Archive for November, 2018

Airport idiocy

Wednesday, November 28th, 2018

On Sunday, I returned to Austin with Dana and the kids from Thanksgiving in Pennsylvania.  The good news is that I didn’t get arrested this time, didn’t mistake any tips for change, and didn’t even miss the flight!  But I did experience two airports that changed decisively for the worse.

In Newark Terminal C—i.e., one of the most important terminals of one of the most important airports in the world—there’s now a gigantic wing without a single restaurant or concession stand that, quickly and for a sane price, serves the sort of food that a child (say) might plausibly want to eat.  No fast food, not even an Asian place with rice and teriyaki to go.  Just one upscale eatery after the next, with complicated artisanal foods at brain-exploding prices, and—crucially—“servers” who won’t even acknowledge or make eye contact with the customers, because you have to do everything through a digital ordering system that gives you no idea how long the food might take to be ready, and whether your flight is going to board first.  The experience was like waking up in some sci-fi dystopia, where all the people have been removed from a familiar environment and replaced with glassy-eyed cyborgs.  And had we not thought to pack a few snacks with us, our kids would’ve starved.

Based on this and other recent experiences, I propose the following principle: if a customer’s digitally-mediated order to your company is eventually going to need to get processed by a human being anyhow—a fallible human who could screw things up—and if you’re less competent at designing user interfaces than Amazon (which means: anyone other than Amazon), then you must make it easy for the customer to talk to one of the humans behind the curtain.  Besides making the customer happy, such a policy is good business, since when you do screw things up due to miscommunications caused by poor user interfaces—and you will—it will be on you to fix things anyway, which will eat into your profit margin.  To take another example, besides Newark Terminal C, all these comments apply with 3000% force to the delivery service DoorDash.

Returning to airports, though: whichever geniuses ruined Terminal C at Newark are amateurs compared to those in my adopted home city of Austin.  Austin-Bergstrom International Airport (ABIA) chose Thanksgiving break—i.e., the busiest travel time of the year—to roll out a universally despised redesign where you now need to journey for an extra 5-10 minutes (or 15 with screaming kids in tow), up and down elevators and across three parking lots, to reach the place where taxis and Ubers are.  The previous system was that you simply walked out of the terminal, crossed one street, and the line of taxis was there.

Supposedly this is to “reduce congestion” … except that, compared to other airports, ABIA never had any significant congestion caused by taxis.  I’d typically be the only person walking to them at a given time, or I’d join a line of just 3 or 4 people.  Nor does this do anything for the environment, since the city of Austin has no magical alternative, no subway or monorail to whisk you from the airport to downtown.  Just as many people will need a taxi or Uber as before; the only difference is that they’ll need to go ten times further out of their way as they’d need to go at a ten times busier airport.  For new visitors, this means their first experience of Austin will be one of confusion and anger; for Austin residents who fly a few times per month, it means that days or weeks have been erased from their lives.  From the conversations I’ve had so far, it appears that every single passenger of ABIA, and every single taxi and Uber driver, is livid about the change.  With one boneheaded decision, ABIA singlehandedly made Austin a less attractive place to live and work.

Postscript I.  But if you’re a prospective grad student, postdoc, or faculty member, you should still come to UT!  The death of reason, and the triumph of the blank-faced bureaucrats, is a worldwide problem, not something in any way unique to Austin.

Postscript II.  No, I don’t harbor any illusions that posts like this, or anything else I can realistically say or do, will change anything for the better, at my local airport let alone in the wider world.  Indeed, I sometimes wonder whether, for the bureaucrats, the point of ruining facilities and services that thousands rely on is precisely to grind down people’s sense of autonomy, to make them realize the futility of argument and protest.  Even so, if someone responsible for the doofus decisions in question happened to come across this post, and if they felt even the tiniest twinge of fear or guilt, felt like their victory over common sense wouldn’t be quite as easy or painless as they’d hoped—well, that would be reason enough for the post.

Teaching quantum in junior high: special Thanksgiving guest post by Terry Rudolph

Thursday, November 22nd, 2018

Happy Thanksgiving!

People have sometimes asked me: “how do you do it?  how do you do your research, write papers, teach classes, mentor grad students, build up the quantum center at UT, travel and give talks every week or two, serve on program committees, raise two rambunctious young kids, and also blog and also participate in the comments and also get depressed about people saying mean things on social media?”  The answer is that increasingly I don’t.  Something has to give, and this semester, alas, that something has often been blogging.

