All current programmable quantum processors are made from unicorn hairs and fairy dust.
The current state of the art is trying to prove that the thing did something quantum. Programmability is at best swapping some wires around to change that something.
Programming current quantum computers does not require swapping wires around. You can go to https://quantum-computing.ibm.com right now, drag some operations around in an editor, and have their quantum computer run those operations. No one is madly dashing around changing wires when you do that.
That was how they did it back in the old days though.
Q: "Ask HN: What's the Equivalent of 'Hello, World' for a Quantum Computer?" https://news.ycombinator.com/item?id=22707580 [ IBM Qiskit, Microsoft Q#, Google TFQ TensorFlow Quantum, Google Cirq ([NumFOCUS,] SymPy) ]
Not if you try to simulate the actual device under test, and not an idealized version of what it isn't. Nothing suspicious about it, they're small and noisy. Scalability is extremely hard for quantum computers, and there's distinct tradeoffs that need to be made in their engineering.
As far as I know, nobody in the industry is making absurd claims about their current offerings. That said, some claims of projected growth I've seen appear to be batshit.
And, I believe that you're wrong: Google's quantum supremacy result appears to be genuine, and IBM's refutation was, essentially, "we can use a ginormous supercomputer to match that" -- not your ordinary desktop.
46 qubits of a perfectly ideal quantum computer require a petabyte of RAM to simulate, and about 10 quadrillion—if not more—arithmetic instructions to perform a single quantum instruction.
This is false. Quantum computers are not programmed like telephone switch boards. They're programmed typically by sending signal pulses (of RF, DC, or light) of the right shape at the right time to physical elements and they react. It really is programming.
Moreover, it's remarkably easy to see that the machines—universal gate-based machines—are doing something quantum. What's not easy to see is if they'll ever break away from their scaling challenges and become useful, large scale machines.
Well, NMR has been doing that for quite some time but nobody considers it "computing". Realistically, you're encoding a problem within a physical system by perturbing its energy levels, letting them evolve, and then doing readout. This works because computation is universal.
Again, no. Maybe you're thinking about adiabatic quantum computation, something popularized by D-Wave. But most people in the industry don't call their machines "(universal) computers," but rather "quantum annealers."
This is not the same as universal, gate-based computation, which does take an actual program containing instructions, and executes those instructions more-or-less sequentially, just as any computer programmer would expect. The state of the system begins with what is essentially equivalent to a large array of 0's, and you manipulate it accordingly, without resorting to evolution as your primary computational means of marching forward. In fact, left alone, these computers are designed to remain static, like an ordinary computer. (However, they still couple with the environment and decohere, which is one of the major challenges we, as humanity, face in building a useful quantum computer.) This is what Rigetti, IBM, Google, HRL, Amazon, IonQ, ColdQuanta, etc. are doing. They use superconducting transmon qubits, ion qubits, neutral atom qubits, or silicon quantum dot qubits. (There are other companies and qubit technologies still.)
Hey, um, I think I do know what I'm talking about :). You can see more of this historical side effort, which didn't pan out for a number of reasons: https://en.wikipedia.org/wiki/Nuclear_magnetic_resonance_qua...
A 7 qubit QC NMR was implemented in 2001.
All these systems are just evolution of spin systems or other similar systems. The big difference with an NMR quantum computer is that it manipulates ensembles of spins.
The comment about universality of computing is that many physical processes can be used to compute things. I didn't say that qcs were universal computers.
Please try to read what i'm writing more carefully.
I know what NMR is. The systems are not evolutions of NMR. They're fundamentally different in their design, construction, and physics. An exchange-only silicon dot quantum computer is very different than an NMR quantum computer, which is very different than a neutral atom quantum computer.
These computers, these days, are by and large considered computers which execute programs. Perhaps in the 90s-00s that wasn't the case since physicists cared more then about the construction of a laboratory apparatus and less about turning it into a programmable device with program and data I/O.
I can't comment on what you know or don't know, but what you're writing is misleading and not accurate. The two points you've made (about what's considered computing and how modern quantum computers work) are what I refute.
