Renting quantum on the cloud: data privacy, adoption and error correction


If an individual or commercial enterprise resorts to renting quantum computing power on the cloud, wouldn't the company who owns and hosts the quantum computer be able to see/monitor/eavesdrop on the electronic transmissions that their customers make, causing privacy concerns? And about adoption, first of all, how has the business and usage statistics of renting cloud services been for supercomputers anyway?

About error correction, How would the user be able to verify the reliability of the estimates and solutions returned by the superconductor they are renting, given how the current state of the technology is widely known to lack error correction mechanisms, or is that just a misconception?


Posted 2019-11-20T14:38:45.667

Reputation: 835



These are two separate questions, so I will (try to) answer them separately as well.

Concerning the (reduction of) privacy in remote or cloud computing.

Without any alterations, the instructions for a quantum computation that is to be run on a remote computer can be seen by that remote computer. That is to say, if you want to conceal the computations that you want to perform (be it the type of computations or the input to said operation), this is not possible using this approach.

There is a parallel to be made to classical remote computation: there exist methods to hide/encrypt the input parameters to some remote computation, which is known as Homomorphic encryption. This can be used to hide the input parameters of the quantum computation as well, in principle.$^{1}$.

To conceal the type of computation that you want to perform, is a much more challenging task. There is no way of doing this in classical computation that I know of, but there is a method that can be used in quantum computing. For this to work, you need to have a quantum channel to the remote machine (i.e. an internet-type link over which you can send qubits). This method is known as blind quantum computing, and is very much in an active area of research.

Concerning error correction, or the lack thereof

You are right in saying that current quantum computers have too few qubits to perform active reliable error correction. This is true for both local and remote quantum computers. However, the answers of most computations that quantum computers perform are relatively easy to test: they are hard to compute, but easy to assert. This means that one just tests the answer returned by the remote machine and through that determines the reliability of the remote machine.

Furthermore, there exist types of computations that are less affected by the current noise levels of qubits; see for instance VQE's. These types of algorithms are more readily implemented on current hardware.

$^{1}$ I believe that this is the case, I am not an expert on the matter.


Posted 2019-11-20T14:38:45.667

Reputation: 3 248