And that’s why, today, I’m delighted to have a special guest post by my good friend Terry Rudolph.  Terry, who happens to be Erwin Schrödinger’s grandson, has done lots of fascinating work over the years in quantum computing and the foundations of quantum mechanics, and previously came up on this blog in the context of the PBR (Pusey-Barrett-Rudolph) Theorem.  Today, he’s a cofounder and chief architect at PsiQuantum, a startup in Palo Alto that’s trying to build silicon-photonic quantum computers.

Terry’s guest post is about the prospects for teaching quantum theory at the junior high school level—something he thought about a lot in the context of writing his interesting recent book Q is for Quantum.  I should stress that the opinions in this post are Terry’s, and don’t necessarily reflect the official editorial positions of Shtetl-Optimized.  Personally, I have taught the basics of quantum information to sharp junior high and high school students, so I certainly know that it’s possible.  (By undergrad, it’s not only possible, but maybe should become standard for both physics and CS majors.)  But I would also say that, given the current state of junior high and high school education in the US, it would be a huge step up if most students graduated fully understanding what’s a probability, what’s a classical bit, what’s a complex number, and any of dozens of other topics that feed into quantum information—so why not start by teaching the simpler stuff well?  And also, if students don’t learn the rules of classical probability first, then how will they be properly shocked when they come to quantum? 🙂

But without further ado, here’s Terry—who’s also graciously agreed to stick around and answer some comments.

Can we/should we teach Quantum Theory in Junior High?

by Terry Rudolph

Should we?

Reasons which suggest the answer is “yes” include:

Economic: We are apparently into a labor market shortage in quantum engineers.  We should not, however, need the recent hype around quantum computing to make the economic case – the frontier of many disparate regions of the modern science and technology landscape is quantum.  Surely if students do decide to drop out of school at 16 they should at least be equipped to get an entry-level job as a quantum physicist?

Educational: If young peoples’ first exposures to science are counterintuitive and “cutting edge,” it could help excite them into STEM.  The strong modern quantum information theoretic connections between quantum physics, computer science and math can help all three subjects constructively generate common interest.

Pseudo-Philosophical: Perhaps our issues with understanding/accepting quantum theory are because we come to it late and have lost the mental plasticity for a “quantum reset” of our brain when we eventually require it late in an undergraduate degree.  It may be easier to achieve fluency in the “language of quantum” with early exposure.

Can we?

There are two distinct aspects to this question: Firstly, is it possible at the level of “fitting it in” – training teachers, adjusting curricula and so on?  Secondly, can a nontrivial, worthwhile fraction of quantum theory even be taught at all to pre-calculus students?

With regards to the first question, as the child of two schoolteachers I am very aware that an academic advocating for such disruption will not be viewed kindly by all.  As I don’t have relevant experience to say anything useful about this aspect, I have to leave it for others to consider.

Let me focus for the remainder of this post on the second aspect, namely whether it is even possible to appropriately simplify the content of the theory.  This month it is exactly 20 years since I lectured the first of many varied quantum courses I have taught at multiple universities. For most of that period I would have said it simply wasn’t possible to teach any but the most precocious of high school students nontrivial technical content of quantum theory – despite some brave attempts like Feynman’s use of arrows in QED: The Strange Theory of Light and Matter (a technique that cannot easily get at the mysteries of two-particle quantum theory, which is where the fun really starts).  I now believe, however, that it is actually possible.

A pedagogical method covering nontrivial quantum theory using only basic arithmetic

My experience talking about quantum theory to 12-15 year olds has only been in the idealized setting of spending a few hours with them at science fairs, camps and similar.  In fact it was on the way to a math camp for very young students, desperately trying to plan something non-trivial to engage them with, that I came up with a pedagogical method which I (and a few colleagues) have found does work.

I eventually wrote the method into a short book Q is for Quantum, but if you don’t want to purchase the book then here is a pdf of Part I,, which takes a student knowing only the rules of basic arithmetic through to learning enough quantum computing they can understand the Deutsch–Jozsa algorithm.  In fact not only can they do a calculation to see how it works in detail, they can appreciate conceptual nuances often under-appreciated in popular expositions, such as why gate speed doesn’t matter – it’s all about the number of steps, why classical computing also can have exponential growth in “possible states” so interference is critical, why quantum computers do not compute the uncomputable and so on.

Before pointing out a few features of the approach, here are some rules I set myself while writing the book:

  • No analogies, no jargon – if it can’t be explained quantitatively then leave it out.
  • No math more than basic arithmetic and distribution across brackets.
  • Keep clear the distinction between mathematical objects and the observed physical events they are describing.
  • Be interpretationally neutral.
  • No soap opera: Motivate by intriguing with science, not by regurgitating quasi-mythological stories about the founders of the theory.
  • No using the word “quantum” in the main text! This was partly to amuse myself, but I also thought if I was succeeding in the other points then I should be able to avoid a word almost synonymous with “hard and mysterious.”

