The tightrope of truth and courtesy
A reader calling him- or herself “A Merry Clown” left a comment on my previous post which was so wise, I decided it had to be promoted to a post of its own.
Scientific discourse is the art of juggling decorum, truth and humor. A high-wire feat, attempted under imposing shadows cast by giants and above the distraction of merry dancing clowns.
The “appropriate” tone for scientific discourse seems to be:
(a) Cordial. Always credit others for their hard work and good intentions (allow or at least pretend that others are basically well-intentioned, except in rare situations where there is proof of egregious misconduct).
(b) Biting, merciless and hard-nosed on the substantive issues. The truth deserves no less.
Perhaps the harsher (b) is, the gentler and more thorough (a) should be. After-all, human beings are what they are.
Certainly, provided one adequately treads through the niceties in (a), there’s no reason to worry about hurting anyone’s feelings in (b). Anyone who makes scientific claims in a professional or public arena should be prepared to put on their big boy pants or their big girl pants and have their claims face the brutal gauntlet of scientific scrutiny. All attempts should be made to avoid even the appearance that any part of (b) contains personal barbs or insults (unless these barbs happen to be to be hilarious.)
Outside of science the rule is: whoever flings the horseshit the hardest wins.
Essentially, what Shtetl-Optimized readers got to see this past week was me falling off the high wire (with tenure the safety net below? 
). I failed at a purely human level—though admittedly, while attempting a particularly difficult tightrope walk, and while heavily distracted by the taunts of both giants and clowns. I’ve already apologized to Cathy McGeoch for insulting her, but I reiterate my apology now, and I extend the apology to any colleagues at MIT who might have been offended by anything I said. I’ll strive, in future posts, to live up to a higher standard of cordiality, composure, and self-control.
At the scientific level—i.e., at level (b)—I stand by everything I wrote in the previous post and the comments therein.
Follow
Comment #1 June 6th, 2013 at 1:54 am
This cordiality stuff gets confusing. It would be easier if you were a hard-ass skeptic all the time.
Comment #2 June 6th, 2013 at 2:13 am
Also…please don’t forget people like me who have very little interest in whether D-wave is telling the truth and lots of interest in whether their latest machine is a scientific advance and doing something non-classical and potentially exciting.
On that note, I prefer to think of D-wave’s machines like all those demonstrations of the grand factoring of the number 15 (one day they will surely crack that one ). In particular cases one could imagine this could be exciting even if there is no clear way to extend the method to 35. (Yes I have read the paper describing the flaw in most of the existing demonstrations.) I don’t care that there is some genius out there who can factories 35 in their head.
In any case, more detailed analysis of their experiments and less of their descriptions of their own success would be great.
Comment #3 June 6th, 2013 at 2:23 am
Ignorant bystander #2: OK, but a huge difference between the two cases, which hardly anyone seems to acknowledge, is that the people using quantum computers to factor 15 don’t claim they’re doing anything beyond what you could do classically! And personally, I find it impossible to do science in a mode where you ignore the truth or falsehood of what’s being claimed, though maybe some people can function that way. (For more about this, see the “Wrap-Up” in my previous post.)
Comment #4 June 6th, 2013 at 2:39 am
I should perhaps have added that, not withstanding my previous comment, I completely agree that some of the claims in previous post comments of what is acceptable in terms of hype ( or what I would call lying) are frankly shameful.
But this would have distracted from my plea for less chat about chat.
Comment #5 June 6th, 2013 at 8:54 am
Bystander #4: Unfortunately, one of the main lessons I drew from the previous post was that I should do precisely the opposite of what you suggest! It’s when people goad me into trying to address the “bigger” questions—where will this lead, etc.—that I leave myself most vulnerable to attacks by D-Wave partisans, and am also most likely to lose my cool. After all, I have arguments and opinions, but I don’t have a crystal ball any more than anyone else. Instead, I should just keep hammering away at what I know, which is the huge gap between what D-Wave claims and what’s actually been established.
Comment #6 June 6th, 2013 at 11:10 am
“A Merry Clown” sounds remarkably similar, in tone and substance, to the banished John Sidles. What a comeback if so!
Comment #7 June 6th, 2013 at 1:22 pm
notamerryclown #6,
I seriously doubt it. The post had no references or quotations in it and was directly related to what was being discussed. Plus I just can’t imagine John Sidles using the word “horseshit”.
Comment #8 June 6th, 2013 at 2:56 pm
Is there a tendency to consider certain modes of online communication (e.g., message boards, fora, blogs, etc.) as more “insulated” or “isolated” from other modes communication, online or in-person (e.g., e-mail or phone call) that predisposes the former to more acerbic or confrontational approches?
I don’t think that it is co-incidental that one may be much more willing to say negative things about people who have made themselves, in some sense, public figures on one’s blog or on an internet forum one belongs to than to say those same things to those same people in their presence.
Comment #9 June 6th, 2013 at 3:12 pm
For those of you who are interested, here is A Merry Clown’s post.
Comment #10 June 6th, 2013 at 3:20 pm
You’re apologising to a clown.
Man up – 98% of what you write is eminently defendable in the harsh climate of scientific discourse. ( maybe 97% 😉 )
Now just blog about your Free-will paper already
Comment #11 June 6th, 2013 at 5:42 pm
James Gallagher–
Comment #12 June 6th, 2013 at 5:44 pm
Ack…I forgot to close the blockquote tag.
Comment #13 June 6th, 2013 at 9:04 pm
Started the “free-will” paper. What I read so far is beautifully written.
Comment #14 June 6th, 2013 at 9:59 pm
Essentially, what Shtetl-Optimized readers got to see this past week was me falling off the high wire
The combination of tenure and children is well-known to be deleterious to brain function.
But it was certainly unrealistic to expect the D-Wavers to crumble like Hamentashen.
Having missed the entire discussion (though inferred the outcome from the original post), it was very informative to catch up by reading only the 35 Sidles posts from 17 May to 3 Jun. I would have assumed someone with so much time saying so many content-free things must not have a day job, and the content and tone suggest a typical garden variety quack, likely operating under a pseudonym, or even some “Mark V Shaney”-like SidlesBot.
But plugged into web search to find what the rest of you perhaps already know … “a Professor in the Department of Orthopaedics and Sports Medicine at the University of Washington School of Medicine … with a 1982 PhD in particle physics from the University of Washington … In addition to an active research career studying the biomechanics of ankle, knee, hip, and shoulder surgery, … also the inventor of magnetic resonance force microscopy (MRFM), a technology for uniting the non-destructive 3D imaging capability of magnetic resonance, with the atomic resolution of force microscopy … co-Directs the UW Quantum Systems Engineering (QSE) Laboratory.”
So independent of the ignorance, irrelevance, and confusion, why does he carve out the time to pester this blog; and web search further suggests it’s not just here … seems like an addiction and you’d be doing him a favor with 3 months rather than 2 weeks.
Comment #15 June 6th, 2013 at 11:36 pm
As an aside, I think this whole business of publishing academic papers based on paid consulting assignment is iffy. I’m not saying it’s wrong, but ought to be handled very carefully if potential exists where people may use the conclusions of your paper to take decisions relevant to the client-firm’s products.
