Archive for September, 2015

Six announcements

Monday, September 21st, 2015
  1. I did a podcast interview with Julia Galef for her series “Rationally Speaking.”  See also here for the transcript (which I read rather than having to listen to myself stutter).  The interview is all about Aumann’s Theorem, and whether rational people can agree to disagree.  It covers a lot of the same ground as my recent post on the same topic, except with less technical detail about agreement theory and more … well, agreement.  At Julia’s suggestion, we’re planning to do a follow-up podcast about the particular intractability of online disagreements.  I feel confident that we’ll solve that problem once and for all.  (Update: Also check out this YouTube video, where Julia offers additional thoughts about what we discussed.)
  2. When Julia asked me to recommend a book at the end of the interview, I picked probably my favorite contemporary novel: The Mind-Body Problem by Rebecca Newberger Goldstein.  Embarrassingly, I hadn’t realized that Rebecca had already been on Julia’s show twice as a guest!  Anyway, one of the thrills of my life over the last year has been to get to know Rebecca a little, as well as her husband, who’s some guy named Steve Pinker.  Like, they both live right here in Boston!  You can talk to them!  I was especially pleased two weeks ago to learn that Rebecca won the National Humanities Medal—as I told Julia, Rebecca Goldstein getting a medal at the White House is the sort of thing I imagine happening in my ideal fantasy world, making it a pleasant surprise that it happened in this one.  Huge congratulations to Rebecca!
  3. The NSA has released probably its most explicit public statement so far about its plans to move to quantum-resistant cryptography.  For more see Bruce Schneier’s Crypto-Gram.  Hat tip for this item goes to reader Ole Aamot, one of the only people I’ve ever encountered whose name alphabetically precedes mine.
  4. Last Tuesday, I got to hear Ayaan Hirsi Ali speak at MIT about her new book, Heretic, and then spend almost an hour talking to students who had come to argue with her.  I found her clear, articulate, and courageous (as I guess one has to be in her line of work, even with armed cops on either side of the lecture hall).  After the shameful decision of Brandeis in caving in to pressure and cancelling Hirsi Ali’s commencement speech, I thought it spoke well of MIT that they let her speak at all.  The bar shouldn’t be that low, but it is.
  5. From far away on the political spectrum, I also heard Noam Chomsky talk last week (my first time hearing him live), about the current state of linguistics.  Much of the talk, it struck me, could have been given in the 1950s with essentially zero change (and I suspect Chomsky would agree), though a few parts of it were newer, such as the speculation that human languages have many of the features they do in order to minimize the amount of computation that the speaker needs to perform.  The talk was full of declarations that there had been no useful work whatsoever on various questions (e.g., about the evolutionary function of language), that they were total mysteries and would perhaps remain total mysteries forever.
  6. Many of you have surely heard by now that Terry Tao solved the Erdös Discrepancy Problem, by showing that for every infinite sequence of heads and tails and every positive integer C, there’s a positive integer k such that, if you look at the subsequence formed by every kth flip, there comes a point where the heads outnumber tails or vice versa by at least C.  This resolves a problem that’s been open for more than 80 years.  For more details, see this post by Timothy Gowers.  Notably, Tao’s proof builds, in part, on a recent Polymath collaborative online effort.  It was a big deal last year when Konev and Lisitsa used a SAT-solver to prove that there’s always a subsequence with discrepancy at least 3; Tao’s result now improves on that bound by ∞.

Bell inequality violation finally done right

Tuesday, September 15th, 2015

A few weeks ago, Hensen et al., of the Delft University of Technology and Barcelona, Spain, put out a paper reporting the first experiment that violates the Bell inequality in a way that closes off the two main loopholes simultaneously: the locality and detection loopholes.  Well, at least with ~96% confidence.  This is big news, not only because of the result itself, but because of the advances in experimental technique needed to achieve it.  Last Friday, two renowned experimentalists—Chris Monroe of U. of Maryland and Jungsang Kim of Duke—visited MIT, and in addition to talking about their own exciting ion-trap work, they did a huge amount to help me understand the new Bell test experiment.  So OK, let me try to explain this.

While some people like to make it more complicated, the Bell inequality is the following statement. Alice and Bob are cooperating with each other to win a certain game (the “CHSH game“) with the highest possible probability. They can agree on a strategy and share information and particles in advance, but then they can’t communicate once the game starts. Alice gets a uniform random bit x, and Bob gets a uniform random bit y (independent of x).  Their goal is to output bits, a and b respectively, such that a XOR b = x AND y: in other words, such that a and b are different if and only if x and y are both 1.  The Bell inequality says that, in any universe that satisfies the property of local realism, no matter which strategy they use, Alice and Bob can win the game at most 75% of the time (for example, by always outputting a=b=0).