> Among semiconductor qubits, the electron and nuclear spins of donors in silicon play a special role for their conceptual simplicity (a 31P donor in silicon is similar to hydrogen in vacuum) and their exceptional coherence times [1] and 1-qubit gate fidelities [2]. Here I will present experimental progress on multi-qubit logic operations with donor spins, which point to several credible pathways for scalability using ion-implanted donors in MOS-compatible devices. The current state of the art is a hybrid electron-nuclear 3-qubit processor [3], where two 31P nuclear spin qubits are coupled to the same electron. The shared electron enables a geometric nuclear two-qubit CZ gate, which we perform with 99.37% average fidelity. NMR single-qubit gates reach fidelities up to 99.95%, and state preparation and measurement are performed with 98.95% fidelity. These three metrics show how close this system is to operating at fault-tolerance thresholds. Further, we entangle the two nuclei with the electron to prepare a 3-qubit GHZ state with 92.5% fidelity. Electron-nuclear entanglement unlocks the ability to connect nuclear qubits via the electrons, for instance using exchange interactions [4]. We have operated a weakly (~10 MHz) exchange-coupled 31P donor pair as a 2-qubit electron system, with native CROT gates performed by resonant microwaves. Gate fidelity benchmarks are underway and will be reported at the Meeting. On the engineering side, we have demonstrated the ability to implant single donors in silicon with confidence up to 99.85% [5]. This striking result identifies ion implantation as a scalable and accurate manufacturing strategy for spin-based quantum computers in silicon.
QoS: Quantum-on-Silicon
The survey article above just says "[Quantum] Output"? Is that different from registers? How long are those states ah coherent?
> By managing to store a qubit in a crystal (a "memory") for 20 milliseconds, a team from the University of Geneva (UNIGE) has set a world record and taken a major step towards the development of long-distance quantum telecommunications networks.
What are repeaters, and what are [quantum] prepared states in re: registers and longer-term storage for non-collapsed (or just probabilistic?) qubit outputs?
I don't understand your point. Everything I said above is completely and totally technically correct from a physical point of view. You are ascribing to my statements meaning which I did not intend.
Modern trapped ion computers are doing the same underlying operations as NMR: you are using RF or other energy to perturb the energy states of underlying particles or other components, and reading out the results.
There was also a whole field called 'dna computing' which you would call biologists in a lab, but they were absolutely doing computing
I'm intently curious what sort of common-sense explanation you'd ascribe to how one programs an ordinary classical computer. Would you describe a computer programmer as one who orchestrates a delicate movement of charge through an intricate arrangement of n–p–n bipolar transistors?
Quantum computers really can be instructed to do an abstract operation on an abstract quantity, just like I can on an ordinary computer. You tell quantum computers to add and multiply, just as you do in your favorite programming language. In the programming language Quil, which a couple quantum computers of different base technologies use, one can literally write down a matrix of numbers in a program to define a new mathematical function, and later use that function to manipulate the contents of the computer's RAM[1]. On top of this, you can have loops, boolean conditions, and all that jazz a programmer expects. You don't even need a physicist to do this; a completely physics-ignorant programmer could do this.
All this business about energy levels, evolution, etc. are distractions, just as the electrodynamics of a transistor are distractions from what it means to program a computer.
[1] I'm abusing the word "RAM" here, where I truly mean the state of your quantum register(s).
> one can literally write down a matrix of numbers in a program to define a new mathematical function, and later use that function to manipulate the contents of the computer's RAM[1].
This "Quantum Computing for Computer Scientists" video https://youtu.be/F_Riqjdh2oM explains classical and quantum operators as just matrices. What are other good references?
> All this business about energy levels, evolution, etc. are distractions, just as the electrodynamics of a transistor are distractions from what it means to program a computer.
But a classical simulator - like e.g. qiskit - for a quantum circuit/experiment/function must run the experiment very^very^very many times to even probabilistically approximate a sufficient quantum system; because of the combinatorial probabilistic explosion that results from adding just one more basis state.
What are the fundamental limitations of quantum simulators? Maybe it's possible.
Your assessment of how a quantum
simulator works is not quite right. These simulators represent the entire probability distribution of basis states succinctly as an array. This array grows very large (exponentially) in the number of qubits.
A simulator only needs to run a computation once (which is multiplication of matrices in a tensor product space) and look at the resulting state. You don't need to run anything multiple times to approximate a quantum state.
The questions you're asking are the whole point of the field of quantum information science. On an ordinary quantum computer where quantum state will collapse to a basis state upon readout, indeed you might need to gather statistics to determine the answer to whatever you've asked your computer. However, "very many times" is mathematically bounded in some way for an ideal quantum computer. It's like saying "we need to do many^many^many^many comparisons to do quicksort". Well yes, but we have a relationship between the size of the input (N) and the average number of comparisons needed (N log N), which makes the algorithm feasible in practice. This is the same with quantum algorithms.