I’m a tad surprised no one seems to be thinking of this dimension. There might be some points that Cathy McGeoch ought to apologize for.
Comment #16 June 7th, 2013 at 1:33 am
I’d like to stick to the main topic of this post for a moment. First, I do not believe in demanding apologies from other people. As I see it, an apology demanded is an apology devalued. Second, I don’t see that Cathy McGeoch ought to apologize for anything. (Nor did she demand an apology from anyone as far as I know.) Because, third, I disagree with her on some of the science, but it shouldn’t really go any further than that. For one particular reason I may have been too sarcastic in expressing my disagreement.
Namely, since she is faculty at a liberal arts college, it’s very honorable for her to continue research. In comparison, my own position looks somewhat decadent, for one reason because my teaching load is about half of hers. Even though the interpretation of her paper with Wang went awry, that’s not really her fault, and the work itself is honorable. Even if some aspect of a paper lead one to a blind alley, it can still be valid as research.
There is a swirling controversy about D-Wave (or I would say, a clownish controversy), and McGeoch-Wang is not at the same level as Yuting Yang’s paper that proves that gaps between primes are bounded. (In other words, that the lim inf of gaps between primes is bounded.) But otherwise, work by people like Cathy McGeoch is part of the same fine tradition as Zhang’s stunning work.
D-Wave itself is a different story.
Comment #17 June 7th, 2013 at 2:45 am
I have no horse in this race and only a very limited understanding of the science. I read Scott’s blog and the comments precisely for the mixture of graciousness and ferocity which he serves up brilliantly post after post. His rare excursions into anger merely serve to confirm that he is human and not in fact the emerged part of a secret Quantum AI program at MIT.
Comment #18 June 7th, 2013 at 4:09 am
I wholeheartedly second the above comment from Luca!
I was wondering if someone might be interested in summing up their thoughts on the best current understanding of the various performance comparisons that have been made with D-Wave, including:
(1) D-wave in McGeoch and Wang (were optimal parameters for the various solvers adequately achieved?)
(2) D-wave versus Troyer
(3) D-wave versus C-Plex with problem reformulation (http://arxiv.org/abs/1306.1202)
I am particularly curious about which comparisons are thought to be the most apples-to-apples at this point in time (and conversely, which might be apples-to-oranges).
Honestly, the question that most interests me is what exactly is the cheapest option available for reliably performing as well as D-wave on any given input distribution 🙂
Comment #19 June 7th, 2013 at 8:17 am
Misunderstanding – I’ll tell you what the bottom line is: As Peter Shor says, D-Wave hasn’t proven that their device is useful for anything. Considering the context, that’s a big deal. Their whole gambit is that academic experimentalists are distracted by perfectionism and academic theorists are distracted by abstraction, and they’re succeeding by concentrating on what’s actually useful.
Speaking as a theorist, although I have no rigorous proof that a D-Wave device can’t work, I have seen no satisfactory non-rigorous theoretical argument that it ever can work. (Where here “work” means, calculate anything potentially useful with a quantum acceleration.)
Comment #20 June 7th, 2013 at 8:45 am
Greg Kuperberg #16:
But otherwise, work by people like Cathy McGeoch is part of the same fine tradition as Zhang’s stunning work. D-Wave itself is a different story.
You don’t think any part of this is a pro-academia / anti-industry bias?
I mean I’ve no arguments with your very valid and relevant critiques of D-Wave but my impression is you are being a tad too charitable towards Cathy McGeoch’s work.
Is it really in the same class as the work on primes you cite? Is this really academic tradition at its best?
Comment #21 June 7th, 2013 at 9:09 am
Rahul – I feel “biased” against D-Wave for having spent 9 figures to get to where they are now, and for their irresponsible level of hype. Besides, if you look at where Dave’s two sold devices actually ended up (USC and NASA Ames) you could view it as an academic project that wears the clothes of industry.
No, McGeoch-Wang is not the same *class* as Zhang’s paper in the Annals, but it is in the same *tradition*. (Sorry, I had the name wrong, it’s Yitang Zhang. I hope Scott can correct it above.) Here I do not mean academia in general, but specifically teaching-based universities and colleges such as Amherst College. Compared to research universities, these are places of high toil and low pay for the faculty, and places with more quiet dignity and less overflowing ego.
Zhang is at UNH which actually is a research university albeit a relatively quiet one, but he is only a lecturer there, so it is the same type of thing.
Comment #22 June 7th, 2013 at 9:45 am
Rahul #15:
As an aside, I think this whole business of publishing academic papers based on paid consulting assignment is iffy … I’m a tad surprised no one seems to be thinking of this dimension.
Yes, in retrospect, it would have been far better if I had focused on the fact that McGeoch and Wang took on their work as a paid consulting assignment, with the basic parameters of the experiments already set by someone else—and discussed the general issues for academia raised by that sort of thing—rather than making what were all too easily construed as ad hominems. That’s especially true, since my only real interest here is what can or can’t be concluded from the experiments. There are many, many people who I have good reasons to dislike at a personal level, but Cathy McGeoch isn’t one of them! 🙂
Comment #23 June 7th, 2013 at 10:20 am
Scott #22:
Yes, in retrospect, it would have been far better if I had focused on the fact that McGeoch and Wang took on their work as a paid consulting assignment, with the basic parameters of the experiments already set by someone else
Conversely, it’d have been far better if Cathy McGeoch had made it clear in her paper about the paid consulting aspect and the fact that she was working under external directions about what aspects to test.
Comment #24 June 7th, 2013 at 11:21 am
Well, “follow the money” is the common denominator of all counterarguments, the rebuttal that even the most unschooled outsider can understand. And I think that in this case it is not an important argument. I would be very surprised if D-Wave has actually swayed Cathy McGeoch’s loyalty with cash. The real issue for D-Wave is that they cherry-picked their favorite benchmark. And the real issue for Cathy is mainly that (as the paper admitted) her benchmark work wasn’t really finished. The media ran away with it as if it was finished work, but, live and learn.
Comment #25 June 7th, 2013 at 12:10 pm
Scot #22:
> There are many, many people who I have good reasons to dislike at a personal level
Poking around for Sidles turned up within his absurd repetitiveness his repeated use on multiple different blogs of the supposed Dirac quote, “A Golden Era is an era in which ordinary people can make extraordinary contributions” (i.e., in quotation marks, and attributed to Dirac, though it does not sound at all like Dirac). Occasionally he dials it back ever so slightly to call it Telegdi quoting Dirac, but has also given a supposed reference to 1975 lectures by Dirac in which the comment is made. In that reference, however, while Dirac does discuss how much easier it was in the early days of quantum mechanics “for any second-rate physicist to do first-rate work”, he *never* makes the precise comment repeatedly attributed to him, and indeed never even utters the words “golden era”.
Apparently Sidles has a non-conventional notion of what constitutes a quotation, and while this may seem picayune, it is symptomatic of the damage he causes. Web search reveals that others have subsequently picked up the quotation, and attributed it directly to Dirac without mention of Sidles (clearly the font of the misinformation, having repetitively seeded wherever he could).
