The headline is, as popular articles on QM often are, misleading. One of the central mysteries of QM has always been: how does the classical world with its arrow of time emerge from the quantum world where the governing equation is time-symmetric? And the answer is: the classical world emerges by a process of decoherence, which is to say, by the creation of large (O(10^23)) networks of entanglements which (it can be shown mathematically) have behavior that is indistinguishable from classical systems. It is very similar to how thermodynamics and the time-irreversibility of the second law emerge from time-reversible Newtonian mechanics (except that the time-asymmetry that emerges from QM is even more fundamental).
Time-reversibility is part-and-parcel of quantum computation because the whole point of quantum computation is to keep the system operating in the regime where its behavior is distinguishable from classical. As soon as you lose coherence, you simultaneously lose all of the quantum behavior, including time-reversibility. Time-reversibility isn't anything special in this regard. It's all part of the same weirdness.
It's all very interesting, but none of it is news.
> And the answer is: the classical world emerges by a process of decoherence, which is to say, by the creation of large (O(10^23)) networks of entanglements which (it can be shown mathematically) have behavior that is indistinguishable from classical systems.
This is the motivation for decoherence and for the Everettian view where the psi function is "the only real thing" and everything else is derived. It is often talked about as if actually accomplished. But it was not; the behaviour of quantum systems, even with decoherence, was not yet shown to be indistinguishable from the classical systems. Neither the fact that measurements have one outcome, nor Born's rules for probability were derived from this decoherence idea yet. One thing that follows from decoherence hypothesis is decay of coherence elements in density matrices, but this is not enough to get explanation for single outcomes or for Born's rules.
Ironically, I am just in the process of preparing a talk on the history of the atomic theory. Even as late as 1905 when Einstein published his work on brownian motion it was still highly controversial whether atoms were "really real" or just a mathematical contrivance. Some people are able to maintain skepticism even in the face of overwhelming evidence.
> Neither the fact that measurements have one outcome, nor Born's rules for probability were derived from this decoherence idea yet.
That is because it is not the case that measurements have one outcome. Measurements appear to have one outcome from the perspective of a given classical observer, and the outcome observed by any given observer will be classically consistent with all the other observations that observer makes [1]. The Born rule (which, it must be stressed, is not fundamental, but only apparent from the point of view of any given observer) follows from that, not mathematically but logically [2].
> "That is because it is not the case that measurements have one outcome. Measurements appear to have one outcome from the perspective of a given classical observer [but actually have none]"
is a big claim, but it lacks "overwhelming evidence". Quantum theory self-consistency is a problem worthy of investigation, but achieving self-consistency is only one of the many possible valid results. This result is not needed as much as many people seem to think; certainly not as much as to reify psi function above objective results of measurements and claim the quoted part. It is a methodological error; instead of shaping the theory around the measurement results, it is rejecting the fact that those measurement results exist.
> The Born rule (which, it must be stressed, is not fundamental, but only apparent from the point of view of any given observer) follows from that, not mathematically but logically
I skimmed the paper [2] but found no such thing there. Could you locate it in more precisely or elaborate on the idea? All these kinds of proofs from algebraic formalism seem to achieve is something like "if Born's rule would not be true, we would have problems with the theory" which is not very interesting; we have some big problems with the theory even without decoherence.
> it is rejecting the fact that those measurement results exist
Well, sort of. I prefer to think of it as replacing one explanatory hypothesis with a superior one, one that is actually consistent with the evidence. To wit:
The observation that requires explaining is that classical measurements are consistent across space and time.
The usual explanation for this is that a classical objective reality exists, and that measurement faithfully reflects the state of this reality. This is a plausible explanation. Einstein believed it. But Bell's theorem and subsequent experiments show it to be false.
The correct explanation is that classical correlation emerges from the Shroedinger equation alone. There is no objective classical reality. If there were, it would be possible to construct a local hidden variable theory consistent with all observations. But it's not so there isn't.
It is only because our subjective experience is (necessarily) classical that we find all this hard to accept.
> I skimmed the paper [2] but found no such thing there. Could you locate it in more precisely or elaborate on the idea?
> But Bell's theorem and subsequent experiments show it to be false.
I think this is an unfounded conclusion. Sure, Bell's theorem and the experiments it inspired did make lots of contribution to investigation of these foundational problems.
But Bell's theorem is about a specific class of probabilistic descriptions. It does not rule out "classical objective reality" unless you restrict meaning of this term so much as to achieve that. Bell himself and many other people who studied the theorem insisted that the theorem is about local hidden variable models, as opposed to nonlocal models. Nothing general is implied about "objective reality".
On the experimental side, I am not aware of any experiment that had ruled out objective reality. How would you even test such a nebulous concept experimentally? We can test only specific predictions.