There are also different kinds of quantum algorithms. Some are more probabilistic in nature. Others are—again in purely ideal circumstances—give you the right answer in one go.
As a side note: It is very hard for me to read, understand, and respond to your comments. They seem like random buzzword soups and aren't very coherently put together, mixed with random links and references.
> Peter Shor first discovered this method of formulating a quantum error correcting code by storing the information of one qubit onto a highly entangled state of nine qubits. A quantum error correcting code protects quantum information against errors of a limited form.
> Quantum Monte Carlo encompasses a large family of computational methods whose common aim is the study of complex quantum systems. One of the major goals of these approaches is to provide a reliable solution (or an accurate approximation) of the quantum many-body problem. [...] The difficulty is however that solving the Schrödinger equation requires the knowledge of the many-body wave function in the many-body Hilbert space, which typically has an exponentially large size in the number of particles. Its solution for a reasonably large number of particles is therefore typically impossible,
What sorts of independent states can or should we map onto error-corrected qubits in an approximating system?
HRL Laboratories is very active in the field (if you happened to catch them at APS March Meeting) and hasn't been public since their founding in the 60s. It's one of the most respected private research laboratories of the United States. (It's currently jointly owned by Boeing and General Motors, but enjoys decision-making autonomy.)
Frankly, they had no other choice: public funding for quantum computing research is unacceptably low in the US. While China is funneling $10B into a single quantum lab [1], the US only spent $800M in the whole area for 2021 [2].
I fear that we're repeating the same mistake we made with Machine Learning funding. A decade ago, the US was quite clearly ahead of China in machine learning, but they're comparable now. In quantum information science, the US is currently ahead of China in terms of research. But will it remain that way?
For more information, here's a panel discussion on the quantum computing race between China and the US (I was quite frustrated by it):
The article mention that NSA have objected to quantum key distribution, then go on to state: However, if you’re a commercial company these systems may be worth exploring. The problem that NSA and many others have pointed out is that it doesn't solve any actual problems. It creates a bunch of hassles but can't replace asymmetric cryptography.
Correctly implemented QKD gives you key distribution without assumptions about how difficult certain mathematical problems are in relation to how much compute your adversary has. Key distribution is nowadays done with assymetric cryptography, so QKD can replace some assymmetric cryptography. You can also have authentication (Wegman-Carter) with symmetric keys. What's not quite clear is how you would do certificates and PKI. However, given key distribution, you could probably use symmetric keys for that as well.
It's unlikely that your adversaries can decrypt your traffic right now (break things like RSA). However, advances in number theory and/or computing power might enable them to do that in the future. Your adversary can just record your encrypted traffic and wait until the means to decrypt it become available. Thus, for data that has to stay secure for a long time (and where you want to be as sure as possible that it will) it's not good to rely on predictions into the future about advances in number theory or computing. This is the niche that QKD is aiming at.
For what it's worth, China has a huge QKD network, which cost them a lot of money. Their QKD satellite also cost a lot of money. They are in fact world leaders in quantum communication technology as well and spend a lot on researching it. I wonder why they made this investment, whether it was smart, and what they get out of it.
I also have doubts that QKD will see much use in the coming decades and even more doubts that its use will be done properly and actually make a lot of systems more secure. Securing systems is very hard and securing individual communication links (what QKD does) is not the main problem. In the current landscape, securing your data and communications to a reasonable level just isn't worth it for the vast majority of buisnesses, since they can offload most of the damages of being breached to their customers. There is a danger that QKD will be seen as "magic fairy dust" that you sprinkle over your systems just to claim you're trying very hard to secure them (this image is still widespread about standard cryptography as well).
Messages sent using classical crypto should be viewed as being public after an unknown delay. They can be decoded at your adversary's leisure with techniques and equipment invented in the future.
Quantum cyrpto must be broken immediately to be broken at all.
If what you are encrypting is, for example, credit card information, it's perfectly fine if that becomes public in a decade. Your information will have changed.
If what you are encrypting needs to remain secret for the next fifty years, do not use classical encyrption and a public channel. It may well be made public while the information is still sensitive. This is why QKD has some early adopters. It's the only long-term secure alternative to having people carry one time pad's back and forth in suitcases full of hard drives, which has its own security issues.