In contrast to Prof. J.Preskill’s well-known humane tendencies, I would maintain that a mere 3 month ban was far too kind for someone who can make 35 comments in 17 days in a single thread yet not contribute usefully to discussion. And that was just one thread on one blog.
Comment #26 June 7th, 2013 at 12:42 pm
Greg, #21 after your instance that somehow Tesla and Virgin Galactic are terrible enterprises, I really didn’t think you could surprise me any more. But the fact that I can utterly not predict your mind, certainly lends strong empirical evidence to Scott’s Knightian uncertainty ideas.
You are utterly confusing me with your disdain for “an academic project that wears the clothes of industry.”
So, getting commercial money to enrich academia is a *bad* thing?
Comment #27 June 7th, 2013 at 4:51 pm
MC = JS and the last sentence is a response to the ban.
Comment #28 June 8th, 2013 at 1:29 am
While I sympathize with Cathy McGeoch in that she presumably stepped into a deep bog that she might not have known was there, I want to point out that she does have some egg on face.
A CS professional absolutely should know that QC is unproven and highly controversial. A minute’s search into the background of who is offering the cash would reveal that D Wave is the ONLY company in the world that is trying to sell QC devices. Thus, a highly favorable report should darn well raise pointed questions.
Comment #29 June 8th, 2013 at 1:47 am
Raoul Ohio #28:
Exactly! I find arguments like the ones offered by Greg Kuperberg a bit naive. It’s a stretch to believe that Cathy McGeoch was totally misquoted by journalists or duped by DWave.
Just read this verbatim quote from Cathy’s conclusions and then decide if we really ought to blame the journalists?:
In all three experiments the V5 hardware or its hybrid counterpart Blackbox found solutions that tied or bettered the best solutions found by
software solvers. Performance was especially impressive on instances that can be solved directly in hardware. On the largest problem sizes tested, the V5 chip found optimal solutions in less than half a second, while the best software solver, CPLEX, needed 30 minutes to nd all optimal solu-
tions.
My gut feeling is she knew exactly what she was getting into here. Yes, DWave and journalists need to be blamed but to treat Cathy as an innocent victim is not realistic.
She has to shoulder some significant blame in this saga.
Comment #30 June 8th, 2013 at 7:33 am
Scott’s recent lapse from his own high standards follows hot on the heels of a surprising amount of time spent on Luboš Motl’s blog. Could the two perhaps be related?
Comment #31 June 8th, 2013 at 7:46 am
Hi all,
I think Dwave should really look at paper
http://arxiv.org/abs/1306.1202)
I think that having a Convex Programming Problem with Penalty (Relaxation as a competitor is essentially a losing battle. Those beasts are Awesome! With all of the attention that Dwave has got, I cannot see Dwave surviving the onslaught.
Zelah
Comment #32 June 8th, 2013 at 8:08 am
Hi all,
To be fair to Dwave, Has there been for the Quantum Gate Model a proven speedup for Quadratic Convex Programming? Unfortunately utilizing Grover Speed naively does not work as the search space is exponential!
I ask this question as Quadratic Convex Programming Relaxations are a formidable competitor to Dwave at this time and comparisons would be interesting.
Zelah
Comment #33 June 8th, 2013 at 2:02 pm
Slipper-mystery, I personally find John Sidles’ contributions full of content and interesting. (Often he assumes knowledge that I do not have (even in mathematics) which makes it difficult to understand.) I cannot see why his quote of Dirac is damaging or a big deal at all, but if the quote is indeed inaccurate, I am quite certain that John will be happy to stand corrected. Of course, if you don’t like John’s comments you can simply skip them.
Comment #34 June 8th, 2013 at 3:20 pm
Simpe question but….has anyone tried using Gurobi instead of Cplex? It’s much faster in some cases.
Comment #35 June 8th, 2013 at 3:40 pm
Gil Kalai #33:
Of course, if you don’t like John’s comments you can simply skip them.
Isn’t that a generic argument that could be made for all kinds of content. Spam, off-topic, ad-hominems, hate-speech, trolls, robots; sure we could just skip all of them.
“You-can-simply-skip-them” does not seem a good reason to not censor certain unwelcome content. It’s not my place to decide what is unwelcome, but Scott seems to have made a judicious decision here.
Comment #36 June 8th, 2013 at 7:02 pm
I always thought of JS as a physics guy who knows much more than I do about physics, but who at least in the past was not very well informed about TCS and logic (not that I’m any expert in those topics myself). Plus he was obsessed with the word “STEM”. So some of his posts here and on stackoverflow got a little bit annoying sometimes. I usually just sort of skimmed over his posts and occasionally found interesting things in them.
Comment #37 June 8th, 2013 at 8:04 pm
Rahuls #28, 29 Cathy et. al. conducted the tests to the specifications of the customers, i.e. Google and Lockheed Martin, in this kind of situation they will have retained her rather than D-Wave (through an intermediary to guarantee confidentiality). Since she was clear that she had the publication rights, the paper would have been published even if the results would have been too disappointing for the customers to ultimately purchase a D-Wave machine.
This has been pointed out repeatedly, yet all Rahuls here keep insisting that D-Wave paid her. I am not an open source clearing house, but frankly IMHO this kind of incorrect info should not be posted here, after what transpired in the previous thread. At this point there is little excuse to still get these details wrong.
As to Greg K. I’d say he is obviously sarcastic in his praise for Cathy. But we already established that I have no idea how his mind works.
Comment #38 June 8th, 2013 at 9:07 pm
Henning@37,
Given how frustrated you’ve become dealing with so many incorrect, naive or nefarious characters on this cite, perhaps your herculean efforts are simply going to waste. I too find it hard to understand why these people can’t see the error of their ways and simply agree with your interpretation of the data and evaluation of D-Wave. Frankly, IMHO, I’m absolutely certain that history will make it clear for all to see! 😉
Comment #39 June 8th, 2013 at 9:55 pm
Henning:
I think my points in #28 are 100% correct. You appear to be stating in #37 that either (1) she did not get paid, or (2) someone other than DW put up the money. That is at odds with the established facts in the matter.
Comment #40 June 8th, 2013 at 10:20 pm
Scott:
I want to register my vote of appreciation for the time and effort you put into being the front man in many important debates. The resulting discourse is important in advancing knowledge. Not many are sharing the load.
This endeavor takes a lot of gumption, because the front position makes you target #1 for slings and arrows from knowable critics and morons alike. I admire your ability to (usually) keep your replies at a high level. Hang in there. Or, Illegitimi non carborundum, as I would say when I am speaking fake Latin.
Comment #41 June 8th, 2013 at 10:33 pm
I hereby propose reducing John Sidles’ sentence to time served.
Here is my take:
John reportedly contributes to various blogs, etc., many times a day. Some contributions are really good, and you can’t expect anyone to be on the money with everything. The downside is that sometimes he gets on a rant and can’t let go. I suggest to Scott that when this happens, either (1) ignore him, and he will probably stop, or (2) give him a four hour time out to calm down.
Comment #42 June 8th, 2013 at 11:44 pm
Gil Kalai #33:
>: Of course, if you don’t like John’s comments you can simply skip them.