One of the main issues to confront is how to represent and explain superposition.  It is typical in popular expositions to draw analogies between a superposition of, say, a cat which is dead and a cat which is alive by saying it is dead “and” alive.  But if superposition was equivalent to logical “and”, or, for that matter, logical “or”, then quantum computing wouldn’t be interesting, and in this and other ways the analogy is ultimately misleading.  The approach I use is closer to the latter – an unordered list of possible states for a system (which is most like an “or”) can be used to represent a superposition. Using a list has some advantages – it is natural to apply a transformation to all elements of a list, for instance doubling the list of ingredients in a recipe.  More critically, given two independent lists of possibilities the new joint list of combined possibilities is a natural concept.  This makes teaching the equivalent of the Kronecker (tensor) product for multiple systems easy, something often a bit tricky even for undergrads to become comfortable with.

Conceptually the weirdest part of the whole construction, particularly for someone biased by the standard formalism, is that I use a standard mathematical object (a negative or minus sign) applied to a diagram of a physical object (a black or white ball).  Moreover, positive and negative balls in a diagram can cancel out (interfere).  This greatly simplifies the exposition, by removing a whole level of abstraction in the standard theory (we do not need to use a vector containing entries whose specific ordering must be remembered in order to equate them to the physical objects).  While it initially seemed odd to me personally to do this, I have yet to have any young person think of it as any more weird than using the negative sign on a number.  And if it is always kept clear that drawing and manipulating the whole diagram is an abstract thing we do, which may or may not have any correspondence to what is “really going on” in the physical setups we are describing, then there really is no difference.

There are some subtleties about the whole approach – while the formalism is universal for quantum computing, it can only make use of unitary evolution which is proportional to a matrix with integer entries.  Thus the Hadamard gate (PETE box) is ok, the Controlled-NOT and Toffoli likewise, but a seemingly innocuous gate like the controlled-Hadamard is not capable of being incorporated (without adding a whole bunch of unintuitive and unjustified rules).  The fact the approach covers a universal gate set means some amazing things can be explained in this simple diagrammatic language.  For example, the recent paper Quantum theory cannot consistently describe the use of itself, which led to considerable discussion on this blog, can be fully reproduced.  That is, a high school student can in principle understand the technical details of a contemporary argument between professional physicists.  I find this amazing.

Based on communication with readers I have come to realize the people at most risk of being confused by the book are actually those already with a little knowledge – someone who has done a year or two’s worth of undergraduate quantum courses, or someone who has taken things they read in pop-sci books too literally.  Initially, as I was developing the method, I thought it would be easy to keep “touching base” with the standard vector space formalism.  But in fact it becomes very messy to do so (and irrelevant for someone learning quantum theory for the first time).  In the end I dropped that goal, but now realize I need to develop some supplementary notes to help someone in that situation.

Q is for Quantum is certainly not designed to be used as a classroom text – if nothing else my particular style and choice of topics will not be to others’ tastes, and I haven’t included all the many, many simple examples and exercises I have students doing along with me in class when I actually teach this stuff.  It should be thought of as more a “proof of principle,” that the expository challenge can be met.  Several colleagues have used parts of these ideas already for teaching, and they have given me some great feedback.  As such I am planning on doing a revised and slightly expanded version at some point, so if you read it and have thoughts for improvement please send me them.

Ten updates

Wednesday, November 7th, 2018

If you like quantum, complexity, etc., then please read to the end! I’ve gotten a bunch of emails lately of the form “why haven’t you ever blogged about such-and-such?,” when it turned out that I damn well did blog about it; it was just somewhere down in a multi-item post.

1. Like many of you, I watched the US midterm election results with (mostly…) disappointment and dismay.  I think that history will judge us harshly for not totally and unequivocally rebuking everything Trump stands for and every politician associated with him.  But that’s not what I wanted to blog about today.

2. There was a breakthrough in communication complexity by Arkadev Chattopadhyay, Nikhil Mande, and Suhail Sherif: the first exponential separation between randomized communication complexity and log approximate rank for a total Boolean function f.  This falsifies the longstanding conjecture that these measures are polynomially related (though it doesn’t resolve the original log rank conjecture).  For those of you keeping score at home, the quantum communication complexity of f is sandwiched in between randomized CC and log approximate rank.  So, at least one of the following must now be true: either randomized CC is exponentially separated from quantum CC, or else quantum CC is exponentially separated from log approximate rank.  My money’s on the latter.