For anyone interested, IBM quantum experience lets you simulate quantum computer operation, and even gives an allowance of running actual simple programs, made in their visual tool “composer”, on a real quantum device, through the api. Signup required for some things.
https://quantum-computing.ibm.com/composer/files/new
You can also sign up for D-Wave Leap and use our online IDE, problem visualizer, and submit problems to our QPUs with a free signup. https://cloud.dwavesys.com/leap/
This is somewhat frustrating. I think everyone is interested in alternatives to IBM's offering, but putting it behind account sign-up makes it hostile to 'just checking it out'. I believe this is also why IBM allow some of their features to be accessed without an account.
Please remove the account requirement if possible. I will happily check it out once the account requirement is removed.
32 comments
[ 663 ms ] story [ 1942 ms ] threadCategory:Quantum mechanics https://en.wikipedia.org/wiki/Category:Quantum_mechanics
Applications_of_quantum_mechanics https://en.wikipedia.org/wiki/Applications_of_quantum_mechan...
List of emerging technologies https://en.wikipedia.org/wiki/List_of_emerging_technologies may have inspiration for applications of currently-discovered quantum mechanical phenomena.
#Q12 is the Quantum K12 talent.
QoS: Quantum-on-Silicon may very well scale; but how can we store un-collapsed output qubits?
Quantum tagging,
Aren't all current programmable quantum processors essentially made of qubit memory cells capable of certain in-memory operations?
The current state of the art is trying to prove that the thing did something quantum. Programmability is at best swapping some wires around to change that something.
Q: "Ask HN: What's the Equivalent of 'Hello, World' for a Quantum Computer?" https://news.ycombinator.com/item?id=22707580 [ IBM Qiskit, Microsoft Q#, Google TFQ TensorFlow Quantum, Google Cirq ([NumFOCUS,] SymPy) ]
- https://www.tensorflow.org/quantum/tutorials/hello_many_worl...
A: Set a register to zero and see how many times it reads as zero: A) in the local software simulator; and B) with just one modern day qubit register.
As far as I know, nobody in the industry is making absurd claims about their current offerings. That said, some claims of projected growth I've seen appear to be batshit.
And, I believe that you're wrong: Google's quantum supremacy result appears to be genuine, and IBM's refutation was, essentially, "we can use a ginormous supercomputer to match that" -- not your ordinary desktop.
Moreover, it's remarkably easy to see that the machines—universal gate-based machines—are doing something quantum. What's not easy to see is if they'll ever break away from their scaling challenges and become useful, large scale machines.
This is not the same as universal, gate-based computation, which does take an actual program containing instructions, and executes those instructions more-or-less sequentially, just as any computer programmer would expect. The state of the system begins with what is essentially equivalent to a large array of 0's, and you manipulate it accordingly, without resorting to evolution as your primary computational means of marching forward. In fact, left alone, these computers are designed to remain static, like an ordinary computer. (However, they still couple with the environment and decohere, which is one of the major challenges we, as humanity, face in building a useful quantum computer.) This is what Rigetti, IBM, Google, HRL, Amazon, IonQ, ColdQuanta, etc. are doing. They use superconducting transmon qubits, ion qubits, neutral atom qubits, or silicon quantum dot qubits. (There are other companies and qubit technologies still.)
All these systems are just evolution of spin systems or other similar systems. The big difference with an NMR quantum computer is that it manipulates ensembles of spins.
The comment about universality of computing is that many physical processes can be used to compute things. I didn't say that qcs were universal computers.
Please try to read what i'm writing more carefully.
These computers, these days, are by and large considered computers which execute programs. Perhaps in the 90s-00s that wasn't the case since physicists cared more then about the construction of a laboratory apparatus and less about turning it into a programmable device with program and data I/O.
I can't comment on what you know or don't know, but what you're writing is misleading and not accurate. The two points you've made (about what's considered computing and how modern quantum computers work) are what I refute.