This is a sensible methodology, and the one I’ve long employed. In this case, however, I had not been following the long discussion and learned of the ban instead from Comment #6 above, so was enticed to look at only his postings in order to learn what line had finally been crossed.
> (Often he assumes knowledge that I do not have (even in mathematics) which makes it difficult to understand.)
From physics so I have the background to understand the comments, and to understand why they are irrelevant, wrong, or misleading.
> I cannot see why his quote of Dirac is damaging or a big deal at all,
It simply serves as a clear-cut illustration of his pernicious methodology, to repeat something that is factually incorrect in as many places as possible. And in most (if not all) cases turned by web search, many blogs, many discussion threads, its use had little relevance to the topic under discussion. If you are not concerned to have fictitious quotations, as well as mistaken physics, inserted into your memory then by all means continue to:
> find John Sidles’ contributions full of content and interesting.
Though that is worrisome, presuming you need accurate and concise information. In any event, the primary desideratum is whether the comments are entertaining to the maintainer of the blog.
> but if the quote is indeed inaccurate, I am quite certain that John will be happy to stand corrected.
He knows it is inaccurate (because he gives a false reference), yet continues to use it. Perhaps not a “big deal”, except insofar as it is symptomatic of larger systematic problems.
This is why I was surprised to learn that he has a legitimate day job, and the web search was an informative exercise. If he were to throttle his blog commentary and focus more on his day job, then perhaps both would benefit, hence my comment that three months was already too kind.
Comment #43 June 9th, 2013 at 12:02 am
Rahul – I wouldn’t call Cathy an “innocent victim” and I don’t see that I’m actually being naive. (And to Henning Dekant: No, I’m not being sarcastic. Speak for yourself, bucko.) You could call McGeoch-Wang itself a naive paper. However, you can in principle say that about any paper or scientist who’s wrong about anything.
I’m just trying to explain various ways in which Cathy has been reasonable enough. I can still make those points even if I disagree with the thrust of her paper.
Comment #44 June 9th, 2013 at 12:43 am
Henning Comment #37:
If I was factually wrong, I apologize. But I fail to see how. Take your assertion that “all Rahuls here keep insisting that D-Wave paid her. “
Are you saying this isn’t factually true? I quote from Cathy herself: “D-Wave retained me as a consultant in September 2012 to help them with a project.[…]I did the work, wrote my report for D-Wave, and yes, collected a consultant’s fee.’
Clearly, D-Wave paid her. I don’t see why you say they didn’t. Whether this did or did not influence anything is another story.
Comment #45 June 9th, 2013 at 4:44 am
Mike #38, the frustration you detected, had more to do with the fact that I currently only have an iPad and had to retype three times after arriving at my hotel room in Palm Springs.
Greg K. glad to hear you weren’t sarcastic. Told you that I can’t figure you out 🙂
Rahul #44, you are right, I thought I remembered that the customers retained Cathy through an intermediary. But they apparently used D-Wave in this role, which indeed makes for a more messy situation, although not uncommon if a sufficient level of trust has been established between the customers and a vendor.
Still don’t think that this is any grounds to doubt Cathy’s professionalism, especially as it had been made clear from the beginning that this work was on behalf of said customers.
Comment #46 June 9th, 2013 at 4:01 pm
I am not qualified to analyze the experiments on the D-Wave machine (DWM). I request those who are to comment on the following question.
Has it been demonstrated that the DWM does anything at all?
One possible alternative is that the DWM is outputting some random noise and the conventional computers involved turn this into output.
Comment #47 June 9th, 2013 at 7:14 pm
Raoul, the D-Wave machine certainly does *something*. I’m sure it makes a fabulous space heater. It can be used to solve some optimization problems. What you probably mean is ‘Does it do something non-classical?’
The short answer is that there currently isn’t any experimental evidence indicating that the D-Wave machine is anything but a classical annealer. There’s a phenomenon in how it comes up with solutions which it was hypothesized might be indicative of quantum phenomena, but a classical mechanism for generating that phenomenon has since been proposed. There’s a question of whether the proposed classical phenomenon matches *exactly* the observed phenomenon, but this is a silly game over what is at best very weak evidence.
A more direct way of indicating that quantum phenomena is going on is to demonstrate a big speed-up. The paper which started the recent hubbub compared the D-Wave machine to solving the same problem using classical techniques and found that the machine did better. Such a result should be taken cautiously, because applying heuristics to solve optimization problems is more art than science and there’s no reason to think any particular ones can’t be beat besides a track record of nobody being able to beat them. The classical runtimes given in that paper have since been improved on considerably and now the D-Wave machine’s performance is clearly inferior.
As bad as that sounds, the theoretical problems are even worse. D-Wave recently scaled back what they’re claiming their machine is doing, and it turns out the newly claimed phenomena are all things which can be simulated classically without a polynomial increase in runtime. Even with the old claims, there’s still the problem that noone has come up with a problem for which there’s a reason to believe that quantum annealing should produce a result faster than classical annealing. It seems entirely plausible that there is such a problem, but building an actual machine in the hope that it will demonstrate some unforeseen phenomenon which will be useful for actual competition is a very, shall we say, optimistic way of investing money.
Comment #48 June 10th, 2013 at 2:17 am
Bram – I don’t think that it’s quite that simple. The evidence in arXiv:1304.4595 seems pretty strong that the D-Wave device really is doing some form of quantum annealing. Even the Smolin-Smith model is not that convincing, because even though it has generally similar annealing behavior, it is not hard and easy in the same cases as the D-Wave device.
However, in order to be interesting as any kind of quantum computing, the D-Wave device would have to be more than non-classical. After all, any atom or covalent bond is already non-classical. No, it would only be outright useful if it’s faster than classical computers. And it would only be potentially exciting if it is not classically simulable — whether or not it has some quantum effects. Other evidence in arXiv:1304.4595 suggests that it is both uncompetitive and classically simulable. And, the clincher for me is that the design never had good theoretical backing anyway.
In fact, if they make D-Wave devices with more and more “qubits”, then because it is at a fixed temperature that cannot be sent to zero, eventually it could be worse than classical annealing. It could have a coarse stage that is quantum annealing, and a fine stage that is just random guessing. That’s a motivation for so-called error correction in the D-Wave setting, which will mean encoding a logical “qubit” in many physical “qubits”. But I don’t know if the number of logical “qubits” can be sent to infinity in this game, or if it is effectively bounded.
Comment #49 June 10th, 2013 at 5:03 am
Hi Greg Kuperberg,
Addressed to Greg Kuperberg Comment 48:
“In fact, if they make D-Wave devices with more and more “qubits”, then because it is at a fixed temperature that cannot be sent to zero, eventually it could be worse than classical annealing.”
Dwave has been moving the goalpost for a long time.
The Quantum Complexity Community still think that Dwave are trying to build a Quantum Adiabatic Computer.
In reality, they are building a Quantum Annealing Machine.
This means that they cannot be worst that Classical Annealing. All of the proven slowdowns are for Quantum Adiabatic Computing
Finally, Classical Anealling for Convex Optimization problems have a prove slight speedup over Classical Convex Optimization. It be interesting to see if this is transferable to the Quantum arena
Regards
Zelah
Comment #50 June 10th, 2013 at 10:50 am
Okay the first answer that I got to the question is that even in this context, the overhead scaling for the repetition code should be logarithmic, so the number of logical qubits would not be bounded.