> Nothing general is implied about "objective reality".
I disagree. The fact that nature cannot be described by any local hidden variable theory tells you a lot about the nature of reality. There is a real difference between quantum randomness and classical ignorance, and so there is a real difference -- one which you can experimentally measure -- between a particle whose position you don't know because you haven't looked at the result of a measurement, and a particle whose position you don't know because it hasn't been measured at all. It is still, perhaps, debatable whether that difference lies in the particle or in you. But if you want to argue that the difference lies in the particle then the burden is on you to provide an account of when and how the transition happens. No one has succeeded in this, and it's not for want of trying.
> The fact that nature cannot be described by any local hidden variable theory tells you a lot about the nature of reality.
If assumed to be true, it means one has to use non-local models such as the quantum theory. Those do not explicitly deny objective reality. Perhaps by "objective reality" you actually mean non-contextual hidden variable theory (a theory that assigns well-defined values to all quantum observables at all times). Then yes, those are incompatible with quantum theory predictions. But I would not use the term objective reality. It is unclear and too general.
> there is a real difference -- one which you can experimentally measure -- between a particle whose position you don't know because you haven't looked at the result of a measurement, and a particle whose position you don't know because it hasn't been measured at all.
Indeed, but this difference has nothing to do with objective reality, and everything to do with the actions of the experimenter or apparatus. In the first option, there is interaction with the particle but experimenter ignores it, and in the second, there is no interaction with the particle. Of course the two situations are different and can result in different results of other measurements.
I am not denying objective reality, I am denying objective classical reality. There is an objective reality, but it is quantum.
> the actions of the experimenter or apparatus
You are missing the point. Before you can talk about "the actions of the experimenter or the apparatus" you have to tell me what an experimenter or an apparatus is. AFAICT, these are classical objects. But theory tells us, and experiment confirms, that classical objects do not actually exist. So why do they appear to exist? And the answer to that is: decoherence/QIT. The classical world is an illusion, an approximation. A very good approximation, but an approximation nonetheless.
If you have a better theory, you should publish it.
> Before you can talk about "the actions of the experimenter or the apparatus" you have to tell me what an experimenter or an apparatus is.
This is a very common error of armchair worldview building. Experiment-based sciences such as physics do not work exclusively that way; we can make progress without being so ambitious as to describe and exhaust the whole world in a single scheme. There are things in the real world that have no useful definition or model in physics, such as matter, or experimenter, or apparatus, or measurement. And this is fine, because we know them by experience, and the subject of the investigation is usually something else and rather more specific.
> But theory tells us, and experiment confirms, that classical objects do not actually exist.
Which theory, which experiment? The studies in quantum foundations showed that explaining expensive experiments with light and particles in the classical terms is hard, and some naive intuitive models, surprisingly, are not consistent with quantum theory. This is very far from "classical objects do not exist". They can easily exist, just with properties that are contextual and interactions that are non-local. And those are categories we are aware of; the possibilities are endless.
If you're thinking about the Everettian viewpoint where indeed no classical objects exist, this is just one possible theoretical scheme of thinking, not an experiment-based physical theory. There is Bohmian mechanics, where particles do exist. There are other theories where particles do exist.
> If you have a better theory, you should publish it.
A better theory in place of "classical objects don't exist" is "don't try to explain the whole world with a single scheme". There is the statistical interpretation due to Einstein, Ballentine and the Bohmian viewpoint which get us useful models and predictions in the realm of atomic physics. There is the classical theory, which gets us understanding of civil engineering, Earth-scale events and celestial mechanics. There is the general theory of relativity, which gets us accurate description of gravity. None of these overlap very well. Neither is explained by Everettian viewpoint.
> Experiment-based sciences such as physics do not work exclusively that way; we can make progress without being so ambitious as to describe and exhaust the whole world in a single scheme.
Nope The object of the game in the physical sciences is reductionism. Reducing the number of explanations required to account for observations is the definition of progress.
> "don't try to explain the whole world with a single scheme"
That's not a theory. That's throwing in the towel.
Thanks. But I would actually not say that QM "allows retrocausality." Instead I would say that the whole concept of "causality" simply meaningless in a time-symmetric system, just as the concept of (say) "up" is meaningless in a spatially symmetric system.
It’s so funny that quantum physicists chose the term “retrocausality” because the literal meaning of its parts (“ex post facto”) can only be understood in the context of the “arrow of time”.
Perhaps wordsmithing that alludes to a superordinate context (beyond time) would be more appropriate..
Microsoft told me that any operation a quantum computer performs HAS to be reversible: something like XOR doesn't work because there're four possible inputs and only two possible outputs, so the state of the inputs is destroyed. However, something like NOT works fine because the input-output mapping is 1:1.
