4 ms·

I am not an expert, but my assessment is that it's pure marketing. They talk about needing millions of qubits to do useful computations. This is flat out false.

by wbhart 9y ago

I am not an expert, but my assessment is that it's pure marketing. They talk about needing millions of qubits to do useful computations. This is flat out false. Thousands will already solve interesting problems that no classical computer can currently solve, e.g. factoring RSA moduli. They also say that many qubits operating together will be required to build a practical quantum computer (true), and that their technology allows this (false). The problems involved in maintaining coherence and correcting errors, at reasonable operating temperatures do not boil down to how to fit multiple qubits on a silicon wafer. This is like saying a new process for manufacturing nuts and bolts will allow the construction of Boeing 747's. The paper may contain more sensible statements, but as usual, the press release is probably nonsense.

fudged71 9y ago
When I took a course on quantum computing in 2007 with a standford prof, he said that things would get interesting after we get to 128 qubits :)
- wbhart 9y ago
  I guess someone will eventually introduce some kind of terminology like, "qubit equivalent". But I think that if you could build an error free 128 qubit machine, this would be interesting. But the main reason I see for that is if you can build one with 128 qubits, almost certainly someone with a lot more money than you, with a black budget, can build one much larger. The basic thrust of the article, and the paper, seems to be that if you can engineer a large enough quantum computer than doesn't really work, you can simulate one that doesn't exist. It's truly great if they are right, but I can't find a single statement about this paper on the entire web by an expert that is enthused about this announcement. The paper is short enough that there should be some comment by now.
  - mtgx 9y ago
    Someone has already done that: https://www.research.ibm.com/ibm-q/resources/quantum-volume.pdf https://www.research.ibm.com/ibm-q/resources/quantum-volume....
- eberkund 9y ago
  That's a very specific number, do you know what the relevance of 128 is?
  - garmaine 9y ago
    He is probably referencing the 128 bit security threshold most crypto systems seek to achieve. However this doesn’t really line up with the number of qubits needed to break said crypto, which is typically much larger (thousands of qubits, plus error correction to survive millions of operations).
    - garmaine 9y ago
      Edit: less than a doubling. Eg from 128 to 192 or from 256 to 384. I meant to say a doubling is sufficient.
  - fudged71 9y ago
    No I don't think it was specific. I believe he was just guesstimating the exponential power required to do anything more interesting than basic math. We were all rather young students at the time so I don't think he had a specific application in mind. It was all rather new back then anyways, hard to imagine all the applications.
jessriedel 9y ago
You're probably confusing physical qubits (inside the actual physical device) and logical qubits (on which the platonic ideal calculation takes place). If we had 1000 logical qubits, we could indeed do useful things, but real-world implementations require huge amounts of error correction, which is much tricker on quantum computers than classical ones. Asymptotically the ratio of physical to logical qubits can approach 1 (just polylog overhead), but for early calculations the ratio will literally be more than a factor of a thousand. (This is true for all concrete proposals to date; people have speculated that there might be something useful to do with 1000 physical qubits for certain problems if the error could be small enough to be ignored and not corrected, but no one has actually shown explicitly how to do this, or produced the components with the quality that would be necessary.)
- wbhart 9y ago
  Yes, I was certainly confused. When IBM announced they are building a 7x7 = 49 qubit machine, I naturally assumed they meant logical qubits, not physical ones. Using surface codes, they can probably get one logical bit. But what confuses me are past records where they used Shor's algorithm to factor small integers on a quantum computer. How is this possible if we haven't even built a single logical qubit yet? It's surely necessary to have ~2N logical qubits to factor an integer of N bits. I have an idea that might help though: if logical qubits are so difficult to build and require thousands of times more physical qubits, and physical qubits are also difficult to build, perhaps we can simulate physical qubits with thousands of times more "ultra-physical" qubits.
  - jessriedel 9y ago
    Two excellent questions. Regarding claims of factoring small numbers: First note that error correction becomes exponentially more important for longer computations. Second, see this https://arxiv.org/abs/1301.7007 https://arxiv.org/abs/1301.7007 Regarding building a tower of ever less noisy qubits: it turns out there is something called a threshold theorem which requires a certain minimal level of noise before concatenating error correction schemes makes things better rather than worse. https://en.m.wikipedia.org/wiki/Quantum_threshold_theorem https://en.m.wikipedia.org/wiki/Quantum_threshold_theorem Reaching this threshold for physical qubits in the laboratory is perhaps the primary goal of experimental quantum computing.
marcosdumay 9y ago
> Thousands will already solve interesting problems that no classical computer can currently solve, e.g. factoring RSA moduli. With hundreds you can already speed-up chemistry simulations, what will have a much larger impact than breaking RSA.
- jabl 9y ago
  > With hundreds you can already speed-up chemistry simulations You can? How?
  - mtgx 9y ago
    You need an exponential increase in performance (power of N) to simulate an increasing number of molecules, while for quantum computers you may need only a polynomial increase in performance (power of 2). So it's possible to simulate 25-qubit-level chemistry with a laptop, 50-qubit-level chemistry with a supercomputer, and 100-qubit-level simulation with a supercomputer is impossible and will be impossible for a very long time. But going from a 50-qubit quantum computer to a 100-qubit one may take just 2 years (or 4, it doesn't really matter), and then another 2 years to get to 200 qubits, and so on. Quantum computers will solve chemistry problems in minutes that would otherwise have taken 100 years to simulate.
    - jabl 9y ago
      So what I'm asking is how does a quantum computer solve quantum chemistry problems? To make it concrete, say I want to calculate the band structure of a crystal using a quantum computer? What do I do? What kind of algorithms can the QC run that solves this problem faster/better than, say, a DFT code? I don't think it's a given that just because I want to simulate a quantum mechanical system in some way, that a quantum computer is a useful tool. Or it might be, I don't know. I just don't think it's enough justification that there happens to be the word "quantum" in both the problem description and in the name of the computer.
      - adrianN 9y ago

