Is there any reason to think breakthroughs like this and similar could one day lead to quantum computers being applied to general computation? It's been explained to me that there's no point in asking questions like "when will quantum computers be able to run Doom?" because they will never be used for non-specialized tasks.
Making non-quantum computer components faster is already a problem that involves quantum physics; quantum computers are faster at specific algorithms, therefore it only really makes sense to use them for the things they are faster for.
> Is there any reason to think breakthroughs like this and similar could one day lead to quantum computers being applied to general computation?
None whatsoever.
After learning a bit about them, I think of quantum computers as being a bit like GPUs: in theory you can do anything with them, but in practice, getting things done with them is hard enough that it’s best to use them for what they’re best for. You can certainly run classical algorithms on a quantum computer, but they probably wouldn’t perform any better than the equivalent on a classical computer; the only real advantage of quantum computers is that they let you run quantum algorithms to give you an exponential speedup on certain problems.
Quantum computations (the computer science concept) is not needed for operations where the classical algorithms are already efficient.
Quantum computer technology used to perform classical computations - almost certainly there will be some engineering considerations where this will make sense.
There are a few abstract, mathematical proofs that a theoretical quantum computer could be faster at certain specific specialized tasks (such as prime number factoring, which has important implications for cryptography). No one has actually really built such a quantum computer yet, it remains a theoretical idea. The theory of quantum physics itself is also incomplete. So our knowledge of the universe and the possibilities it contains is very limited. Someone who says "never" is operating from a mindset where they equate the whole of reality with their current, limited knowledge of the universe. I don't think it's far fetched to imagine that our understanding of quantum physics might one day lead to new breakthroughs in general computation. Given that silicon processors are already near the quantum limits, there would be a huge incentive to find a new breakthrough. That said, this is in the realm of pure science fiction and not something that people are actually working on today or something we are likely to see in our lifetimes.
>No one has actually really built such a quantum computer yet
What are you talking about? We've had very small quantum universal computers for some time, and quantum annealers for two decades. How did you miss that?
I regret to inform you that low bit strength RSA was even broken with a quantum annealer, not even a universal quantum machine. the NSA is a customer of D-Wave.
IBM's quantum roadmap has them shipping 1121 qubit machine next year, this year they've shipped a 400+ qubit machine. So far they've been able to keep doubling the number of qubits every year. These are universal quantum computers.
If they can keep scaling at that rate, both RSA and ECDSA are toast within a decade.
Here's a bit of an ELI5 about quantum computers, and why this type of news are not nearly as much of a big deal as they appear.
To create quantum computers you need to be able to manipulate small particles like electrons, or ions. You need to be able to "freeze" them in a "superposition" state. If you think of a particle as a tiny magnetic needle, you need to be able to make the particle be in a "half North, half South" state, like Schroedinger's cat, which is half dead, half alive. What this means, is that you don't know if the needle points North or South (the usual terms in quantum computing are "spin up" and "spin down"), and nobody can know that; but you know that if you perform a measurement you will have exactly 50% chance to be North and 50% chance to be South. And by "manipulate", I mean, you can choose different such probabilities, for example 70% and 30%.
So far, so good, nothing too interesting happens.
But then, if you want quantum computation, you need to manipulate more than one particle. Let's say you want to manipulate 2 particles.
Now you can say, if I can freeze one particle, then I can freeze 2, what's the big deal. The big deal, is that the concept of freezing 2 particles involves "entanglement", or, as Einstein called it, "spooky action at a distance". For example, not only are you able to arrange then so that each has a probability to be, let's say 70% spin up and 30% spin down, but you can arrange them so that they always point in the same direction. If you measure one and you get up, then you know for sure that the other one will be up.
Now, this entanglement can lead to wonderful things, when it comes to computation. If you are smart and know how to entangle, let's say about 2000 particles, then you can factor large composite numbers or about 1000 bits length. And some people think this is a big deal (because of breaking passwords and stuff). Lots of people think it's not such a big deal after all, but let's not dwell on that.
Now, here's the problem. Entangling many particles is difficult. If you are able to entangle 5 particles, going to 6 is going to be difficult. But going from 6 to 7 is going to be about as difficult, and then from 7 to 8.
Here's [1] a link about IBM creating a quantum computer with 127 qubits (particles). They claim they'll achieve 1000 qubits sometime next year. That seems to contradict what I said in the paragraph above. Well, you can achieve more qubits if you relax what you call "entanglement", but then your algorithms will not perform as well, so it's not clear you can still achieve the exponential speedup needed to factor large numbers.
Anyway, back to the principles of quantum computation. The way you perform a computation is by first "freezing" or "entangling" the qubits, let's say you make them all fully entangled to point up with 50% probability. If you measure one, and you get up, you know all the others are up.
Now, here's the big deal: you can perform changes to this ensemble of qubits, while keeping them entangled. You can "rotate" them. And if your rotation is judiciously chosen, then when you perform the measurement, you'll get a 50% chance to get, let's say up, up, down, up, down, etc, or 11010... where this binary number gives you the answer to a problem you care about. Of course, you'll get a 50% chance to get a different number that's garbage, but you don't mind, because you can check your solution, and you can repeat the computation several times.
But the number of problems where such a smart rotation is possible is not that large. Many times you don't get an exponential speedup, but just a quadratic one.