Comment #51 June 10th, 2013 at 11:37 am
ASDF#36: “STEM” is very much used by educational authorities and community funders all the way down—my wife was just given a blurb to craft into an awards citation that included “STEM” and “stakeholders” in the same sentence. IMHO the term is also used as “code” for the need to advance womens’ opportunities in these fields.
My remark in the long thread about the role for “color commentary” at an important moment in science was meant not only regarding John Sidles but several others including Scott himself. 🙂 John had substantial contribution to the QC debate at GLL and his comments and suggestions are always welcome there—and occasionally we’re able to act on some…
Comment #52 June 11th, 2013 at 3:31 pm
Law professor Eugene Volokh has an analogue to Scott’s advice:
“Be in their face–but with a breath mint.”
Comment #53 June 12th, 2013 at 5:21 am
Hey look! There’s a new popular introduction to quantum stuff, see http://quantumfrontiers.com/2013/06/11/quantum-matter-animated/#content
Comment #54 June 12th, 2013 at 6:02 am
Concerning truth and courtesy.
I have just read Aaronson’s lecture 14 “Skepticism of Quantum Computing” http://www.scottaaronson.com/democritus/lec14.html and found the following passage concerning my 2006 paper:
On to argument 7. This is an argument that Dyakonov makes many times in his recent paper. The argument goes that all the systems we have experience with involve very rapid decoherence, and thus that it isn’t plausible to think that we could “just” engineer some system which is not like any of the systems in nature that we have any experience with.
Q: Could we sic the “brains are quantum computers” people on these guys?
Scott: That’ll be good… put them in a room together. I hadn’t thought of that.
I actually had a less amusing reaction, which is that a nuclear fission reactor is also unlike any naturally occurring system in many ways. What about a spacecraft? Things don’t normally use propulsion to escape the earth. We haven’t seen anything doing that in nature. Or a classical computer. I don’t know if more than that needs to be said.
Well, I think that something more DOES need to be said. How would Aaronson like if I gave and then published a lecture in which HIS paper were treated in a similar manner that leaves the audience bewildered by the stupidity of the author:
<>
Anybody, Aaronson included, can easily find the following sentence in my Conclusions section:
“It seems likely that the (theoretical) success of fault-tolerant computation is due not so much to the ”quantum tricks”, but rather to the backdoor introduction of ideal (flawless) elements in an extremely complicated construction.”
Isn’t this somewhat different from what Aaronson describes as my main argument?
In his quest for truth (forget the courtesy), Aaronson could have discovered the half-page section 10 of my paper (the “Dyakonov challenge”).
After explaining what is a realistic noise model which “cannot be relaxed further without entering an imaginary world where something is ideal”, I pose the challenge:
“Let us focus on the simplest, almost ”trivial” task of storing just one qubit and let us verify the statement that its initial state can be maintained indefinitely in the presence of low-amplitude noise. Presumably, to maintain our single qubit close to its initial state, a certain sequence of operations (with possible branching depending on the result of intermediate measurements) should be applied periodically. PROVIDE A FULL LIST OF THESE ELEMENTARY OPERATIONS, so that anybody can use a PC to check whether qubit storage really works. The future quantum engineer will certainly need such a list! If it works, this demonstration would be a convincing, though partial, proof that the idea of fault-tolerant quantum computation is sound.”
After seven years since the publication of my paper, this challenge to my knowledge was never answered, theorem-proving is so much easier.
Presumably nobody knows how this tiny brick in the magnificient castle of fault-tolerant quantum computing should be constructed. Then where does the religious belief in the feasibility of QC come from?
Anyway, my arguments can hardly be refuted by the fact that we have built an aircraft (or a kettle) that does not exist in Nature.
Having not read the rest of Aaronson’s book, I can only hope that the way he discusses my arguments is not a manifestation of his general polemic strategy.
Comment #55 June 12th, 2013 at 6:19 am
In my preceding post the example of how I could have commented Aaronson’s paper, if I had followed his method, got lost. Here it is:
“In a recent paper Aaronson et al many times make the statement that one can construct maximally-nontrivial theories in every finite dimension and pretend to provide such a construction.
Q: Could we sic the “loop quantum gravity” people on these guys?
Michel: That’ll be good… put them in a room together. I hadn’t thought of that.
I actually had a less amusing reaction, which is that whenever the term “theories” in the plural is used this is an unmistakable sign of some intellectual deficiency. Is there any more to say?”
Comment #56 June 12th, 2013 at 12:03 pm
Michel Dyakonov #54, I suppose, the argument about single qubit with noise may be replied with more or less efforts (and even not enough to get ban here). I think, the idea of such answer might be based on logical qubits and quantum error correction. For me, idea of John Sidles about QED looks more promising, because in such a case we have to compensate regular laws of physics, not a random noise.
Comment #57 June 12th, 2013 at 2:45 pm
Michel Dyakonov, I’m not sure you want to hear the answer to your first question, but yes, in substance if not in tone, “fault-tolerant computation [implicates] ideal (flawless) elements” sounds very similar to “[we can’t engineer] some system which is not like any of the systems in nature that we have any experience with”.
By the way, could you clarify your question in “PROVIDE A FULL LIST OF THESE ELEMENTARY OPERATIONS”? Of course you know about the stabilizer codes, so what are looking for exactly?
Comment #58 June 13th, 2013 at 3:24 am
Jay, I think that the following two statements are completely different:
1) A method for fighting noise by introducing noiseless elements is not valid. It is like fighting the 2nd law by introducing elements which are not affected by thermal fluctuations, e.g. Maxwell demon
2) We cannot engineer a system which is not like any of the systems in nature that we have any experience with. This is an obvious nonsense.
I make statement 1 (among other ones, for example describing what a a realistic noise model, where NOTHING is exact). Aaronson ascribes statement 2 to me, as my main argument, which is false.
I believe that this is foul play.
Comment #59 June 13th, 2013 at 4:47 am
Alexander Vlasov (#56) and Jay (#57):
Storing one qubit is an elementary block of quantum memory. Now suppose that you are a physicist-engineer and you are very clever at manipulating qubits, making measurements etc.
You want to implement qubit storage and you are familiar with the existing theory of encoding, error correction, stabilizer codes, and all that, but you want to do it practically.
Moreover, before building the hardware with quite a lot of lasers, lock-in detectors, temperature control, (none of which can work perfectly), you would like to make a computer simulation to verify that for a sufficiently low noise level in all of your auxiliary devices qubit storage will indeed work.
You are aware of the fact that a large enough quantum system cannot be simulated by a classical computer, but you hope that in your (simplest) case the system is not so large, so that this simulation might be possible.
You then seek the advice of theorists and ask them, what practically should be done. You expect to get a list of operations (what exactly you should do, and in what order, gates, measurements, verifications, repetitions and so forth) to be performed (probably) periodically.
Moreover, to understand whether your equipment is good enough (or you should purchase a more precise equipment for more megabucks), you ask them to give you the required precision of all your manipulations. You would like to know also, how many qubits you need, and what is the number of gates and measurements to be performed during one storage cycle.