"Multi-qubit quantum logic operations with ion-implanted donor spins in silicon" (2022) https://scholar.google.com/citations?view_op=view_citation&h... https://meetings.aps.org/Meeting/MAR22/Session/G39.1 (Veritasium video)
> Among semiconductor qubits, the electron and nuclear spins of donors in silicon play a special role for their conceptual simplicity (a 31P donor in silicon is similar to hydrogen in vacuum) and their exceptional coherence times [1] and 1-qubit gate fidelities [2]. Here I will present experimental progress on multi-qubit logic operations with donor spins, which point to several credible pathways for scalability using ion-implanted donors in MOS-compatible devices. The current state of the art is a hybrid electron-nuclear 3-qubit processor [3], where two 31P nuclear spin qubits are coupled to the same electron. The shared electron enables a geometric nuclear two-qubit CZ gate, which we perform with 99.37% average fidelity. NMR single-qubit gates reach fidelities up to 99.95%, and state preparation and measurement are performed with 98.95% fidelity. These three metrics show how close this system is to operating at fault-tolerance thresholds. Further, we entangle the two nuclei with the electron to prepare a 3-qubit GHZ state with 92.5% fidelity. Electron-nuclear entanglement unlocks the ability to connect nuclear qubits via the electrons, for instance using exchange interactions [4]. We have operated a weakly (~10 MHz) exchange-coupled 31P donor pair as a 2-qubit electron system, with native CROT gates performed by resonant microwaves. Gate fidelity benchmarks are underway and will be reported at the Meeting. On the engineering side, we have demonstrated the ability to implant single donors in silicon with confidence up to 99.85% [5]. This striking result identifies ion implantation as a scalable and accurate manufacturing strategy for spin-based quantum computers in silicon.
QoS: Quantum-on-Silicon
The survey article above just says "[Quantum] Output"? Is that different from registers? How long are those states ah coherent?
"Researchers store a quantum bit for a record-breaking 20 milliseconds" (2022) https://phys.org/news/2022-03-quantum-bit-fora-record-breaki...
> By managing to store a qubit in a crystal (a "memory") for 20 milliseconds, a team from the University of Geneva (UNIGE) has set a world record and taken a major step towards the development of long-distance quantum telecommunications networks.
What are repeaters, and what are [quantum] prepared states in re: registers and longer-term storage for non-collapsed (or just probabilistic?) qubit outputs?
Modern trapped ion computers are doing the same underlying operations as NMR: you are using RF or other energy to perturb the energy states of underlying particles or other components, and reading out the results.
There was also a whole field called 'dna computing' which you would call biologists in a lab, but they were absolutely doing computing
Quantum computers really can be instructed to do an abstract operation on an abstract quantity, just like I can on an ordinary computer. You tell quantum computers to add and multiply, just as you do in your favorite programming language. In the programming language Quil, which a couple quantum computers of different base technologies use, one can literally write down a matrix of numbers in a program to define a new mathematical function, and later use that function to manipulate the contents of the computer's RAM[1]. On top of this, you can have loops, boolean conditions, and all that jazz a programmer expects. You don't even need a physicist to do this; a completely physics-ignorant programmer could do this.
All this business about energy levels, evolution, etc. are distractions, just as the electrodynamics of a transistor are distractions from what it means to program a computer.
[1] I'm abusing the word "RAM" here, where I truly mean the state of your quantum register(s).
This "Quantum Computing for Computer Scientists" video https://youtu.be/F_Riqjdh2oM explains classical and quantum operators as just matrices. What are other good references?
Quantum state: https://en.wikipedia.org/wiki/Quantum_state
Quantum logic; quantum logical operators: https://en.wikipedia.org/wiki/Quantum_logic
> All this business about energy levels, evolution, etc. are distractions, just as the electrodynamics of a transistor are distractions from what it means to program a computer.
But a classical simulator - like e.g. qiskit - for a quantum circuit/experiment/function must run the experiment very^very^very many times to even probabilistically approximate a sufficient quantum system; because of the combinatorial probabilistic explosion that results from adding just one more basis state.
What are the fundamental limitations of quantum simulators? Maybe it's possible.
Quantum simulator: https://en.wikipedia.org/wiki/Quantum_simulator
- [ ] Maybe Twistor theory has insight into a classical geometrical formulation that could be run on a non-QC?
Amplituhedron: https://en.wikipedia.org/wiki/Amplituhedron
[Photon] wave-particle constructive superpositions approximate which operators, which may form a neat topology like this:
- [ ] > A research question for a new school year:
> The classical logical operators form a neat topology. Should we expect there to be such symmetry and structure amongst the quantum operators as well? https://commons.m.wikimedia.org/wiki/File:Logical_connective...
A simulator only needs to run a computation once (which is multiplication of matrices in a tensor product space) and look at the resulting state. You don't need to run anything multiple times to approximate a quantum state.
The questions you're asking are the whole point of the field of quantum information science. On an ordinary quantum computer where quantum state will collapse to a basis state upon readout, indeed you might need to gather statistics to determine the answer to whatever you've asked your computer. However, "very many times" is mathematically bounded in some way for an ideal quantum computer. It's like saying "we need to do many^many^many^many comparisons to do quicksort". Well yes, but we have a relationship between the size of the input (N) and the average number of comparisons needed (N log N), which makes the algorithm feasible in practice. This is the same with quantum algorithms.