What does local realism mean?  It means that, after she receives her input x, any experiment Alice can perform in her lab has a definite result that might depend on x, on the state of her lab, and on whatever information she pre-shared with Bob, but at any rate, not on Bob’s input y.  If you like: a=a(x,w) is a function of x and of the information w available before the game started, but is not a function of y.  Likewise, b=b(y,w) is a function of y and w, but not of x.  Perhaps the best way to explain local realism is that it’s the thing you believe in, if you believe all the physicists babbling about “quantum entanglement” just missed something completely obvious.  Clearly, at the moment two “entangled” particles are created, but before they separate, one of them flips a tiny coin and then says to the other, “listen, if anyone asks, I’ll be spinning up and you’ll be spinning down.”  Then the naïve, doofus physicists measure one particle, find it spinning down, and wonder how the other particle instantly “knows” to be spinning up—oooh, spooky! mysterious!  Anyway, if that’s how you think it has to work, then you believe in local realism, and you must predict that Alice and Bob can win the CHSH game with probability at most 3/4.

What Bell observed in 1964 is that, even though quantum mechanics doesn’t let Alice send a signal to Bob (or vice versa) faster than the speed of light, it still makes a prediction about the CHSH game that conflicts with local realism.  (And thus, quantum mechanics exhibits what one might not have realized beforehand was even a logical possibility: it doesn’t allow communication faster than light, but simulating the predictions of quantum mechanics in a classical universe would require faster-than-light communication.)  In particular, if Alice and Bob share entangled qubits, say $$\frac{\left| 00 \right\rangle + \left| 11 \right\rangle}{\sqrt{2}},$$ then there’s a simple protocol that lets them violate the Bell inequality, winning the CHSH game ~85% of the time (with probability (1+1/√2)/2 > 3/4).  Starting in the 1970s, people did experiments that vindicated the prediction of quantum mechanics, and falsified local realism—or so the story goes.

The violation of the Bell inequality has a schizophrenic status in physics.  To many of the physicists I know, Nature’s violating the Bell inequality is so trivial and obvious that it’s barely even worth doing the experiment: if people had just understood and believed Bohr and Heisenberg back in 1925, there would’ve been no need for this whole tiresome discussion.  To others, however, the Bell inequality violation remains so unacceptable that some way must be found around it—from casting doubt on the experiments that have been done, to overthrowing basic presuppositions of science (e.g., our own “freedom” to generate random bits x and y to send to Alice and Bob respectively).

For several decades, there was a relatively conservative way out for local realist diehards, and that was to point to “loopholes”: imperfections in the existing experiments which meant that local realism was still theoretically compatible with the results, at least if one was willing to assume a sufficiently strange conspiracy.

Fine, you interject, but surely no one literally believed these little experimental imperfections would be the thing that would rescue local realism?  Not so fast.  Right here, on this blog, I’ve had people point to the loopholes as a reason to accept local realism and reject the reality of quantum entanglement.  See, for example, the numerous comments by Teresa Mendes in my Whether Or Not God Plays Dice, I Do post.  Arguing with Mendes back in 2012, I predicted that the two main loopholes would both be closed in a single experiment—and not merely eventually, but in, like, a decade.  I was wrong: achieving this milestone took only a few years.

Before going further, let’s understand what the two main loopholes are (or rather, were).

The locality loophole arises because the measuring process takes time and Alice and Bob are not infinitely far apart.  Thus, suppose that, the instant Alice starts measuring her particle, a secret signal starts flying toward Bob’s particle at the speed of light, revealing her choice of measurement setting (i.e., the value of x).  Likewise, the instant Bob starts measuring his particle, his doing so sends a secret signal flying toward Alice’s particle, revealing the value of y.  By the time the measurements are finished, a few microseconds later, there’s been plenty of time for the two particles to coordinate their responses to the measurements, despite being “classical under the hood.”

Meanwhile, the detection loophole arises because in practice, measurements of entangled particles—especially of photons—don’t always succeed in finding the particles, let alone ascertaining their properties.  So one needs to select those runs of the experiment where Alice and Bob both find the particles, and discard all the “bad” runs where they don’t.  This by itself wouldn’t be a problem, if not for the fact that the very same measurement that reveals whether the particles are there, is also the one that “counts” (i.e., where Alice and Bob feed x and y and get out a and b)!

To someone with a conspiratorial mind, this opens up the possibility that the measurement’s success or failure is somehow correlated with its result, in a way that could violate the Bell inequality despite there being no real entanglement.  To illustrate, suppose that at the instant they’re created, one entangled particle says to the other: “listen, if Alice measures me in the x=0 basis, I’ll give the a=1 result.  If Bob measures you in the y=1 basis, you give the b=1 result.  In any other case, we’ll just evade detection and count this run as a loss.”  In such a case, Alice and Bob will win the game with certainty, whenever it gets played at all—but that’s only because of the particles’ freedom to choose which rounds will count.  Indeed, by randomly varying their “acceptable” x and y values from one round to the next, the particles can even make it look like x and y have no effect on the probability of a round’s succeeding.