If the theorists were serious about really building a QC, they would jump on the occasion and spend the time necessary to answer your legitimate questions.
However, they are not interested, because it is much easier and better for their academic careers to prove another theorem about N qubits. So, they tell you “Of course you know about the stabilizer codes, so just go ahead”.
In fact, the theorists rely on you for doing their job.
Then you start thinking: One cannot build anything without first having some sort of blueprint. If nobody knows what should be done to store just one qubit, why is everybody so sure that it is possible to build a full scale QC with millions of qubits? Is it a religious belief?
Comment #60 June 13th, 2013 at 9:31 am
Michel, I find your most recent comment quite bizarre. There are numerous papers enumerating the steps and counting exact resources, and you seem to be suggesting that this entire literature does not exist. I can only assume you are some how missing the relevant plots in such papers. Generally the answer to “How many qubits do I need?” is not 7, but rather is some function which depends on how far below the threshold you are, and how long your computation is to last. However, some author are even more explicit. See for example some of Rod Van Meter’s papers, such as quant-ph/0607065.
Comment #61 June 13th, 2013 at 10:50 am
#58
You’re perfectly right that if noiseless element were required then almost no one would bet quantum computers are physically possible -in the same way almost no one thinks we’ll go beyond Church-Turing using noisless analog computer. But that is not the statement you quoted yourself in #54!
Your statement in #54 was that quantum computers DO require noiseless elements.
Look, if you were anyone but Michel Dyakonov, I would just have said: plain false, noiseless elements are not required, please search for “threshold theorem”.
But you are Michel Dyakonov, so my reading was what you actually meant was something much more subtle and not obviously wrong, such as “whatever the physical design you will try, you will never make the noise small enough or for long enough or with the right kind of (non)correlation”.
Is there any other non trivial reading of your statement in #54 that we should prefer?
#59
Sorry I still don’t get the point you’re trying to make. If you are a physicist-engineer very familar with things such as stabilizer code and error correction, then what do you need from theorists you can do by yourself?
As Joe Fitzsimons mentionned, many phycisist-engineer already did what you seem to ask for, for many different implementations, and published both theorical steps and sometime experimental demonstrations. We can’t believe you don’t know this, so what is the subtlety we miss?
Comment #62 June 14th, 2013 at 7:02 am
Joe #60 and Jay #61,
Thanks to both of you for your remarks. I must admit that I am not adequately familiar with this literature. So, I have studied Rod Van Meter’s dissertation quant-ph/0607065 presenting ideas and numerical estimates for the creation of a quantum multicomputer.
I found this work very interesting and of unusually high quality, I also liked his sober statements with which I absolutely agree:
“Indeed, the real-world feasibility of creating entanglement across thousands of qubits remains very much open to question” and
“Although small-scale quantum computers exist, prospects for large-scale ones remain uncertain”.
This means that for the time being it is a entirely a question of belief, and
it is a pity that one can never find this point of view expressed in mainstream literature, either scientific or popular. Please tell me, if you know of any examples. Also, do you agree that “prospects are uncertain”, and if yes, where in your opinion does the incertitude come from.
However, this work, which considers the optimum architecture of a full scale QC, does not answer my simple challenge. I believe that the storage of a single qubit is for QC what the Hydrogen atom is for atomic physics. It is the simplest problem where the efficiency of proposed error correction procedures can be tested by numerical simulation.
You tell me that this has been done many times already, and I believe you, since unlike me you are experts. Could you please give me references (for the particular problem of qubit storage, not for something much more general). Especially, I am interested to see the work explicitly taking into account that NOTHING can be exact.
Jay #61: Your statement in #54 was that quantum computers DO require noiseless elements.
That’s not quite exact. I said that the existing (by 2006) error-correcting schemes include flawless elements, which makes those schemes questionable. The model that I consider as realistic is described in my 2006 paper.
Comment #63 June 14th, 2013 at 8:08 am
Re sobers statements:
Well Scott will probably cry loud on this one, as he often points (in papers, his blog, his book) that his main interest in seeing a quantum computer is to test whether it’s possible in the first place. The main reason he (and most) thinks it’s possible is because it seems hard to get models in which both QM is right and QC are impossible. Do you think your paper in 2006 answered this challenge?
More recently there was an interesting, name-calling free, debate on Godel Lost Letter (I’d suggest the summary on Gil Kalai’s blog as the original discussion was a bit of a mess)
Re expert:
Contrary to Joe I’m just an “amateur éclairé” (hard to translate this one…), but you might find these ones interesting:
http://arxiv.org/pdf/1002.4659.pdf
10.1126/science.1057357
re #61: ok, I agree these claims are slightly different.
Comment #64 June 14th, 2013 at 10:44 am
Jay #63: I’m just an “amateur éclairé”
Moi aussi, et même pas assez éclairé…
Re: it seems hard to get models in which both QM is right and QC are impossible.
It also seems hard, even very hard, to understand why though classical mechanics is correct, the Maxwell demon is impossible. You will say: thermodynamics, 2nd law etc, but thermodynamics is nothing more than a rough scale description of a large system of particles obeying classical mechanics. Essentially, the answer is that we endow the demon with capabilities that are possible in principle, but highly (exponentially) improbable. Well, given the huge amount of literature devoted to the subject, let’s leave it at that.
You will say: do you mean that there is some general principle similar to the 2nd law that forbids scalable QC? I don’t know. Gil Kalai is looking for such a principle, and it certainly will be an important discovery, if it exists. But it well could be that there is no such general principle, like there is no principle forbidding your cat to learn QED.
This is why, I would like to study the question empirically by laptop simulation of qubit storage. Let the egg-headed theorists give me the required sequence of operations, I will then add noise and uncertainties in all possible places (like in reality) and see what happens.
Comment #65 June 14th, 2013 at 11:59 am
Michel Dyakonov #62 (#64) Just for certainty. Principles of quantum error correction may not be applied to single physical system/qubit. You must have n>4 systems with two basic values encoded in particular states of that systems. You must also have permanent supply of auxiliary qubits/systems in fixed/zero states to “transfer” noise from your “memory” to that auxiliary qubits…
Comment #66 June 14th, 2013 at 12:00 pm
I like this comparison with Maxwell demon very much. Yes classical mecanics seems at first sight to allows it. But I disagree that there is no fundamental reason to find (other than linearity of QM better describes physical reality than classical mechanics).
To me Bennett, following Landauer’s principle, has shown that a Maxwell demon can’t break the second law because what it does is to exchange information of one kind (enthalpy) for information of another kind (memory storage).
As this applies to any proposal an engineer can think of, not to some single species in particular, I consider that as a fundamental insight on nature. You may call that a very narrow sense if you wish. 🙂
On the other side, comparison with a single species, be it a cat or a donkey, is incorrect.
Yes there is no fundamental reason but it can’t read or learn QED. In the same way there is no fundamental reason but one can’t run windows 7 on abacus or the Babbage machine. In the same way there is no fundamental reason but most proposals for quantum computing are probably dead ends.
