nonsense.]

Let me see whether I understand you correctly and am not missing something vital:

You say: Alice chooses vector a, and calculates a fixed bivector using the fixed basis.

Practical: a = (0, 1, 0) and the fixed bivector {a_j beta_j} of the fixed basis{beta_1, beta_2, beta_3} is thus beta_2.

You say: She receives lambda, and then calculates the random bivector using the random basis and beta_j beta_k = – delta_jk – lambda epsilon_jkl beta_l.

Practical: lambda = +/–1, thus the random basis is {+/– beta_1, +/– beta_2, +/–beta_3} and the random bivector is +/– beta_2. Not all that random; merely two possible and simple cases.

You say: She then takes the product of the fixed bivector and the random bivector,as defined in equation (1).

Practical: There is a further minus in equation (1) inside {– a_j beta_j} which I did not count towards the fixed bivector {a_j beta_j} in the above, and with this further minus sign the equation works out to be A = {– beta_2} {+/– beta_2} =–/+ (– delta_22) = +/– 1 exactly as you claim.

You say: Bob follows exactly the same procedure for a direction b.

Practical: b = (0, cos(Pi/8), sin(Pi/8)) and the fixed bivector {b_k beta_k} is calculated using the same (“exactly the same procedure”) fixed basis {beta_1, beta_2, beta_3}.
So, Bob’s fixed bivector is cos(Pi/8) beta_2 + sin(Pi/8) beta_3 (or did I somehow mess up with the epsilon? Please correct this into the notation you would prefer so we do not talk past each other).

His random bivector is something like +/– {b_k beta_k} and thus B = +/– (cos cos beta_2 beta_2 + cos sin …. – sin cos … + sin sin beta_3 beta_3) = ...(please correct if I am totally wrong)
Which together with cos cos + sin sin = 1 results in B = –/+ 1, exactly as you claim.

Surely I did something wrong, because I should not get exactly equation (2), as it would imply that Bob gets every time the exact opposite answer of Alice’s measurement even when the angle is Pi/8. That is not what is experimentally observed. Nevertheless, even if there is some epsilon and cos sin factors I got wrong, we surely agree on that the resulting B is obviously completely determined by lambda, and that it is either plus or minus one.

The big problem I have with this is that it allows only two combined outcomes, namely either [A=1 and B=B(b, lambda = +1) with the b as given], or the only other possible outcome, which is [A= –1 and B(b, lambda = – 1)]. Whatever that B really happens to be in the above calculation, there are only two different cases! This is clearly wrong since if Alice and Bob actually do this experiment at these angles, they will get all four possibilities [A=1, B=1], [A=1, B= –1],
[A= –1, B=1], and [A= –1, B= –1].

Something here is very wrong. You either mean a different, non-naïve lambda (as it is currently in your last paper, it really is just plus or minus 1), or you mean something like that the fixed basis is actually also randomly chosen every time freshly again before the next photon arrives, either by the EPR pair or even by Alice and Bob randomly and independently (locally) every time (yet without them realizing that this is what they actually effectively do).

[Comment: Here in these last three sections I am trying very hard to give him plenty of opportunities to save his face with some sort of copout about that I got the true complexity of the hidden variables wrong or whatever. This is to trigger him to either start mumbo-jumbo pseudo-science talk right now or admit that I understand his maths perfectly (it is trivial after all).]

I would be very grateful indeed if you could clear up my difficulties here,




Joy’s Answer:

Hi Sascha,

You are doing nothing wrong in your calculations. [Comment: That was a pretty easy catch. Now how will he claim that I do not understand the maths? Sure he will anyways!] You are interpreting some things incorrectly however. Equations (1) and (2) correspond to independent observations of Alice and Bob. It does not matter what angles Alice and Bob choose. For all angles their results are exactly opposite, and therefore the product of their result is always -1. You are of course right to worry that this is not what we observe in experiments. [Comment: Closed the trapdoor behind him and started frying himself. Thanks. This is basically where you can stop reading. From here on it just goes deeper into the crackpot sewer.] Remember, however, that Alice and Bob are not aware of each other. It is only after all of the data is collected and the results are compared we see the correlations. Alice and Bob themselves only see random outcomes, +1 or -1, for all possible angles.

To understand the correlations and the occurrences of all four possible joint outcomes,+ +, + –, – +, and + +, let me first explain the physical meaning of the bivectors appearing in equations (1) and (2).

The random bivectors are supposed represent the physical spin:

      { a_k beta_k (lambda) } = spin "up" about the direction a if lambda = +1
                                            = spin "down" about the direction a if lambda = -1

      { b_j beta_j (lambda) } = spin "up" about the direction b if lambda = +1
                                          =  spin "down" about the direction b if lambda = -1

The fixed bivectors, {- a_j beta_j }, on the other hand, are supposed to represent
the measuring devices that measure the random bivectors (spins). In other words,

            spin { a_j beta_j (lambda) }  is measured with respect to {-- a_j beta_j }

    and  spin { b_j beta_j (lambda) } is measured with respect to {+ b_k beta_k }

The crucial point here is that Alice's spin and Bob's spin are measured with respect to two different measuring devices, {-- a_j beta_j } and {+ b_k beta_k}.

Suppose now we set b = a. Then the two spins will be measured by two oppositely aligned measuring devices; that is to say, the two measuring devices will differ by a minus sign, and we will get the perfect anti-correlation ( -1 ), as your calculations confirm. This corresponds to the cases + – and  – +.