3. Ewin Tang, who achieved fame with a quantum-inspired classical algorithm for recommendation systems (which I blogged about in July), is now back with quantum-inspired classical algorithms for principal component analysis and supervised clustering.  Well, with the announcements of such algorithms; details of the analysis are to come later.

4. A bunch of people asked me about the paper by Sergey Bravyi, David Gosset, and Robert Koenig, Quantum advantage with shallow circuits.  tl;dr: it’s great!  And it was deservedly a highlight of the QIP conference back in January!  That’s why it confused me when everyone started asking about it a couple weeks ago.  The resolution is that the paper was just recently published in Science magazine, which led to popular coverage like this, which in turn led to people asking me whether this result unconditionally proves P≠BQP (that is, quantum computers can solve more problems in polynomial time than classical computers), and if not why not.  The answer is no: the paper proves an unconditional separation, but one that’s a long long way from P≠BQP, or anything else that would entail solving the central open problems of complexity theory like P vs. PSPACE.  Basically, it shows there are problems solvable in constant time with a quantum computer that aren’t solvable in constant time classically, for suitable meanings of “problem” (namely, a relation problem) and “in constant time” (namely, NC0 circuits, in which each output bit depends on only a constant number of input bits).  I understand that a stronger separation has since been achieved, between quantum NC0 and classical AC0, in work that’s not yet on the arXiv.  The problems in question, however, are all easy to solve in P, or even in classical logarithmic time, given a polynomial number of parallel processors.

5. A bunch of people also asked me about the paper by Xun Gao and Luming Duan, Efficient classical simulation of noisy quantum computation.  This paper tries to formalize something that many of us have suspected/feared for years, namely that random quantum circuits (the whole thing is specific to random circuits) can tolerate only a tiny amount of noise and decoherence before they become efficiently simulable classically.  If true, this has obvious implications for the sampling-based quantum supremacy experiments that Google and others are planning for the next few years: namely, that all the engineering effort they’ve already been investing anyway to push down the noise rate, will actually be necessary!  However, correspondence with the authors revealed that there’s a significant gap in the analysis as it currently stands: namely, the current proof applies only to closed quantum systems, which would (for example) rule out all the techniques that people eventually hope to use to achieve quantum fault-tolerance—all of which are based on constantly measuring subsets of the qubits, doing essentially error-free classical computation on the measurement outcomes, throwing away noisy qubits, and pumping in fresh qubits.  Xun and Duan say that they’re currently working on an extension to open systems; in my personal view, such an extension seems essential for this interesting result to have the informal interpretation that the authors want.

6. My friend Bram Cohen asked me to announce that his company, Chia, has launched a competition for best implementation of its Verifiable Delay Functions (VDFs), with real money rewards.  You can find the details at this Github page.

7. The second Q2B (“Quantum 2 Business”) conference, organized by QC Ware Corp., will be held this coming December 10-12, at the Computer History Museum in Mountain View.  There will be two keynote addresses, one by John Preskill and the other by me.  I hope I’ll get a chance to meet some of you there!

8. Longtime colleague and friend-of-the-blog Ashwin Nayak asked me to announce that the 2019 Conference on Computational Complexity, to be held July 18-20 in exciting New Brunswick, NJ, is now accepting submissions.  I hope to be there!

9. OK, what the hell: the 21st annual, and nearly impossible to capitalize correctly, SQuInT (Southwest Quantum Information and Technology) workshop will be held February 2019 in Albuquerque, NM.  UT Austin is now a node of the SQuInT network, and I’ll hopefully be attending along with a posse of students and postdocs.  The deadline for abstract submission is coming up soon: Monday November 12!

10. I went to morning Shabbat services in Austin this past weekend, exactly one week after the tragedy in Pittsburgh.  There was massively increased security, with armed guards interrogating us, Israeli-style, about why we had no membership sticker on our car, whether we knew the name of the rabbi, etc.  Attendance was maybe a factor of three higher than usual.  About thirty priests, ministers, and Christian seminary students, and one Muslim, came up to the bima to say a prayer of solidarity with Jews.  The mayor of Austin, Steve Adler, was also on hand to give a speech.  Then the rabbi read a letter to the synagogue by Sen. Ted Cruz denouncing antisemitism (well, parts of it; he said the letter was really long).  There were murmurs of disapproval from the crowd when Cruz’s name was mentioned, but then everyone quieted down and listened.  The thing is, the US and large parts of the world are now so far outside the norms of liberal democracy, in territory so terrifyingly uncharted since the end of World War II, that shooting up synagogues is bad is actually something that it’s better than not for powerful people to affirm explicitly.  Anyway, while I’m neither a believer nor much of a synagogue-goer, I found the show of unity moving.