But does that says anything on the possibility to breed animals able to read, construct classical computer, or to find a way toward scalable quantum computers?
Of course not! Actually I breed some of these animals that can read and perhaps will learn QED one day. 🙂
Comment #67 June 14th, 2013 at 2:30 pm
Dear Michel,
I think your “simple challenge” is in fact anything but simple. As I understand your challenge, you are asking to be pointed to a paper showing how to construct a logical qubit which never succumbs to noise, and that you consider this the necessary fundamental building block for quantum computation. If this is indeed what you mean, then I think I see why you are skeptical about the possibility of large scale quantum computation, since you will never find such a paper, or if you do it will be incorrect. This is something which should not be possible, even in principle.
But fortunately, we do not need such a beast to perform fault-tolerant computation. We need only protect our qubits for long enough that logical errors are unlikely to have occurred during the computation.
So your hypothetical physicist-engineer needs to decide how long they need their qubits to last for, or equivalently, what the maximum probability of error is that they are willing to accept. Given that probability, you can easily find a sufficiently low per gate probability of error to guarantee the error probability for the computation as a whole is below this threshold. The remaining problem is simply to work backwards from this to determine the necessary parameters for the error correction code.
Sometimes there are closed form equations, and it is simply a matter of solving them. In other cases we need to rely on numerics. Ultimately, though, the number you are looking for depends on a several parameters (error probability per physical gate, the distribution of errors, the desired per gate error probability and the specific fault-tolerance scheme you use). What Rod does in his paper, and what others do in other similar papers, is to basically pick some reasonable representative choices for these parameters, and calculate the requirements. If you have a specific system in mind and a specific task you wish to accomplish, you can perform the equivalent calculation for that system.
Comment #68 June 14th, 2013 at 5:59 pm
There is no proof at all yet that algorithms like Shor’s algorithm will even work on a quantum computer. Isn’t it just as likely that someone will find a polynomial time algorithm for the integer factorization problem which will run on a classical computer? Or rather equally unlikely. Since no-one has mathematically ruled out this possibility, and there aren’t very many people working seriously on it.
Comment #69 June 15th, 2013 at 7:04 am
Hi Fitzsimons,
I have been thinking about the noise issues in regards to Quantum Error Correction. I accept you arguments that in principle that Quantum Error Correction is achievable.
However in pratice, it apprears that there is an enormous overhead of error correction qubits which in my opinion makes Quantum Error Correction impratical.
Has recent research made any headway regarding Error Correction overhead?
Thanks in advance
Zelah
Comment #70 June 15th, 2013 at 9:44 am
“Realistic Sceptic” #68: One could drive a Mack truck through the misconceptions in your comment. There is a proof that Shor’s algorithm would work on an (ideal) quantum computer: Shor gave it, and you can find it in any QC textbook. And even if a fast classical factoring algorithm were someday discovered, it wouldn’t invalidate that proof. And contrary to what you say, many people have searched for decades for a fast classical factoring algorithm, but the Number Field Sieve (which runs in mildly subexponential time) is the fastest they’ve found. And in general, “no-one has mathematically ruled out this possibility” strikes me as an absolutely terrible basis for judging something just as likely as its opposite: would you also say that (e.g.) Goldbach’s Conjecture or the Riemann Hypothesis are just as likely to be false as true?
Comment #71 June 15th, 2013 at 1:02 pm
Alexander Vlasov #65: Principles of quantum error correction may not be applied to single physical system/qubit.
I do not ask for anything diffrent from the standard error-correcting methods, I just want them applied to just one logical qubit (not thousand).
#65: You must have n>4 systems with two basic values encoded in particular states of that systems. You must also have permanent supply of auxiliary qubits/systems in fixed/zero states to “transfer” noise from your “memory” to that auxiliary qubits
I will accept whatever you want, just tell me what do you want to do to keep my logical qubit intact. This is called quantum memory (in its simplest form).
Comment #72 June 15th, 2013 at 1:15 pm
Michel Dyakonov #71, I wrote just because it sounds like you wanted to apply some scheme to single atom or something like that. Sorry if I was wrong.
Comment #73 June 15th, 2013 at 1:47 pm
To put Michel Dyakonov’s point of view, my own, and that of Robert Alicki’s in perspective, let me quote Scott himself in his response to Michel’s remark about nonlinear effects in quantum physics:
“You [Michel] ask about how to reconcile the nonlinearity of classical physics with the linearity of the Schrödinger equation. While this isn’t directly relevant to the “main” discussion, I think I can answer that. Quantum mechanics is linear only at the level of the amplitude vectors. It can be just as nonlinear as classical physics at the level of measurable objects, whether they’re qubits, electrons in a field, or whatever. Conversely, if you formulate (say) classical statistical physics abstractly enough, as a theory of transformations of probability vectors into other probability vectors by means of stochastic matrices, then it too will be linear; it’s only nonlinear if you look at the actual elements of the ensemble rather than at the distributions. Therefore, this isn’t a classical vs. quantum issue at all, but simply an issue of which level of description you’re interested in. (It so happens that in the quantum case, people more often prefer the abstract level of description, both because that level already contains the highly-nontrivial phenomenon of interference, and because a “concrete” description tends to be difficult or impossible. But again, classical probability is just as linear as quantum probability when you formulate them in a parallel way.)”
I remember reading at the time this paragraph several times and finding it completely reasonable! Scott talked about a large uncharted territory of non-linear phenomena in quantum physics (just like in classical physics), and explained that people mainly studied the abstract “linear” level as it already exhibited highly-nontrivial phenomena and because moving forward is “difficult or impossible”.
Of course, when we try to implement quantum computation we may not be able to shut off the large uncharted territory of nonlinear effects in quantum physics whose study is “difficult or impossible.” My interpretation of Michel Dyakonov’s opinion (but correct me if I am wrong, Michel) is that such non-linear phenomena will exclude quantum computation (or anything remotely close to it) for many ad-hoc reasons that cannot be clearly put into principles. So attempts to build scalable quantum computers will fail and yet they will not leave us with anything to show for our efforts. This is certainly a possibility. Robert Alicki’s opinion is that thermodynamics is the crucial area where obstacles to quantum computation comes from, and his research is centered around quantum thermodynamics. John Sidles raised (over here, and previously) the possibility that eventually further understanding of the foundation of particle physics will reflect negatively on the possibility of quantum computers. My point of view is that understanding the major phase transition reflected by quantum fault-tolerance, which I call the fault-tolerance barrier, is crucial, and that the impossibility of crossing this barrier is a basic physics principle, needed to be formalized and explored. (Of course, Robert’s attitude and mine are complementary and are not in contradiction also with John’s view.)
So my expectation is that the novel concepts of quantum error-correction and quantum fault-tolerance will not lead to the constructions of superior quantum computers, but rather will supply theoretical basis needed to move forwards in these “difficult or impossible” theoretical areas of nonlinear aspects of quantum physics, and will eventually explain why QC are not possible.
