> tl;dr: no crypto is broken, some people just did some tricks with misusing fancy big words in an otherwise alright but now-retracted paper, and i’m Very Mad On Line about it, also everyone should give me shells on their supercomputers
I believe you might be right, interference is exactly what you want from a quantum computer I think.
And the test they did show that they actually achieved interference. How you'd call that quantum supremacy is a bit of a strech tough if you ask me. Looks to be like saying 'hey, a quantum system can simulate quantum behavior quicker than a classical computer' which is kinda implied
“Quantum supremacy” has been a long-standing term invented by the quantum information science research community for the QIS community many years ago. It wasn’t made recently and settled being “the” term to mean a specific mathematical/empirical result.
Quantum advantage may not be the perfect term, but it seems to be a better fit than quantum supremacy. The former term conveys a sense that it has a limited scope, while the latter seems to imply that it beats classical across the board, or in the majority of cases, or even just in the ones we care about.
I get that the term quantum supremacy was coined long ago, by people who were most likely not looking at the event with the degree of granularity that our present technologies require. It's just unfortunate that what made sense on a macro level becomes illogical when looking at the minutiae.
Although I suppose this IS fitting in a way, since it parallels quantum mechanics itself.
From now on, quantum computers will beat classical computers, one problem at a time, until they are the only viable option for all the quantum easy problems (no we don't know exactly what those are). Supremacy seems quite fitting.
So? A huge number of scientific ideas are badly named for laypeople. We have both "somewhat homomorphic encryption" and "partially homomorphic encryption" and they mean different things. That's done bad jargon. But when targeting the research community it works fine.
There are no naming committees. Names pop up largely by accident. This is true for all fields. Heck the entire field "artificial intelligence" is named in a way that causes no end of incorrect beliefs among laypeople.
The solution is better journalism, not fighting over how names are understood outside the community.
A real reason for papers such as this, even with seemingly bombastic claims as "supremacy" with modest improvements, is to prove real progress at a time where QIS funding is sharply ramping up by DOE/DOD/(NSA/CIA) while also identifying core areas for improvement (where to spend that money).
$300MM has been set aside for DOE QIS alone (with more of a focus on simulation), and that is likely to ramp up with papers like this, probably somewhere over $1B over the next 5 years for DOE Science alone. Throw in DOD/NSA and the number is easily going to be in the $5-10B total over the next 5-10 years, assuming there's no quantum winter.
The paper claims that a classical supercomputer is MUCH slower at simulating a noiseless 53 logical qubit system,
but doesn't exclude the possibility of a classical supercomputer simulating an approximation of the actual noisy 53 bit physical qubit system, that still gives sufficiently decent estimates of the observed final state probabilities. Such a simulation would compute orders of magnitude fewer than 2^53 amplitudes, and use whatever shortcuts are available to get approximately correct results.
Quantum supremacy cannot be claimed until this possibility is ruled out.
To folks in the field, it is my impression that this is about as minor as assuming P != NP. Conjectured, unproven, but the world would be a lot more interesting if wrong.
Quantum physicist here: just to correct the mistaken claims in this, interference is definitely a feature of quantum computing, and not a bug.
Interference is also NOT why there are are no error corrected quantum computers. The reason there are no error corrected quantum computers is because we need way (way way) many more physical qubits, and they would need to be of better quality (less decoherence).
I don’t think it’s a mistaken claim so much as a misused term-of-art by conflating it with its layman meaning. If you replace “interference” with “noise”, what’s being said makes sense.
However, the second bit you said is true, even after correcting the term.
TLDR; I don't think this twitter thread is good public science communication, and should not appear on the HN frontpage.
As a quantum computing researcher, I can't make head and tail of what OP is saying after and including tweet 7. I don't think OP understands interference correctly. As someone points out elsewhere in the comment section, she might mean decoherence, but I can only evaluate what she has written.
