4 ms·

So take what I'm about to say with a grain of salt but I believe that current generation of quantum computers (quantum digital) is fundamentally flawed. Basical

So take what I'm about to say with a grain of salt but I believe that current generation of quantum computers (quantum digital) is fundamentally flawed. Basically all architectures I've seen still use bits (qubits are still bits and use entanglement) as opposed to the superior signals. The class of computers I'm talking about is called continuous-variable quantum computers (https://en.wikipedia.org/wiki/Continuous-variable_quantum_information https://en.wikipedia.org/wiki/Continuous-variable_quantum_in...). Unlike DQ, it doesn't use entanglement. They are similar to the old school analog electric computers. They have some interesting properties and I can actually imagine programming one unlike DQ. There's a fourth class, continuous analog with entanglement which are superior to both, DQ, and continuous-variable quantum computers but right now we should really be looking into the continous variable ones.

reikonomusha 9y ago
Why do you think using two-level systems is fundamentally flawed? Even those are analog devices that span a finite dimensional space, but they’re nonetheless dense vector spaces over C^2^n. You still get theoretically infinitely parameterizable operations, though we know some discrete subset of those is sufficient for universal computation. If one were to work with an infinite dimensional system, you’d still be employing finite truncations of infinite dimensional operators, leading you back to, more or less, a finite dimensional subspace. Digitization—or rather, discretization—is important, and the reason computers have managed to be so successful. And programming a system like a universal gate-based quantum computer can be done now, with languages like Quil and libraries like pyQuil [0]. [0] http://pyquil.readthedocs.io/en/latest/ http://pyquil.readthedocs.io/en/latest/
- adamnemecek 9y ago
  Because you lose integration and differentiation in hw. You can also model signals "natively".
  - reikonomusha 9y ago
    You actually get those benefits, differentiation and integration, in the usual classical analog circuits, but, except for a few kinds of analog tricks, they’ve not outperformed their digital counterparts in precision, accuracy, or speed. You can make for neat demos, like wiring up integrators to solve the Lorenz attractor equations to make a nice oscilloscope plot, but the circuits fall short practically for anything more difficult.
    - adamnemecek 9y ago
      I'm well aware. Those that you are talking about are electric though which brings a whole class of issues. I think that photonic might work very well.
krastanov 9y ago
There is a serious misconception in your claim. Yes, analog computers, whether quantum or classical solve even NP-complete problems in polynomial time. No, they can not be constructed in the real world because analog computing does not permit error correction, and in the real world you have to deal with noise. Only very small analog computers (nothing scalable, nothing solving general problems) can be constructed before noise becomes an issue. Three good references: Book on quantum and classical computing: Aaronson's "Quantum Computing since Democritus" for gentle-for-newbies but rigorous discussion Very old paper on classical computing (analog vs digital): Von Neumann's "Probabilistic logics and the synthesis of reliable organisms from unreliable components" (pretty advanced) Newish (old for the field) paper on quantum computing: Calderbank's "Good Quantum Error-Correcting Codes Exist" Edit and addition: I work at Yale's Quantum Institute and we are some of the biggest proponents of "continuous variable" quantum computing. We use the continuous variables to encode a discrete "qudit" (with a "d") representation for the information, for all the reasons mentioned above (noise and error correction).
- adamnemecek 9y ago
  They aren’t misconceptions, I’m questioning some currently held assumptions. You are still talking about electric right? Why are assuming I’m unaware of your claims? I’m aware of e.g. error corection problems but I would definitely not say it’s impossible. Photonic systems have plenty of error correcting schemes.
  - krastanov 9y ago
    Does not matter whether it is electric or photonic (and many systems are in between). Scalable error correction (i.e. the error approaches zero in the limit of a large system) is possible only with digital systems, whether classical or quantum. There are things we can do to suppress errors and make a slightly bigger analog computer (classical or quantum) but there is no way to make an analog system scalable (i.e. not just lower errors to some floor, rather completely eliminate errors). The book cited in my previous comment explains the details in a fairly understandable fashion, but it takes up a whole chapter. The gist of it is, you need some minimal "logic distance" between your data "levels" in order to be able to distinguish them and correct deviations. This requires digitization. If somebody finds a way to error-correct analog representations they will have a way to solve NP-complete problems for instance. The Nobel price will be the least of their recognitions.
mathgenius 9y ago
Terry Rudolph has an excellent paper arguing for continuous-variable quantum computers [1]. CV still uses entanglement. You can think of CV states as infinite dimensional qubits. So in a way it is still "digital". Something about quantum measurement is fundamentally digital. [1] https://arxiv.org/abs/1607.08535 https://arxiv.org/abs/1607.08535

