That there's still shreds of evidence for transformational improvements to the dreadfully arcane, inefficient, proprietary machines that are currently "the best".
Personally, my favorite is the Deutsch-Josza algorithm which illustrates a task which can be trivially understood, and provides exponential speedup over any classical algorithm that I can fathom to perform the same task.
Consider there was something like a half-century delay between the discovery of the photoelectric effect and the invention of the triode.
What's most striking to me is that these inherently optical phenomena apparently haven't been well-investigated for how they can provide speed-up to graphics processing.
Depends how you measure quantum supremacy. If you measure by seconds of computation, then the claim has been debunked since if you rent the most powerful supercomputer in the world you can match the performance of the experiment.
If you measure by e.g. energy usage then the claim remains, since it would cost orders of magnitude more energy to run the algorithm described in the paper than run Google's experiment.
I generally avoid discussing Quantum Computing as I know next to nothing, but I've been wondering:
Are there any inherently stable/noise free/non-Volatile Quantum Computing methods, at all?
Even just maintaining the state of a single Qubit for long periods without an exotic lab setup, or reliably transforming a single Qubit from a state, to a state with a near zero error rate?
>Are there any inherently stable/noise free/non-Volatile Quantum Computing methods, at all?
No, but there are methods (quantum error correction) that allow you to simulate noise free qbits with multiple, slightly noisy qbits. The main challenge here is that you need to start with quite good qbits for this error correction to actually help.
Not disagreeing necessarily but why do you think this? Codes have got better, qbits have got massively better over the last few decades, why wouldn't the two lines eventually cross?
Do you think there is some physical principle that will prevent getting a quantum speedup, or do you just think that the practical engineering challenges are simply too difficult?
I've become increasingly convinced that QM is incomplete for various reasons, and so I no longer have confidence that coherence will scale the way we expect. I think it will hit a hard limit that we can't yet see because of this incompleteness. Maybe that wall will become clear once we have quantum gravity, maybe some successor to QM like [1] will clarify what's happening.
If there is a phsyical principle that stops QCs from working, then trying to build QCs is probably the best way to make progress in fundamental physics.
Yes, quantum computers are good physics research in the same way that ITER can be considered good plasma research. Neither will produce a working product that some people are expecting. I'm not sure that QC will discover the problem before we infer it theoretically due to the many engineering challenges, but it's possible. If any real progress is made in QC, my bet's on boson sampling.
Yeah, seems very win-win. Either we get a working quantum computer or we make revolutionary new discoveries in physics. Either option would be pretty exciting.
Sure, they scaled it in the same way we're able to build bigger and more effective rockets than 50 years ago. We'll never travel faster than light though, because the speed of light is known to be the limit. I'm saying there will be some ultimate limit to coherence as well, but quantum mechanics is incomplete as a model of reality so we just don't see that in the math yet.
If that's the case, the QC computer endeavors are still worthwhile since while failing to scale them we may develop another piece of a more complete model of reality.
Could you briefly summarize one of the reasons you think QM is incomplete? I would agree with you that there is no reason to expect coherence to scale, but I don't know of any reason for it not to scale either.
Nobody knows of any certain reasons to not scale, but there's plenty of speculation. Someone else posted about Gil Kalai in this subthread, so he's one high profile person that has documented his skepticism on quantum computing and explained his reasons.
For my part, I don't think quantum foundations are on a firm footing. I think the measurement problem is still unresolved to a satisfactory degree, which means our understanding of decoherence is incomplete.
Per the link I posted above, there's also a good chance that gravity is inherently decoherent, which means entanglement will naturally breakdown in various conditions around mass.
I've also developed skepticism of quantum field theories. For one, the supposedly "most precise calculation in physics" has been marred by numerous mistakes and even fraud (electron magnetic moment). Renormalization is also sketchy business.
A more fringe reason, but one I think will resolve some of those issues is a growing skepticism of continuity [1], even for classical mechanics; continuous formalisms just seem to lead to logical and physical absurdities, like Norton's Dome and singularities in GR. Discrete theories are only now getting a little attention, but they're promising [2] because they seem to eliminate some of the formal structure (gauges), and the infinities disappear.
I don't believe there is a quantum computer that exists with a single stable logical qubit. I may be wrong, but if so, at best I think it's in the single-digits.
What the media reports is physical qubits (and hence not useful for computation), rather than logical.