She also makes light of the theoretical computational complexity results underlying the demonstration of quantum supremacy using random circuits. This experiment's results and interpretation might be iffy at best, but the underlying strategy of the quantum computing community in this matter are sound. We don't want to keep asking for funding for the next 20 years without proofs of principles in between, and this experiment is one step along the way.
> however, despite other interesting finds in the paper (such as being able to collapse state and read output from multiple qubits at once), claiming quantum supremacy feels to me to be somewhat aggrandising and disingenuous, given what the quantum circuit was actually computing
As the person says, they have been out of the field. For better or for worse, this is the term the field settled on years ago to mean showing that a quantum computer can sample from the Porter-Thomas distribution more quickly than a classical computer, “beyond reasonable doubt.” (What I mean by the last bit is that it must be an empirical demonstration that matches the very well established theory.)
It’s maybe worth repeating: “Quantum supremacy” isn’t some gimmick made up by the authors of this paper. It was used as early as 2012 [0] by John Preskill, a well known scientist in the field. It has well-defined meaning in the research community: a specific experimental verification of a mathematical result which shows plausible rejection of the extended Church-Turing thesis.
This result may not seem like much to the average reader, but—provided it passes the muster of peer review—it’s a huge step of progress in the development of quantum computers.
I'm not a quantum physicist, but I pretend that copious amounts of reading and watching PBS Space Time and Infinite has made me somewhat knowledgeable on this. The author looks to have previously been a theoretical physicist, so I might be very off base here, but... this thread doesn't square up with my understanding of things at all.
>QUANTUM INTERFERENCE IS JUST WHAT QUANTUM PROCESSORS DO, ITS A BUG, NOT A FEATURE
This sort of throws the whole series of tweets into question for me. Of course interference is a feature - you want to create a pattern of interference where paths leading to the wrong answer interfere with each other and cancel out! Without interference I don't understand how quantum computers could even work, what they would look like, and why we would want to use them for anything.
>which, to me, is an infuriating thing to label “quantum supremacy” because quantum interference is EXACTLY WHY WE CAN’T USE CURRENT QUANTUM PROCESSORS TO “BREAK ALL CRYPTO” IN THE FIRST PLACE BECAUSE WE CANNOT DO EFFICIENT QUANTUM ERROR CORRECTING CODES
Error correction requires more qubits to do the error correction. We need the current qubits just to do the calculations, leaving us no qubits to use for error correction. Quantum interference certainly isn't the reason we can't do error correction.
If the author is confusing decoherence and interference, then I can understand why they believe what they do, though that confusion ultimately removes a significant portion of their point. Decoherence is what takes something in superposition and puts it back into the realm of classical physics - we lose the ability to utilize that quantum interference and entangled qubits. Ultimately to get a result we have to de-cohere the qubit, because we have to observe it, so not all decoherence is bad. The problem is all sorts of things besides our desired observation can cause decoherence, which results in errors, and avoiding this requires implementing error correction, which takes additional physical qubits. We need a lot more physical qubits, qubits that are more resistant to decohering when we don't want them to, and/or for someone to come through with a breakthrough that does error correcting much more efficiently than the methods Shor came up with.
Of course, I might be the one that's confused here.
34 comments
[ 3.1 ms ] story [ 84.8 ms ] threadAnd the test they did show that they actually achieved interference. How you'd call that quantum supremacy is a bit of a strech tough if you ask me. Looks to be like saying 'hey, a quantum system can simulate quantum behavior quicker than a classical computer' which is kinda implied
I get that the term quantum supremacy was coined long ago, by people who were most likely not looking at the event with the degree of granularity that our present technologies require. It's just unfortunate that what made sense on a macro level becomes illogical when looking at the minutiae.
Although I suppose this IS fitting in a way, since it parallels quantum mechanics itself.
There are no naming committees. Names pop up largely by accident. This is true for all fields. Heck the entire field "artificial intelligence" is named in a way that causes no end of incorrect beliefs among laypeople.