Until a month ago, the state-of-the-art was that there were experiments that closed the locality loophole, and other experiments that closed the detection loophole, but there was no single experiment that closed both of them.

To close the locality loophole, “all you need” is a fast enough measurement on photons that are far enough apart.  That way, even if the vast Einsteinian conspiracy is trying to send signals between Alice’s and Bob’s particles at the speed of light, to coordinate the answers classically, the whole experiment will be done before the signals can possibly have reached their destinations.  Admittedly, as Nicolas Gisin once pointed out to me, there’s a philosophical difficulty in defining what we mean by the experiment being “done.”  To some purists, a Bell experiment might only be “done” once the results (i.e., the values of a and b) are registered in human experimenters’ brains!  And given the slowness of human reaction times, this might imply that a real Bell experiment ought to be carried out with astronauts on faraway space stations, or with Alice on the moon and Bob on earth (which, OK, would be cool).  If we’re being reasonable, however, we can grant that the experiment is “done” once a and b are safely recorded in classical, macroscopic computer memories—in which case, given the speed of modern computer memories, separating Alice and Bob by half a kilometer can be enough.  And indeed, experiments starting in 1998 (see for example here) have done exactly that; the current record, unless I’m mistaken, is 18 kilometers.  (Update: I was mistaken; it’s 144 kilometers.)  Alas, since these experiments used hard-to-measure photons, they were still open to the detection loophole.

To close the detection loophole, the simplest approach is to use entangled qubits that (unlike photons) are slow and heavy and can be measured with success probability approaching 1.  That’s exactly what various groups did starting in 2001 (see for example here), with trapped ions, superconducting qubits, and other systems.  Alas, given current technology, these sorts of qubits are virtually impossible to move miles apart from each other without decohering them.  So the experiments used qubits that were close together, leaving the locality loophole wide open.

So the problem boils down to: how do you create long-lasting, reliably-measurable entanglement between particles that are very far apart (e.g., in separate labs)?  There are three basic ideas in Hensen et al.’s solution to this problem.

The first idea is to use a hybrid system.  Ultimately, Hensen et al. create entanglement between electron spins in nitrogen vacancy centers in diamond (one of the hottest—or coolest?—experimental quantum information platforms today), in two labs that are about a mile away from each other.  To get these faraway electron spins to talk to each other, they make them communicate via photons.  If you stimulate an electron, it’ll sometimes emit a photon with which it’s entangled.  Very occasionally, the two electrons you care about will even emit photons at the same time.  In those cases, by routing those photons into optical fibers and then measuring the photons, it’s possible to entangle the electrons.

Wait, what?  How does measuring the photons entangle the electrons from whence they came?  This brings us to the second idea, entanglement swapping.  The latter is a famous procedure to create entanglement between two particles A and B that have never interacted, by “merely” entangling A with another particle A’, entangling B with another particle B’, and then performing an entangled measurement on A’ and B’ and conditioning on its result.  To illustrate, consider the state

$$ \frac{\left| 00 \right\rangle + \left| 11 \right\rangle}{\sqrt{2}} \otimes \frac{\left| 00 \right\rangle + \left| 11 \right\rangle}{\sqrt{2}} $$

and now imagine that we project the first and third qubits onto the state $$\frac{\left| 00 \right\rangle + \left| 11 \right\rangle}{\sqrt{2}}.$$

If the measurement succeeds, you can check that we’ll be left with the state $$\frac{\left| 00 \right\rangle + \left| 11 \right\rangle}{\sqrt{2}}$$ in the second and fourth qubits, even though those qubits were not entangled before.

So to recap: these two electron spins, in labs a mile away from each other, both have some probability of producing a photon.  The photons, if produced, are routed to a third site, where if they’re both there, then an entangled measurement on both of them (and a conditioning on the results of that measurement) has some nonzero probability of causing the original electron spins to become entangled.

But there’s a problem: if you’ve been paying attention, all we’ve done is cause the electron spins to become entangled with some tiny, nonzero probability (something like 6.4×10-9 in the actual experiment).  So then, why is this any improvement over the previous experiments, which just directly measured faraway entangled photons, and also had some small but nonzero probability of detecting them?