by wbhart 9y ago

fudged71 9y ago

When I took a course on quantum computing in 2007 with a standford prof, he said that things would get interesting after we get to 128 qubits :)

wbhart 9y ago
I guess someone will eventually introduce some kind of terminology like, "qubit equivalent". But I think that if you could build an error free 128 qubit machine, this would be interesting. But the main reason I see for that is if you can build one with 128 qubits, almost certainly someone with a lot more money than you, with a black budget, can build one much larger. The basic thrust of the article, and the paper, seems to be that if you can engineer a large enough quantum computer than doesn't really work, you can simulate one that doesn't exist. It's truly great if they are right, but I can't find a single statement about this paper on the entire web by an expert that is enthused about this announcement. The paper is short enough that there should be some comment by now.
- mtgx 9y ago
  Someone has already done that: https://www.research.ibm.com/ibm-q/resources/quantum-volume.pdf https://www.research.ibm.com/ibm-q/resources/quantum-volume....
eberkund 9y ago
That's a very specific number, do you know what the relevance of 128 is?
- garmaine 9y ago
  He is probably referencing the 128 bit security threshold most crypto systems seek to achieve. However this doesn’t really line up with the number of qubits needed to break said crypto, which is typically much larger (thousands of qubits, plus error correction to survive millions of operations).
  - garmaine 9y ago
    Edit: less than a doubling. Eg from 128 to 192 or from 256 to 384. I meant to say a doubling is sufficient.
- fudged71 9y ago
  No I don't think it was specific. I believe he was just guesstimating the exponential power required to do anything more interesting than basic math. We were all rather young students at the time so I don't think he had a specific application in mind. It was all rather new back then anyways, hard to imagine all the applications.

jessriedel 9y ago

You're probably confusing physical qubits (inside the actual physical device) and logical qubits (on which the platonic ideal calculation takes place). If we had 1000 logical qubits, we could indeed do useful things, but real-world implementations require huge amounts of error correction, which is much tricker on quantum computers than classical ones. Asymptotically the ratio of physical to logical qubits can approach 1 (just polylog overhead), but for early calculations the ratio will literally be more than a factor of a thousand. (This is true for all concrete proposals to date; people have speculated that there might be something useful to do with 1000 physical qubits for certain problems if the error could be small enough to be ignored and not corrected, but no one has actually shown explicitly how to do this, or produced the components with the quality that would be necessary.)

wbhart 9y ago
Yes, I was certainly confused. When IBM announced they are building a 7x7 = 49 qubit machine, I naturally assumed they meant logical qubits, not physical ones. Using surface codes, they can probably get one logical bit. But what confuses me are past records where they used Shor's algorithm to factor small integers on a quantum computer. How is this possible if we haven't even built a single logical qubit yet? It's surely necessary to have ~2N logical qubits to factor an integer of N bits. I have an idea that might help though: if logical qubits are so difficult to build and require thousands of times more physical qubits, and physical qubits are also difficult to build, perhaps we can simulate physical qubits with thousands of times more "ultra-physical" qubits.
- jessriedel 9y ago
  Two excellent questions. Regarding claims of factoring small numbers: First note that error correction becomes exponentially more important for longer computations. Second, see this https://arxiv.org/abs/1301.7007 https://arxiv.org/abs/1301.7007 Regarding building a tower of ever less noisy qubits: it turns out there is something called a threshold theorem which requires a certain minimal level of noise before concatenating error correction schemes makes things better rather than worse. https://en.m.wikipedia.org/wiki/Quantum_threshold_theorem https://en.m.wikipedia.org/wiki/Quantum_threshold_theorem Reaching this threshold for physical qubits in the laboratory is perhaps the primary goal of experimental quantum computing.

marcosdumay 9y ago

> Thousands will already solve interesting problems that no classical computer can currently solve, e.g. factoring RSA moduli. With hundreds you can already speed-up chemistry simulations, what will have a much larger impact than breaking RSA.

jabl 9y ago
> With hundreds you can already speed-up chemistry simulations You can? How?
- mtgx 9y ago
  You need an exponential increase in performance (power of N) to simulate an increasing number of molecules, while for quantum computers you may need only a polynomial increase in performance (power of 2). So it's possible to simulate 25-qubit-level chemistry with a laptop, 50-qubit-level chemistry with a supercomputer, and 100-qubit-level simulation with a supercomputer is impossible and will be impossible for a very long time. But going from a 50-qubit quantum computer to a 100-qubit one may take just 2 years (or 4, it doesn't really matter), and then another 2 years to get to 200 qubits, and so on. Quantum computers will solve chemistry problems in minutes that would otherwise have taken 100 years to simulate.
  - jabl 9y ago
    So what I'm asking is how does a quantum computer solve quantum chemistry problems? To make it concrete, say I want to calculate the band structure of a crystal using a quantum computer? What do I do? What kind of algorithms can the QC run that solves this problem faster/better than, say, a DFT code? I don't think it's a given that just because I want to simulate a quantum mechanical system in some way, that a quantum computer is a useful tool. Or it might be, I don't know. I just don't think it's enough justification that there happens to be the word "quantum" in both the problem description and in the name of the computer.
    - adrianN 9y ago