-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Nature: Quantum computers ready to leap out of the lab in 2017
- Thread starter Kinitari
- Start date
- Status
Sure to be a success in at least some of all possible universes.
E-Cat
Member
A universe where Microsoft is cool? Sounds optimistic.
I was about to post this, it was a great read for me a while ago.
The reality of quantum computers is so different than what you percieve from reading popular mechanics esque articles. For people who are curious, this article really helped me out:
https://www.ias.edu/ideas/2014/ambainis-quantum-computing
SenjutsuSage
Banned
The better question would be: what the hell would you want to do with a personal quatum computer?
They work on a very different logic than current computers, and are theoretically only better at very specific thing normal people shouldn't have any interest in. It won't run Windows or your Internet browser faster.
chillybright
Banned
Let the hacking wars begin.
hopefully for rendering/lighting solutions. Particle simulations... Cloth simulations... things like that.
Epsilon-delta
Banned
I wouldn't mind AI taking over our economic and social policies.
FunkyPajamas
Member
Post of the thread right here. *therockclap.gif*
The difference is that a classical bit can have exactly 2 values - 0 and 1, whereas a qbit can contain any superposition of those two states. Common examples of qbits are things like particles with spin or polarised light. This all goes back to the basic premise of the wavefunction of a system being the superposition of all its possible states, with the function collapsing into one of its substates when you interact with the system.
One interesting thing about quantum computation is that factorisation problems, which are at the heart of encryption methods like Diffie-Hellman, are easily solvable (eg by using Shor's algorithm), so many classical encryption methods become effectively worthless once quantum computation is available.
archangelmorph
Member
Err I highly doubt that...
Getting quantum computers productized, matured and then miniaturised to the point of being able to slip a qCPU into your smart phone is probably a lifetime or two away.
Miniaturisation depends heavily on the physics involved, and so for say superconductive quantum computers for example, where qubits have to be stored at freezing temperatures, you have about as much chance of seeing a miniaturised processor, as you do seeing a functional tokomak fusion reactor that with fit in your remote controlled toy car.
Basically we're kinda reaching the point right now, where we were with silicon based processors in the the 60s-70s; where productization will be tantamount to machines occupying warehouse-sized server-spaces, with heavily enterprise-orientated service consumption.
It'll also take a very very long time for the utility of these things to become much more widely realised, with the early hardware models being useful for very very specialised, mostly scientific use cases. We'll need new data transformations and encodings, new machine code instruction sets, new high-level programming languages completely with novel idioms and paradigms, new development tools, protocols etc etc.
This is basically millions of man hours of research and development that needs to be spent even after the HW is functional and stable.
On a separate note, productization of quantum computing hardware has already started (MS, Google et al are already LTTP).
D Wave Systems already has a 16 qubit quantum computer available for highly specialised, commercial usage and has done for quite a few years.
paradise_circus
Member
Quantum "computers" for now use superconductivity afaik. I think you won't see liquid helium tanks at home for videogames.
Thanks Linkyn,
Question time:
1. When you say superposition, is that a range of values between 0 and 1. Is a 'fraction' in between a good way of thinking of it? Or are you saying that 0 and 1 can have multiple values, rather than our current two definite 0 or 1 values.
2. How would a quantum machine be utilized in a calculative capacity, that would be superior to today's traditional machines?
I'll LogOff and take my answer offline.
Thank you.
There is a GAFer that already can do that on his own.
Don't underestimate overclocking enthusiasts
I only wish to be a God for my civ 
archangelmorph
Member
1. It's not range, it basically means you can have a state of 1, a state of 0 or 1+0 at the same time.
2. There's some pretty interesting possibilities for specialised quantum algorithms, which exploit the behaviours of the hardware and enable you to reduce incredibly large, problem sets down to a solution super quickly. E.g. if a traditional algorithm wanted to figure out the optimal play in a game of chess, it could sequentially work through every permutation of every possible move and take an non-trivial amount of time (or non-trivial amount of computing resources) to process that information. Even a humble quantum computer with a dozen or so qubits, could basically calculate every possible permutation of every possible move all at the same time and provide you with an answer within fractions of a second.
chromatic9
Member
Only when you're not looking at it.
PeakPointMatrix
Member
Hi everyone, uninformed plebeian here.
I know people are joking about Crysis, but what does this actually for consumer stuff like gaming? Smartphones? Computers? Or is it as simple as the same stuff, but just really, really fast?
I know people are joking about Crysis, but what does this actually for consumer stuff like gaming? Smartphones? Computers? Or is it as simple as the same stuff, but just really, really fast?
1. In quantum theory, superposition refers to the idea that any linear combination of two valid states gives another valid state. What that means is that, since both 0 and 1 are valid states for a qbit, any linear combination of the two is also a valid state that the qbit can be in. You can think of it as having all of one, or all of the other, or a bit of both. The only basic restriction is that if the superposed state contains more of one state, it contains less of the other, so that the total probability of being in any one state is never more than unity (I'm sorry if this all sounds a bit technical). In essence, though, it means that the qbit can potentially be in any of an infinity of possible superposed states.