This leads to the third idea.  The new setup is an improvement because, whenever the photon measurement succeeds, we know that the electron spins are there and that they’re entangled, without having to measure the electron spins to tell us that.  In other words, we’ve decoupled the measurement that tells us whether we succeeded in creating an entangled pair, from the measurement that uses the entangled pair to violate the Bell inequality.  And because of that decoupling, we can now just condition on the runs of the experiment where the entangled pair was there, without worrying that that will open up the detection loophole, biasing the results via some bizarre correlated conspiracy.  It’s as if the whole experiment were simply switched off, except for those rare lucky occasions when an entangled spin pair gets created (with its creation heralded by the photons).  On those rare occasions, Alice and Bob swing into action, measuring their respective spins within the brief window of time—about 4 microseconds—allowed by the locality loophole, seeking an additional morsel of evidence that entanglement is real.  (Well, actually, Alice and Bob swing into action regardless; they only find out later whether this was one of the runs that “counted.”)

So, those are the main ideas (as well as I understand them); then there’s lots of engineering.  In their setup, Hensen et al. were able to create just a few heralded entangled pairs per hour.  This allowed them to produce 245 CHSH games for Alice and Bob to play, and to reject the hypothesis of local realism at ~96% confidence.  Jungsang Kim explained to me that existing technologies could have produced many more events per hour, and hence, in a similar amount of time, “particle physics” (5σ or more) rather than “psychology” (2σ) levels of confidence that local realism is false.  But in this type of experiment, everything is a tradeoff.  Building not one but two labs for manipulating NV centers in diamond is extremely onerous, and Hensen et al. did what they had to do to get a significant result.

The basic idea here, of using photons to entangle longer-lasting qubits, is useful for more than pulverizing local realism.  In particular, the idea is a major part of current proposals for how to build a scalable ion-trap quantum computer.  Because of cross-talk, you can’t feasibly put more than 10 or so ions in the same trap while keeping all of them coherent and controllable.  So the current ideas for scaling up involve having lots of separate traps—but in that case, one will sometimes need to perform a Controlled-NOT, or some other 2-qubit gate, between a qubit in one trap and a qubit in another.  This can be achieved using the Gottesman-Chuang technique of gate teleportation, provided you have reliable entanglement between the traps.  But how do you create such entanglement?  Aha: the current idea is to entangle the ions by using photons as intermediaries, very similar in spirit to what Hensen et al. do.

At a more fundamental level, will this experiment finally convince everyone that local realism is dead, and that quantum mechanics might indeed be the operating system of reality?  Alas, I predict that those who confidently predicted that a loophole-free Bell test could never be done, will simply find some new way to wiggle out, without admitting the slightest problem for their previous view.  This prediction, you might say, is based on a different kind of realism.

Ask Me Anything: Diversity Edition

Saturday, September 5th, 2015

With the fall semester imminent, and by popular request, I figured I’d do another Ask Me Anything (see here for the previous editions).  This one has a special focus: I’m looking for questions from readers who consider themselves members of groups that have historically been underrepresented in the Shtetl-Optimized comments section.  Besides the “obvious”—e.g., women and underrepresented ethnic groups—other examples might include children, traditionally religious people, jocks, liberal-arts majors… (but any group that includes John Sidles is probably not an example).  If I left out your group, please go ahead and bring it to my and your fellow readers’ attention!

My overriding ideal in life—what is to me as Communism was to Lenin, as Frosted Flakes are to Tony the Tiger—is people of every background coming together to discover and debate universal truths that transcend their backgrounds.  So few things have ever stung me more than accusations of being a closed-minded ivory-tower elitist white male nerd etc. etc.  Anyway, to anyone who’s ever felt excluded here for whatever reason, I hope this AMA will be taken as a small token of goodwill.

Similar rules apply as to my previous AMAs:

  • Only one question per person.
  • No multi-part questions, or questions that require me to read a document or watch a video and then comment on it.
  • Questions need not have anything to do with your underrepresented group (though they could). Math, science, futurology, academic career advice, etc. are all fine.  But please be courteous; anything gratuitously nosy or hostile will be left in the moderation queue.
  • I’ll stop taking further questions most likely after 24 hours (I’ll post a warning before closing the thread).

Update (Sep. 6): For anyone from the Boston area, or planning to visit it, I have an important piece of advice.  Do not ever, under any circumstances, attempt to visit Walden Pond, and tell everyone you know to stay away.  After we spent 40 minutes driving there with a toddler, the warden literally screamed at us to go away, that the park was at capacity. It wasn’t an issue of parking: even if we’d parked elsewhere, we just couldn’t go.  Exceptions were made for the people in front of us, but not for us, the ones with the 2-year-old who’d been promised her weekend outing would be to meet her best friend at Walden Pond.  It’s strangely fitting that what for Thoreau was a place of quiet contemplation, is today purely a site of overcrowding and frustration.

Another Update: OK, no new questions please, only comments on existing questions! I’ll deal with the backlog later today. Thanks to everyone who contributed.