reikonomusha 9y ago

Why do you think using two-level systems is fundamentally flawed? Even those are analog devices that span a finite dimensional space, but they’re nonetheless dense vector spaces over C^2^n. You still get theoretically infinitely parameterizable operations, though we know some discrete subset of those is sufficient for universal computation. If one were to work with an infinite dimensional system, you’d still be employing finite truncations of infinite dimensional operators, leading you back to, more or less, a finite dimensional subspace. Digitization—or rather, discretization—is important, and the reason computers have managed to be so successful. And programming a system like a universal gate-based quantum computer can be done now, with languages like Quil and libraries like pyQuil [0]. [0] http://pyquil.readthedocs.io/en/latest/ http://pyquil.readthedocs.io/en/latest/

adamnemecek 9y ago
Because you lose integration and differentiation in hw. You can also model signals "natively".
- reikonomusha 9y ago
  You actually get those benefits, differentiation and integration, in the usual classical analog circuits, but, except for a few kinds of analog tricks, they’ve not outperformed their digital counterparts in precision, accuracy, or speed. You can make for neat demos, like wiring up integrators to solve the Lorenz attractor equations to make a nice oscilloscope plot, but the circuits fall short practically for anything more difficult.
  - adamnemecek 9y ago
    I'm well aware. Those that you are talking about are electric though which brings a whole class of issues. I think that photonic might work very well.

krastanov 9y ago

There is a serious misconception in your claim. Yes, analog computers, whether quantum or classical solve even NP-complete problems in polynomial time. No, they can not be constructed in the real world because analog computing does not permit error correction, and in the real world you have to deal with noise. Only very small analog computers (nothing scalable, nothing solving general problems) can be constructed before noise becomes an issue. Three good references: Book on quantum and classical computing: Aaronson's "Quantum Computing since Democritus" for gentle-for-newbies but rigorous discussion Very old paper on classical computing (analog vs digital): Von Neumann's "Probabilistic logics and the synthesis of reliable organisms from unreliable components" (pretty advanced) Newish (old for the field) paper on quantum computing: Calderbank's "Good Quantum Error-Correcting Codes Exist" Edit and addition: I work at Yale's Quantum Institute and we are some of the biggest proponents of "continuous variable" quantum computing. We use the continuous variables to encode a discrete "qudit" (with a "d") representation for the information, for all the reasons mentioned above (noise and error correction).

adamnemecek 9y ago
They aren’t misconceptions, I’m questioning some currently held assumptions. You are still talking about electric right? Why are assuming I’m unaware of your claims? I’m aware of e.g. error corection problems but I would definitely not say it’s impossible. Photonic systems have plenty of error correcting schemes.
- krastanov 9y ago
  Does not matter whether it is electric or photonic (and many systems are in between). Scalable error correction (i.e. the error approaches zero in the limit of a large system) is possible only with digital systems, whether classical or quantum. There are things we can do to suppress errors and make a slightly bigger analog computer (classical or quantum) but there is no way to make an analog system scalable (i.e. not just lower errors to some floor, rather completely eliminate errors). The book cited in my previous comment explains the details in a fairly understandable fashion, but it takes up a whole chapter. The gist of it is, you need some minimal "logic distance" between your data "levels" in order to be able to distinguish them and correct deviations. This requires digitization. If somebody finds a way to error-correct analog representations they will have a way to solve NP-complete problems for instance. The Nobel price will be the least of their recognitions.

mathgenius 9y ago

Terry Rudolph has an excellent paper arguing for continuous-variable quantum computers [1]. CV still uses entanglement. You can think of CV states as infinite dimensional qubits. So in a way it is still "digital". Something about quantum measurement is fundamentally digital. [1] https://arxiv.org/abs/1607.08535 https://arxiv.org/abs/1607.08535