So I read the paper, and it is kind of a cool experiment, but it does not show that “future choices can affect a past measurement’s outcome.” Explaining why would require a separate article (maybe time to re-open main!) But the TL;DR version is this: if you want to argue that A affects B then you have to show a causal relationship that runs from A to B. If you can do that, then you can always come up with some encoding that will allow you to transmit information from A to B. That’s what “causal relationship” means. But that is (unsurprisingly) not what Aharonov et al. have done. They have merely shown correlations between A and B, and then argue on purely intuitive grounds that there must have been some causal relationship between A and B because “Bell’s theorem forbids spin values to exist prior to the choice of the orientation measured.” While this is true, it’s misleading because it implies that spin values do exist after a strong measurement. But that is not true. There is no fundamental difference between a strong and a weak measurement. There is a smooth continuum between weak and strong measurements, and at no point during the transition from weak to strong does the spin value begin to “actually exist” (a.k.a. wavefunction collapse).
I disagree. Following Pearl, I define “A causes B” to mean something like: (DO:A) raises the probability of B.
Bob’s choice in the evening to make strong measurements along the beta-axis, raises the probability of Alice’s noon measurements along the beta-axis measurements having been the ones that showed the best correlation. It doesn’t raise the probability of any individual measurement being up or down, but that’s OK. Even on a many worlds interpretation, where perhaps every digital up/down pattern happens at some “world” and the overall multi-world distribution is invariant, “probability” refers to what happens in our “world”, so again that’s OK.
Correlation can only be observed after the fact, in the evening, not at noon. So isn’t this just a case of Bob affecting Bob+Alice’s immediate future, where they go over the results? Why do I say Bob’s choice affected Alice’s results? Because correlation is a two-way street, and in this case there isn’t much traffic in the forward direction. Alice’s measurements only weakly affect Bob’s results.
Yes, but there’s still some terminological sleight-of-hand going on here. It is only fair to say that a future A affected a past B if P(B) is well defined without reference to A. In this case it’s not. Because B is defined in terms of correlations between measurements made at T1 (noon) and measurements made at T2 (evening) then B cannot be said to have actually happened until T2.
correlation is a two-way street
No, it’s an n-squared-minus-one-way street. It appears to be a two-way street in one (very common) special case (two macroscopic systems mutually entangled with each other), but weak measurements are interesting precisely because they do not conform to the conditions of that special case. When you go beyond the conditions of the common special case you can’t keep using the rhetoric and intuitions that apply only to the special case and hope to come up with the right answer.
Yeah, probably. It’s actually probably N!-1 because you have to trace over one degree of freedom to obtain a classical universe. But the details don’t really matter. What matters is that it’s >>N.
I disagree: if you interpret EPR experiments as wavefunction collapse rather than many worlds, then you can conclude that either one measurement affects the other, or both affect each other. But you cannot come up with any encoding that will allow you to transmit information.
Yes, of course that’s true. But collapse is only an approximation to the truth. It is a very good approximation in many common cases. But the Aharonov experiment is interesting precisely because it is a case where collapse is no longer a good approximation to the truth, and so of course if you view it through the lens of collapse things are going to look weird. To see why collapse is not always a good approximation to the truth, see the references in the OP.
So I read the paper, and it is kind of a cool experiment, but it does not show that “future choices can affect a past measurement’s outcome.” Explaining why would require a separate article (maybe time to re-open main!) But the TL;DR version is this: if you want to argue that A affects B then you have to show a causal relationship that runs from A to B. If you can do that, then you can always come up with some encoding that will allow you to transmit information from A to B. That’s what “causal relationship” means. But that is (unsurprisingly) not what Aharonov et al. have done. They have merely shown correlations between A and B, and then argue on purely intuitive grounds that there must have been some causal relationship between A and B because “Bell’s theorem forbids spin values to exist prior to the choice of the orientation measured.” While this is true, it’s misleading because it implies that spin values do exist after a strong measurement. But that is not true. There is no fundamental difference between a strong and a weak measurement. There is a smooth continuum between weak and strong measurements, and at no point during the transition from weak to strong does the spin value begin to “actually exist” (a.k.a. wavefunction collapse).
I disagree. Following Pearl, I define “A causes B” to mean something like: (DO:A) raises the probability of B.
Bob’s choice in the evening to make strong measurements along the beta-axis, raises the probability of Alice’s noon measurements along the beta-axis measurements having been the ones that showed the best correlation. It doesn’t raise the probability of any individual measurement being up or down, but that’s OK. Even on a many worlds interpretation, where perhaps every digital up/down pattern happens at some “world” and the overall multi-world distribution is invariant, “probability” refers to what happens in our “world”, so again that’s OK.
Correlation can only be observed after the fact, in the evening, not at noon. So isn’t this just a case of Bob affecting Bob+Alice’s immediate future, where they go over the results? Why do I say Bob’s choice affected Alice’s results? Because correlation is a two-way street, and in this case there isn’t much traffic in the forward direction. Alice’s measurements only weakly affect Bob’s results.
Yes, but there’s still some terminological sleight-of-hand going on here. It is only fair to say that a future A affected a past B if P(B) is well defined without reference to A. In this case it’s not. Because B is defined in terms of correlations between measurements made at T1 (noon) and measurements made at T2 (evening) then B cannot be said to have actually happened until T2.
No, it’s an n-squared-minus-one-way street. It appears to be a two-way street in one (very common) special case (two macroscopic systems mutually entangled with each other), but weak measurements are interesting precisely because they do not conform to the conditions of that special case. When you go beyond the conditions of the common special case you can’t keep using the rhetoric and intuitions that apply only to the special case and hope to come up with the right answer.
You’re right. Good point.
Don’t you mean n-factorial? Anyway, … hmm, I need to think about this more.
Yeah, probably. It’s actually probably N!-1 because you have to trace over one degree of freedom to obtain a classical universe. But the details don’t really matter. What matters is that it’s >>N.
I disagree: if you interpret EPR experiments as wavefunction collapse rather than many worlds, then you can conclude that either one measurement affects the other, or both affect each other. But you cannot come up with any encoding that will allow you to transmit information.
Yes, of course that’s true. But collapse is only an approximation to the truth. It is a very good approximation in many common cases. But the Aharonov experiment is interesting precisely because it is a case where collapse is no longer a good approximation to the truth, and so of course if you view it through the lens of collapse things are going to look weird. To see why collapse is not always a good approximation to the truth, see the references in the OP.