There are also different kinds of quantum algorithms. Some are more probabilistic in nature. Others are—again in purely ideal circumstances—give you the right answer in one go.
As a side note: It is very hard for me to read, understand, and respond to your comments. They seem like random buzzword soups and aren't very coherently put together, mixed with random links and references.
How to best quantize reals into matrices (~= tensors)? https://en.wikipedia.org/wiki/Quantization_(signal_processin...
> Peter Shor first discovered this method of formulating a quantum error correcting code by storing the information of one qubit onto a highly entangled state of nine qubits. A quantum error correcting code protects quantum information against errors of a limited form.
Here's "Quantum Algorithm Zoo" by Microsoft Quantum: https://quantumalgorithmzoo.org/
And "Timeline of quantum computing and communication" https://en.wikipedia.org/wiki/Timeline_of_quantum_computing_...
I have a hard time with the idea that the outcome of the ultimate quantum simulation is a collapsed float.
Quantum Monte Carlo: https://en.wikipedia.org/wiki/Quantum_Monte_Carlo :
> Quantum Monte Carlo encompasses a large family of computational methods whose common aim is the study of complex quantum systems. One of the major goals of these approaches is to provide a reliable solution (or an accurate approximation) of the quantum many-body problem. [...] The difficulty is however that solving the Schrödinger equation requires the knowledge of the many-body wave function in the many-body Hilbert space, which typically has an exponentially large size in the number of particles. Its solution for a reasonably large number of particles is therefore typically impossible,
What sorts of independent states can or should we map onto error-corrected qubits in an approximating system?
Propagation of Uncertainty ... Numerical stability ... Chaotic convergence, ultimately, apparently: https://en.wikipedia.org/wiki/Propagation_of_uncertainty
I fear that we're repeating the same mistake we made with Machine Learning funding. A decade ago, the US was quite clearly ahead of China in machine learning, but they're comparable now. In quantum information science, the US is currently ahead of China in terms of research. But will it remain that way?
For more information, here's a panel discussion on the quantum computing race between China and the US (I was quite frustrated by it):
https://youtu.be/KzFEeQ49HHI
[1] - https://english.ckgsb.edu.cn/knowledges/quantum-wars/
[2] - https://quantumcomputingreport.com/u-s-qis-budget-proposed-t...
It's unlikely that your adversaries can decrypt your traffic right now (break things like RSA). However, advances in number theory and/or computing power might enable them to do that in the future. Your adversary can just record your encrypted traffic and wait until the means to decrypt it become available. Thus, for data that has to stay secure for a long time (and where you want to be as sure as possible that it will) it's not good to rely on predictions into the future about advances in number theory or computing. This is the niche that QKD is aiming at.
For what it's worth, China has a huge QKD network, which cost them a lot of money. Their QKD satellite also cost a lot of money. They are in fact world leaders in quantum communication technology as well and spend a lot on researching it. I wonder why they made this investment, whether it was smart, and what they get out of it.
I also have doubts that QKD will see much use in the coming decades and even more doubts that its use will be done properly and actually make a lot of systems more secure. Securing systems is very hard and securing individual communication links (what QKD does) is not the main problem. In the current landscape, securing your data and communications to a reasonable level just isn't worth it for the vast majority of buisnesses, since they can offload most of the damages of being breached to their customers. There is a danger that QKD will be seen as "magic fairy dust" that you sprinkle over your systems just to claim you're trying very hard to secure them (this image is still widespread about standard cryptography as well).
Messages sent using classical crypto should be viewed as being public after an unknown delay. They can be decoded at your adversary's leisure with techniques and equipment invented in the future.
Quantum cyrpto must be broken immediately to be broken at all.
If what you are encrypting is, for example, credit card information, it's perfectly fine if that becomes public in a decade. Your information will have changed.
If what you are encrypting needs to remain secret for the next fifty years, do not use classical encyrption and a public channel. It may well be made public while the information is still sensitive. This is why QKD has some early adopters. It's the only long-term secure alternative to having people carry one time pad's back and forth in suitcases full of hard drives, which has its own security issues.
Just add one bit and make it 100 years. By that time, nobody will care really.
Please remove the account requirement if possible. I will happily check it out once the account requirement is removed.