> Are there any inherently stable/noise free/non-Volatile Quantum Computing methods, at all?
topological quantum computers are based on braid theory which is invariant to almost any kind of environmental noise, thus qubita stay in a coherent state much longer without error correction
This is the correct answer. Surface codes, like the ones discussed in the article, are essentially a way to simulate a kind of topological quantum computer on other architectures.
Nobody has yet built a topological qubit though, but Microsoft has claimed to be close for at least 5+ years now. On the other hand, TQC is supposed to be able to scale much faster, since the way in which you create a new qubit (by creating several anyons) doesn't necessarily require additional hardware - you could move the anyons for one qubit out of the way, and then use the same device you used to make them to make another[0]. Of course more hardware for manipulating additional qubits simultaneously may be desired - but the point is that the scaling problem is theoretically easier for TQC, even if creating the first qubit seems to be much more difficult.
They are not immune to all forms of errors - for instance, cosmic rays could cause unwanted anyons to form. But they are immune to most typical errors.
[0] This is a bit of an oversimplification. When talking about theoretical TQC, we are often talking about actually moving anyons confined to a 2D surface around. However, in the real world, the medium on which this happens is very disordered, so due to Anderson localization, anyons are actually trapped where they spawned. So this is where Majorana fermions and nanowires come in as a realistic approach where anyons can be moved, or alternatively, "measurement-based TQC" which relies on teleporting anyons instead of actually moving them.
> Are there any inherently stable/noise free/non-Volatile Quantum Computing methods, at all?
No. With quantum computers, you'd be delighted to have physical qubits where all gate error rates stayed below 1 in a thousand as you scaled up. Finding a qubit with massively better error rates, like one error per million gates, would be tantamount to inventing the quantum transistor.
Error correction should be able to reach arbitrarily low error rates. But it has a lot of overhead so, in terms of amount-of-stuff, it'll be more like building your computer out of cogs and gears than like building it out of transistors.
> Are there any inherently stable/noise free/non-Volatile Quantum Computing methods, at all?
Why would there be? Such a thing is impossible, since quantum states naturally decohere upon interaction with the environment, and therefore are inherently volatile unless isolated from the environment
There is a tension between qubits being able to interact with your quantum gates quickly, and not being able to interact with the environment quickly. Superconducting qbits can work well with gates but last for hundreds of microseconds, while nuclear spin qbits can last for seconds - but don't work so well with gates.
Universal quantum computing is still gated behind a technology known as a "magic state". Maintaining coherence is but one technical challenge to the implementation of a quantum computer, and while we have not solved it, I think it is likely to be the lesser of the two hurdles. The main problem with magic states is that the larger our computation needs to be, the more pure our magic state needs to be. The claim is that we can take many "copies" of impure magic states, and leverage QEC to produce a single "copy" of a purer magic state. Fortunately, given this, the number of copies we need in order scale our computation to n qubits for n operations is polynomial in n.
The big caveat, however, is that in order for this scheme to work, the impure copies need to have no entanglement between them. These are partial measurements of a larger quantum system which are by definition not maximally mixed, and as far as I know, no one has come up with a compelling physical or mathematical argument why it should be easy to find n separable mixed states like this. In fact, I think the more natural claim would be that this is hard to do, since separable pure states are a measure zero subset of pure states.
"Towards" in a title usually means the authors didn't do what follows in the title. I have no opinion on the article (and no expertise in the field), but whenever I see "towards", my aim in reading the article often shifts to understand where the authors got stuck.
38 comments
[ 8.9 ms ] story [ 98.0 ms ] threadhttps://www.techradar.com/news/scientists-say-theyve-debunke...
What do we have to be excited about then? Please explain.
Personally, my favorite is the Deutsch-Josza algorithm which illustrates a task which can be trivially understood, and provides exponential speedup over any classical algorithm that I can fathom to perform the same task.
Consider there was something like a half-century delay between the discovery of the photoelectric effect and the invention of the triode.
What's most striking to me is that these inherently optical phenomena apparently haven't been well-investigated for how they can provide speed-up to graphics processing.
If you measure by e.g. energy usage then the claim remains, since it would cost orders of magnitude more energy to run the algorithm described in the paper than run Google's experiment.
Are there any inherently stable/noise free/non-Volatile Quantum Computing methods, at all?