The solution is better journalism, not fighting over how names are understood outside the community.
$300MM has been set aside for DOE QIS alone (with more of a focus on simulation), and that is likely to ramp up with papers like this, probably somewhere over $1B over the next 5 years for DOE Science alone. Throw in DOD/NSA and the number is easily going to be in the $5-10B total over the next 5-10 years, assuming there's no quantum winter.
Quantum supremacy cannot be claimed until this possibility is ruled out.
https://news.ycombinator.com/item?id=21004615
Interference is also NOT why there are are no error corrected quantum computers. The reason there are no error corrected quantum computers is because we need way (way way) many more physical qubits, and they would need to be of better quality (less decoherence).
However, the second bit you said is true, even after correcting the term.
As a quantum computing researcher, I can't make head and tail of what OP is saying after and including tweet 7. I don't think OP understands interference correctly. As someone points out elsewhere in the comment section, she might mean decoherence, but I can only evaluate what she has written.
She also makes light of the theoretical computational complexity results underlying the demonstration of quantum supremacy using random circuits. This experiment's results and interpretation might be iffy at best, but the underlying strategy of the quantum computing community in this matter are sound. We don't want to keep asking for funding for the next 20 years without proofs of principles in between, and this experiment is one step along the way.
As the person says, they have been out of the field. For better or for worse, this is the term the field settled on years ago to mean showing that a quantum computer can sample from the Porter-Thomas distribution more quickly than a classical computer, “beyond reasonable doubt.” (What I mean by the last bit is that it must be an empirical demonstration that matches the very well established theory.)
It’s maybe worth repeating: “Quantum supremacy” isn’t some gimmick made up by the authors of this paper. It was used as early as 2012 [0] by John Preskill, a well known scientist in the field. It has well-defined meaning in the research community: a specific experimental verification of a mathematical result which shows plausible rejection of the extended Church-Turing thesis.
This result may not seem like much to the average reader, but—provided it passes the muster of peer review—it’s a huge step of progress in the development of quantum computers.
[0] https://arxiv.org/abs/1203.5813
>QUANTUM INTERFERENCE IS JUST WHAT QUANTUM PROCESSORS DO, ITS A BUG, NOT A FEATURE
This sort of throws the whole series of tweets into question for me. Of course interference is a feature - you want to create a pattern of interference where paths leading to the wrong answer interfere with each other and cancel out! Without interference I don't understand how quantum computers could even work, what they would look like, and why we would want to use them for anything.
>which, to me, is an infuriating thing to label “quantum supremacy” because quantum interference is EXACTLY WHY WE CAN’T USE CURRENT QUANTUM PROCESSORS TO “BREAK ALL CRYPTO” IN THE FIRST PLACE BECAUSE WE CANNOT DO EFFICIENT QUANTUM ERROR CORRECTING CODES
Error correction requires more qubits to do the error correction. We need the current qubits just to do the calculations, leaving us no qubits to use for error correction. Quantum interference certainly isn't the reason we can't do error correction.
If the author is confusing decoherence and interference, then I can understand why they believe what they do, though that confusion ultimately removes a significant portion of their point. Decoherence is what takes something in superposition and puts it back into the realm of classical physics - we lose the ability to utilize that quantum interference and entangled qubits. Ultimately to get a result we have to de-cohere the qubit, because we have to observe it, so not all decoherence is bad. The problem is all sorts of things besides our desired observation can cause decoherence, which results in errors, and avoiding this requires implementing error correction, which takes additional physical qubits. We need a lot more physical qubits, qubits that are more resistant to decohering when we don't want them to, and/or for someone to come through with a breakthrough that does error correcting much more efficiently than the methods Shor came up with.
Of course, I might be the one that's confused here.
Let the computation power grow exponentially, Until one day, computation faces its vulnerability.
Let all be known, beautiful dawn holds its crown.
Computation is a growing pyramid, da da da, dancing on a brittle tip.