It seems like cheating to say that "reversing cause and effect is no trouble" when the computer was never allowed to perform irreversible operations to begin with. I think I am missing some important information or maybe I am misunderstanding the words used in the article?
Note: I'm not a professional quantum scientist so if any real ones come along please tell me if I'm getting this right, and please correct any mistakes that you see in my logic.
From my understanding so far:
One benefit of using qubits is that you can map multiple of them to one traditional bit.
Let's say we use light for our quantum computer. I like to see a wave of light as a continuous stream of photons, and photons as "an order". I'm assuming all waves and photons in the rest of this comment are of the same frequency.
If you have a vertically polarised wave of light, it orders you this:
"at time 1: jump up. at time 2: jump down. at time 3: jump up, ..."
If you have a horizontally polarised photon, it orders you this:
"at time 1: move left. at time 2: move right. at time 3: move left, ..."
We can start by saying "horizontal polarization is 0" and "vertical polarization is 1".
H = 0
V = 1
If we define XOR using only H and V:
H ^ H = H (0)
H ^ V = V (1)
V ^ H = V (1)
V ^ V = H (0)
Let's now add phase to the qubits to enhance them. Phase can be seen as adding a time delay (or "shift") to all of the photons.
Let's see the vertically polarized photon again:
"at time 1: jump up. at time 2: jump down. at time 3: jump up, ..."
And at 180 degree phase, you shift all orders by 1 time unit:
"at time 1: jump down. at time 2: jump up. at time 3: jump down, ..."
One fun and useful property of phase is that if you have two photons of the same polarization but they are at phases 180 degrees apart from each other, they will cancel each other out because you cannot move up and down at the same time.
Getting back to qubits and XOR:
Let's say H at 180 phase is I, and V at 180 phase is W.
Note that I is still horizontally polarized, and W is still vertically polarized. Let's bring them together, where we have 4 types of qubits mapped to 2 types of bits:
H = 0
I = 0
V = 1
W = 1
Let's add a step to the XOR operation:
- if the right input is V, phase shift the output by 180 degrees
The outputs of our "XOR 2.0: quantum edition":
H ^ H = H, no shift = H (0)
H ^ V = V, shift it = W (1)
V ^ H = V, no shift = V (1)
V ^ V = H, shift it = I (0)
Now we have output that still matches the original definition of XOR when mapped to traditional bits, but with the bonus of reversibility.
You have the right idea, but not the right operation. XOR is reversible.
To give a better example: COPY is not reversible. Imagine that you have two qubits [A, B] as an input and you want to output [A, A]. That is not possible with quantum computers.
Isn't everything reversible if enough information is stored? Instead of COPY [A, B] -> [A, A]; you could have [A, B, 0] -> [A, A, B] which saves the value of B.
Yes, but "keeping the inputs" is non-trivial in quantum-land because of the no-cloning theorem. You can't just make a copy of your inputs and stash them. You have to be more clever than that.
Forgive me for this being off topic: if you use a script blocker this site will blow up your CPU unless you have ajax.googleapis.com allowed. My browser became completely unresponsive. When I checked dev. console there were 50,000+ errors (and quickly climbing) on a video JS import. (2018-07-reversing-effect-quantum.html:1314). Pretty buggy.
If quantum computers can maintain a viewpoint where cause and effect don't matter, it seems like they will be at a point very different from the human viewpoint - where a person inherently has more difficulty predicting the future than determining the past.
It hasn’t got anything to do with that, what the exceedingly clickbaity headline is actually talking about is that QCs are more able to reverse computation.
I’m unclear /exactly/ what they are trying to say as all QC gates are required to be reversible, but it seems that they’re trying to say that to make a computation reversible on a classical computer you in general (eg worst case) need to store all the state (massive memory cost) or recompute from the beginning, or somewhere in between (eg when skipping back in a video, a player is generally going to compute relative to the nearest iframe).
In QC this extra memory isn’t needed as every computation /must/ be reversible so you just run the program backwards, and rely on the increased state that can be stored in each qubit (a lot of the exponential => polynomial performance improvements basically seem to boil down to each individual qubit being able to represent exponentially more state than a single bit).
Maybe that’s helpful? I’m not super sure as the places where QC is faster than classical computation is pretty much only places where you can trade off exponential time for exponential state. E.g. problems that you can’t plausibly use a classical computer in the first place.
Time-reversibility is part-and-parcel of quantum computation because the whole point of quantum computation is to keep the system operating in the regime where its behavior is distinguishable from classical. As soon as you lose coherence, you simultaneously lose all of the quantum behavior, including time-reversibility. Time-reversibility isn't anything special in this regard. It's all part of the same weirdness.
It's all very interesting, but none of it is news.