Even just maintaining the state of a single Qubit for long periods without an exotic lab setup, or reliably transforming a single Qubit from a state, to a state with a near zero error rate?
My apologies if these are general knowledge
No, but there are methods (quantum error correction) that allow you to simulate noise free qbits with multiple, slightly noisy qbits. The main challenge here is that you need to start with quite good qbits for this error correction to actually help.
Do you think there is some physical principle that will prevent getting a quantum speedup, or do you just think that the practical engineering challenges are simply too difficult?
[1] https://arxiv.org/pdf/2102.07795.pdf
For my part, I don't think quantum foundations are on a firm footing. I think the measurement problem is still unresolved to a satisfactory degree, which means our understanding of decoherence is incomplete.
Per the link I posted above, there's also a good chance that gravity is inherently decoherent, which means entanglement will naturally breakdown in various conditions around mass.
I've also developed skepticism of quantum field theories. For one, the supposedly "most precise calculation in physics" has been marred by numerous mistakes and even fraud (electron magnetic moment). Renormalization is also sketchy business.
A more fringe reason, but one I think will resolve some of those issues is a growing skepticism of continuity [1], even for classical mechanics; continuous formalisms just seem to lead to logical and physical absurdities, like Norton's Dome and singularities in GR. Discrete theories are only now getting a little attention, but they're promising [2] because they seem to eliminate some of the formal structure (gauges), and the infinities disappear.
[1] https://arxiv.org/abs/1609.01421
[2] https://arxiv.org/abs/1902.08997
and a nuanced lay-opinion by the same author: https://gilkalai.wordpress.com/2022/05/26/waging-war-on-quan...
Progress has been rather slow:
2001: Shor's algorithm was used to factor 15
2012: Shor's algorithm was used to factor 21
2019: Shor's algorithm was attempted at factoring 35, but failed due to too much error accumulation
What the media reports is physical qubits (and hence not useful for computation), rather than logical.
topological quantum computers are based on braid theory which is invariant to almost any kind of environmental noise, thus qubita stay in a coherent state much longer without error correction
https://en.m.wikipedia.org/wiki/Topological_quantum_computer
Nobody has yet built a topological qubit though, but Microsoft has claimed to be close for at least 5+ years now. On the other hand, TQC is supposed to be able to scale much faster, since the way in which you create a new qubit (by creating several anyons) doesn't necessarily require additional hardware - you could move the anyons for one qubit out of the way, and then use the same device you used to make them to make another[0]. Of course more hardware for manipulating additional qubits simultaneously may be desired - but the point is that the scaling problem is theoretically easier for TQC, even if creating the first qubit seems to be much more difficult.
They are not immune to all forms of errors - for instance, cosmic rays could cause unwanted anyons to form. But they are immune to most typical errors.
[0] This is a bit of an oversimplification. When talking about theoretical TQC, we are often talking about actually moving anyons confined to a 2D surface around. However, in the real world, the medium on which this happens is very disordered, so due to Anderson localization, anyons are actually trapped where they spawned. So this is where Majorana fermions and nanowires come in as a realistic approach where anyons can be moved, or alternatively, "measurement-based TQC" which relies on teleporting anyons instead of actually moving them.
No. With quantum computers, you'd be delighted to have physical qubits where all gate error rates stayed below 1 in a thousand as you scaled up. Finding a qubit with massively better error rates, like one error per million gates, would be tantamount to inventing the quantum transistor.
Error correction should be able to reach arbitrarily low error rates. But it has a lot of overhead so, in terms of amount-of-stuff, it'll be more like building your computer out of cogs and gears than like building it out of transistors.
Why would there be? Such a thing is impossible, since quantum states naturally decohere upon interaction with the environment, and therefore are inherently volatile unless isolated from the environment
The big caveat, however, is that in order for this scheme to work, the impure copies need to have no entanglement between them. These are partial measurements of a larger quantum system which are by definition not maximally mixed, and as far as I know, no one has come up with a compelling physical or mathematical argument why it should be easy to find n separable mixed states like this. In fact, I think the more natural claim would be that this is hard to do, since separable pure states are a measure zero subset of pure states.
(jk, the comment quality is what I come here for)
The part to click is this link at the top of the summary page:
That points to: https://www.nature.com/articles/s41586-022-05434-1>Suppressing quantum errors by scaling a surface code logical qubit [Open Access]
https://www.nature.com/articles/s41586-022-05434-1