Next, suppose we set b = -- a for the two measuring devices. [Comment: See here how he in typical crackpot manner goes back to what is long since dealt with. The “Quantum Crackpot Randi Challenge” is specifically about only the three angles including Pi/8, this is 100% clear since email number one. He insists on talking about angles that show no Bell violation anyway, states that the model is correct there (trivially true), and then claims to have disproved Bell’s argument. Smooth performance of a classical crackpot move.] Then it is clear from the above that the two spins will be measured by "the same" device {-- a_j beta_j } (they are of course still two separate devices at the two opposite end of the experiment, but they will be mathematically exactly the same). Therefore in this case the two spins will be the same, and we will have 50/50 chance of either + + or – – , giving the product +1.

What these two cases show is that the joint outcomes of spins depend on the measuring devices used by Alice and Bob. There is nothing nonlocal going on here. It is just that one must use the same scales on both sides to make any meaningful comparison of the two sets of measurement results (clearly, one cannot use the same amount of Dollars and Euros to buy the same number of bananas). One must measure the spins with the same "scale" on both sides,and with the same "units."

So how does all these affect the correlations? Well, correlations are in general defined as covariance of the measurement results divided by the product of two standard deviations, and these standard deviations take care of the "scaling" problem. [Comment: Super characteristic crackpot gobbledygook. Only missing is some infinite fractal singularity or so. He is desperately trying to make me forget something. But what?] You can check that equations (5) of my paper satisfies both the perfect anti-correlation condition for 0 degrees and the perfect correlation conditions for 180 degrees as we vary the measurement axes from 0degrees separation to 180 separation. [Comment: Oh, right, that his “model” cannot even model the one angle Pi/8. Sorry, but I did not forget that all my emails are precisely not about these 0 and 180 degrees angles as they do not violate the Bell inequality anyway.]

So theoretically everything works out fine.

Now you may wonder how to deal with the scaling problem in your simulation. Well, my guess is that somehow you will have to make sure that your program keeps the fixed bivectors (the measuring devices) completely fixed, and the randomness is introduced only in the random bivectors (i.e., the "spins"). Then, when the results from the two ends are compared, the distribution will be normal distribution in the units of standard deviations, which are the measuring devices, and that should take care of the "scaling" problem.



20 May 2011 02:40

“For all angles their results are exactly opposite. You are of course right to worry that this is not what we observe in experiments. Remember, however, that Alice and Bob are not aware of each other. It is only after all of the data is collected and the results are compared we see the correlations.”

Alice and Bob writing down all the outcomes and afterward comparing; that IS the experiment! Analyzing the data, they will find all outcomes, + +, + –, – +, and + + if the relative angle between directions a and b is Pi/8. What are you trying to say? That the recorded data change inside Bob’s lab-log while he drives to Alice’s place?

[Comment: So, by now you see I am getting a little impatient. How is it that he has still not caught on to what I think about his model?]



Joy’s Answer:

"What are you trying to say? That the recorded data change inside Bob’s lab-log while he drives to Alice’s place?"

Yes! In a sense, that is exactly what is happening. But not in a mysterious way. It is simply a matter of rescaling. Clearly, Alice and Bob must use the same scale to analyze the two sets of data for their comparison to be meaningful.

What I am saying is nothing new. This is how any statistical distribution of numbers is compared. What is new in my model is that the statistical analysis must also respect the topology of the 3-sphere, because that is the topology of the experiment. When this is done correctly, all four outcomes, + +, + –, – +, and – – , appear correctly. There is no mystery here --- only the counter-intuitiveness of the topology of the 3-sphere.

-- Joy

23 May 2011 11:44

I see. Scaling – the non-mysterious kind. Well that explains it. So, the angle is Pi/8 and Alice’s lab book says “Photon 1: +1, Photon 2: +1, Photon 3: -1, Photon 4: +1, …” and according to your local realism, Bob’s lab log shows: “Photon 1: -1, Photon 2: -1, Photon 3: +1, Photon 4: -1, …”

Could you shortly explain how non-mysterious re-scaling while Bob drives to Alice’s place leads to Bob’s lab log changing to read instead “Photon 1: -1, Photon 2: -1, Photon 3: -1, Photon 4: +1, …” or whatever once he arrived and compares with Alice?




Enough brown semi solids. In good old crackpot manner, he just keeps on writing mumbo-jumbo about “scaling” and the “counter-intuitiveness of the topology of the 3-sphere” without ever even once acknowledging the question asked, which is clearly for at least three of my mails now one and only one: What ghosts come flying down from the heaven of Crackpotonia to rewrite the measurement results and Bob’s and Alice’s memories and lab-logs so that his crackpot model’s total anticorrelation is “non-mysteriously” changed into what quantum physics is well known to result in?

The further emails only drove home the already obvious: He does plainly not accept the experimental evidence! But instead of telling people this honestly, he keeps stating that his model can reproduce all the data. This is not an inch better than creationists’ “arguing”: totally dishonest.

He deceives people with a pseudo profound but basically trivial model that supposedly “explains everything”, including what is observed in experiments, while in fact he does not accept the experimental observations and thereby excludes himself from the scientific community, for which scientific experiment is the ultimate judge.

The guy is funded by "Foundational Questions Institute (FQXi)", which funds loads of pseudo-science, so maybe no surprise there. But the big question left is: What is such a clearly obnoxious fraud doing at Oxford University and Perimeter Institute?


By the way: I have NO intention of wasting my time “discussing” with crackpots. Put up or shut up! Means what? Means: Solve the simple “Quantum Crackpot Randi Challenge”, then we can talk.