Regarding my debate with Aram Harrow mentioned by Jay, I would recommend reading the debate posts themselves. My summaries are personal, subjective, and very non-technical. (And attempt to pick entertaining moments.) Let me mention that for me personally, while exploring the secrets of our computational quantum world is a terrific challenge, understanding of how different people (sometimes coming from different disciplines) think about the issue is very exciting as well. I really try to understand how different people view this matter, especially since progress both in theory and in experiments is slow and it will take years for things to unfold. It is also surprising how many different issues are related to the debate regarding quantum computers (see this post and its follow-ups), and even the free-will question is somehow related 🙂 .
Comment #74 June 15th, 2013 at 3:27 pm
Dear Joe,
I am afraid their is some (mutual) misunderstanding.
Joe Fitzsimons #67 says: I think your “simple challenge” is in fact anything but simple. As I understand your challenge, you are asking to be pointed to a paper showing how to construct a logical qubit which never succumbs to noise
A logical qubit, which never succumbs to noise, BECAUSE of continuous error correction. I believe that this is what quantum memory means. Am I wrong?
It is believed that with error correction indefinitely long quantum cimputation can be performed, and even “arbitrary accurately”. Am I correct, that a trivial case is to indefinetly store just one qubit by using appropriate error correction?
If not indefinetly, I would be content to save it for 1000 decoherence times.
THIS is the specific system I have in mind and the specific task that I wish to accomplish. Is it asking too much?
The challenge consists in providing a detailed description of the required sequence of elementary operations or, if it has been done already, to provide a reference.
Comment #75 June 15th, 2013 at 3:56 pm
Dear Michel,
Quantum memories do not last, and cannot last forever, even with continuous error correction. The error-correction code has operators associated with it which cause a transition between the encoded states, and if we have independent noise on each qubit, there is always some finite probability of making this transition within an error correction cycle. The trick is to make the encoding to have a large distance (meaning that the operators that transition between logical states act non-trivially on a large number of qubits), and hence rely on the binomial distribution to keep such high weight errors improbable. So we do need to decide on a target lifetime for our qubits.
In the thesis I linked to, the author finds that two levels of the Steane [23,1,7] code suffice to protect the qubits long enough to factor a 1024 bit number. I would suggest that this is a proper answer to the question, not as asked, but as it might be reasonably reformulated (i.e. how do you protect a qubit for long enough to perform x number of logical gates, where x is sufficient to accomplish some task of interest).
1000 decoherence times is a bit of a moving target, since the exact encoding you need depends not just on this factor of 1000, but also on the ratio of gate time versus decoherence times (and potentially on T1 vs T2), and it is only possible if your gate times are much faster than the decoherence rate (since this ratio effectively gives you the error rate).
But if it is only a factor of 1000 you need, then it might be achievable by dynamic decoupling, decoherence free subspaces or other means, depending on the system. I believe quite a few qubits have seen decoherence times increased by factors of this magnitude by employing clever strategies to combat noise. The point of using quantum error correction is so that we can choose any lifetime we desire and choose an encoding which can protect our qubits for that length of time.
Comment #76 June 15th, 2013 at 4:12 pm
Hi, Gil,
#73
The possibility, that you cite: “So attempts to build scalable quantum computers will fail and yet they will not leave us with anything to show for our efforts.”
does look likely. But true, we should try to learn lessons from our failure (provided there are lessons to learn).
I think that the coming failure is somehow related to the question, why any phyisical quantity can be measured with a limited precision only?
When I tell this to a mathematician, they automatically respond that precision is just a question of time and resources (they have in mind the number of digits of pi that one can calculate). In physics things are different, one will NEVER be able to measure, say, temperature with a precision 10^(-12) .
As to the very interesting linear vs nonlinear issue in quantum mechanics, I am not ready to discuss it professionally, but I do believe that it might be important.
Comment #77 June 15th, 2013 at 5:11 pm
Michel, thanks for the discussion, but for now I feel obliged to concentrate on an other stuff that is fun and easier for me (a vitriolic diatribe about free will, baysian rules, things like that). Please allow one last citation you might find interesting:
“Using this approach, coherence times up to 82 ms have been reported [148]. This time was further increased to 30 s by adding dynamic decoherence control [149].”
Comment #78 June 16th, 2013 at 4:15 am
Gil Kalai #73: I don’t understand you. You quote my explanation, about how to reconcile the nonlinearity of classical physics with the exact linearity of the Schrödinger equation, and say you find it “completely reasonable.” But then you go on to talk at length about the “large uncharted territory of nonlinear effects in quantum physics,” and how something there might prevent quantum computation.
However, the whole point of my explanation was that there’s nothing mysterious whatsoever about this linear/nonlinear issue! It results entirely from a confusion between abstraction layers: the Schrödinger equation is about an amplitude vector rotating through Hilbert space, while classical physics is typically formulated in terms of particles moving around in real space. It would be as if we were soberly discussing how to reconcile the word “duck” printed on a piece of paper, with the fact that real ducks are not made of paper—and then speculating that this paradox might forever prevent ducks from quacking. Well, maybe ducks can’t quack (I’m just a layperson, not an ornithologist 🙂 )—but asking about “the large, uncharted territory of 3-dimensional effects in real ducks, as compared to the 2-dimensionality of the word ‘duck’ written on paper” seems like an unpromising start to figuring out why. 😀
Comment #79 June 16th, 2013 at 12:57 pm
Hi Michel, one very specific plan for building robust (manipulable) quantum qubits based on much less-robust physical qubits is via distance-5 surface codes realized on a little over 100 physical superconducting qubits. As far as I know, there are fairly detailed plans for how to build these protected qubits, and perhaps one can look at these plans and try to figure out what can caused them to fail. (Using classical simulation, or some more detailed modeling, or sheer cleverness.) In any case, we can simply wait a few years since some experimental groups plan to implement this architecture!
Hi Scott, as you correctly said, there are non-linear phenomena in quantum physics just as there are in classical physics. You explained that studying these phenomena is harder, and that the linear aspects already have highly non-trivial exciting phenomena. As you explained, there is nothing mysterious about such non-linearity and nothing which is in conflict with linearity of Schrödinger equation. (Indeed, just like there are interesting non linear aspects of probability theory.) So, the fact that there are non-linear phenomena in quantum physics just like in classical physics is not mysterious, but non linear phenomena in quantum physics is a large uncharted scientific territory which may have a lot of mysteries. Now, all the models of noise/errors regarding quantum computation have ingredients which are nonlinear. The dependence of the rate of noise on the rate of computation is a non-linear phenomenon even for very mundane models that allow fault-tolerant quantum computation, as well as for noise models like mine which probably do not allow fault-tolerance.
Does your duck example, Scott, applies also to three-sexes and five-heads creatures that “almost certainly” have discovered the “Zork’s bloogorithm,” described on this 2D blog? 🙂 Do you believe that everything written on paper can be lifted up to reality?
Regarding Robert Alicki’s point of view, a recent U-tube lecture describes informally Alicki’s recent arXiv:1305.4910 paper which put forward an idea of a fundamental conflict between stability and reversibility (also relevant to the discussion on quantum computing).
Comment #80 June 28th, 2013 at 10:30 pm
http://www.wired.com/wiredenterprise/2013/06/d-wave-quantum-computer-usc/
To paraphrase one of the commenters, when do we get the quantum internet I can’t wait for lolcats that are alive and dead at the same time.