The case with an exponential speedup is called quantum Fourier transform. If you can reduce a problem to the calculation of a quantum Fourier transform, then you are all set, you can achieve exponential sp...
Is there any binding force between neighboring Neon atoms? Or do you make the "solid" by just bunching them together until they have no place to go (like stacking oranges)?
layman question from me would be whether neon acts like a superfluid the way helium does at those temperatures, and whether that property is part of what makes this approach different.
These are called London Dispersion forces. You can think of it somewhat like an electrostatic interaction except on a “temporary” basis, like fluctuations in the electron densities around nuclei spontaneously cause weak attraction between these atoms. The weakness of the attraction of these forces is why Ne atoms are so easily broken by thermal forces, ie why Ne is a monoatomic gas.
So yes, there is a binding force, but it is very weak.
I wonder whether cubits made of Bose-Einstein condensates would be more stable than what they are trying to make them out of now. It seems like decohering a collection of hundreds or thousands of atoms sharing a single state would take a lot more perturbation than single atoms or particles.
17 comments
[ 3.4 ms ] story [ 95.8 ms ] threadNone whatsoever.
After learning a bit about them, I think of quantum computers as being a bit like GPUs: in theory you can do anything with them, but in practice, getting things done with them is hard enough that it’s best to use them for what they’re best for. You can certainly run classical algorithms on a quantum computer, but they probably wouldn’t perform any better than the equivalent on a classical computer; the only real advantage of quantum computers is that they let you run quantum algorithms to give you an exponential speedup on certain problems.
Change is often sudden and unexpected :-)
Quantum computer technology used to perform classical computations - almost certainly there will be some engineering considerations where this will make sense.
What are you talking about? We've had very small quantum universal computers for some time, and quantum annealers for two decades. How did you miss that?
I regret to inform you that low bit strength RSA was even broken with a quantum annealer, not even a universal quantum machine. the NSA is a customer of D-Wave.
IBM's quantum roadmap has them shipping 1121 qubit machine next year, this year they've shipped a 400+ qubit machine. So far they've been able to keep doubling the number of qubits every year. These are universal quantum computers.
If they can keep scaling at that rate, both RSA and ECDSA are toast within a decade.
To create quantum computers you need to be able to manipulate small particles like electrons, or ions. You need to be able to "freeze" them in a "superposition" state. If you think of a particle as a tiny magnetic needle, you need to be able to make the particle be in a "half North, half South" state, like Schroedinger's cat, which is half dead, half alive. What this means, is that you don't know if the needle points North or South (the usual terms in quantum computing are "spin up" and "spin down"), and nobody can know that; but you know that if you perform a measurement you will have exactly 50% chance to be North and 50% chance to be South. And by "manipulate", I mean, you can choose different such probabilities, for example 70% and 30%.
So far, so good, nothing too interesting happens.
But then, if you want quantum computation, you need to manipulate more than one particle. Let's say you want to manipulate 2 particles.
Now you can say, if I can freeze one particle, then I can freeze 2, what's the big deal. The big deal, is that the concept of freezing 2 particles involves "entanglement", or, as Einstein called it, "spooky action at a distance". For example, not only are you able to arrange then so that each has a probability to be, let's say 70% spin up and 30% spin down, but you can arrange them so that they always point in the same direction. If you measure one and you get up, then you know for sure that the other one will be up.
Now, this entanglement can lead to wonderful things, when it comes to computation. If you are smart and know how to entangle, let's say about 2000 particles, then you can factor large composite numbers or about 1000 bits length. And some people think this is a big deal (because of breaking passwords and stuff). Lots of people think it's not such a big deal after all, but let's not dwell on that.
Now, here's the problem. Entangling many particles is difficult. If you are able to entangle 5 particles, going to 6 is going to be difficult. But going from 6 to 7 is going to be about as difficult, and then from 7 to 8.
Here's [1] a link about IBM creating a quantum computer with 127 qubits (particles). They claim they'll achieve 1000 qubits sometime next year. That seems to contradict what I said in the paragraph above. Well, you can achieve more qubits if you relax what you call "entanglement", but then your algorithms will not perform as well, so it's not clear you can still achieve the exponential speedup needed to factor large numbers.
Anyway, back to the principles of quantum computation. The way you perform a computation is by first "freezing" or "entangling" the qubits, let's say you make them all fully entangled to point up with 50% probability. If you measure one, and you get up, you know all the others are up.
Now, here's the big deal: you can perform changes to this ensemble of qubits, while keeping them entangled. You can "rotate" them. And if your rotation is judiciously chosen, then when you perform the measurement, you'll get a 50% chance to get, let's say up, up, down, up, down, etc, or 11010... where this binary number gives you the answer to a problem you care about. Of course, you'll get a 50% chance to get a different number that's garbage, but you don't mind, because you can check your solution, and you can repeat the computation several times.
But the number of problems where such a smart rotation is possible is not that large. Many times you don't get an exponential speedup, but just a quadratic one.
The case with an exponential speedup is called quantum Fourier transform. If you can reduce a problem to the calculation of a quantum Fourier transform, then you are all set, you can achieve exponential sp...
https://en.m.wikipedia.org/wiki/Supersolid
So yes, there is a binding force, but it is very weak.