One big difficulty arises here because of readout. Even if a quantum system can exist in a superposition of its possible 'eigenstates' (1 and 0 in this case), whenever you interact with it, it collapses back into one of them. The nature of the superposed state determines how likely each outcome is, but a priori, it's impossible to predict with certainty which one it'll be (unless the system was already in one of its eigenstates). In practice, this means that whenever you read the value of the qbit, you risk losing information, so quantum computing systems and algorithms need to be created with that in mind.
2. Obvious advantages in computation are in problems that can be very iteration-heavy, like factorisation for instance. To me, the more interesting application lies in quantum key distribution, which uses entanglement to create and share private encryption keys.
SenjutsuSage
Banned
Wonder how many quantum computers it would take to equal one project scorpio? 2? 3?
RE: Quantum computers breaking encryption. While it's true that they can break lots of algorithms, there do exist a number of classical algorithms which appear to be quantum secure. Lattice-based cryptography for example.
Overall, the singular most valuable thing a quantum computer can do right now is serve as a perfect quantum simulator (ie it can efficiently simulate any quantum system). Quantum computers will likely never be more useful than classical computers outside of certain specific instances imo, for a generic problem quantum computers only offer a quadratic speedup (O(N) -> O(sqrt(N))). Given how difficult it is to create coherence between qubits I am sceptical there will exist a time in which a quadratic speedup is worth all the trouble.
Overall, the singular most valuable thing a quantum computer can do right now is serve as a perfect quantum simulator (ie it can efficiently simulate any quantum system). Quantum computers will likely never be more useful than classical computers outside of certain specific instances imo, for a generic problem quantum computers only offer a quadratic speedup (O(N) -> O(sqrt(N))). Given how difficult it is to create coherence between qubits I am sceptical there will exist a time in which a quadratic speedup is worth all the trouble.
Fair enough.
These aren't really applications most people would use.
No, it works on different concepts an current computers.
That means all the programmjng languages and compilers will have to be redone, on top of the fact that quantum computing is only really good at iterative tasks (e.g. factoring) due to the nature of qbits.
Even if the hardware is running, consumer non scientific applications are decades off. For the scientific work though, it should be huge, and lead to improvement in other areas (e.g. Far better weather forecast, better cryptography and so on)
Psycho_Mantis
Banned
Computational sciences have become a massive industry and dominate sectors such as aerospace, mechanical and finance engineering. Could quantum computers not provide a significant leap in available power for these fields?
NYCmetsfan
Banned
Idk why everyone is freaking out about encryption. Its seems people are already looking and have some answers.
https://en.wikipedia.org/wiki/Post-quantum_cryptography
http://www.newyorker.com/tech/elements/hacking-cryptography-and-the-countdown-to-quantum-computing
I also doubt this will be overnight.
https://en.wikipedia.org/wiki/Post-quantum_cryptography
http://www.newyorker.com/tech/elements/hacking-cryptography-and-the-countdown-to-quantum-computing
I also doubt this will be overnight.
Razorback
Member
Underrated post.
archangelmorph
Member
You don't even need to go that far...
Lamport signatures for example, as a rough and ready, efficient alternative to RSA, are quantum secure...
It could... if people develop algorithm tailored to them. Iirc, FEM simulations could be sped up significantly (or be made with much smaller elements), which should enable engineer to attain better result through better simulations.
Berserking Guts
Member
But Doctor. Can these... machines. Can they feel?
archangelmorph
Member
Ok.. I think I get it..
Today we have 0 and 1. With quantum computers you get 0 and 1, 0+1 and 1+0...to compute data....?
as far as the rest... I'll reply after I get my engineering degree... bbl... getting my GED...
maniac-kun
Member
Arent this Google systems just the D-Wave quantum computers? Which use https://en.wikipedia.org/wiki/Quantum_annealing instead of general quantum computing. Which is the reason they only are faster at encryption / searching databases and stuff.
djplaeskool
Member
Or better yet...
Not even quantum computers would make this fun to play.
But how to you build a quantum computer? I mean does it only need more power to generate more stuff faster ?
Unless many more quantum algorithms which offer an exponential speedup compared to the classic case are discovered then I am deeply sceptical. Basically it needs to be cheaper for a company to use a quantum computer than an ordinary one, if N^2 bits could do the same job as N qubits then I doubt that will ever be the case outside of a few special applications (CERN and spy agencies).
Google do own a D-Wave machine but this is there own internal team. They are working with John Martinis who is the best person at superconducting qubits in the world. The D-wave machine can't actually implement Shor's algorithm. A general quantum annealer could be universal but because of the limited connectivity of the D-Wave chip it isn't universal.
There are as many ways to build a quantum computer as there are quantum systems (although most of them are crap). What you basically need is 1. Something to be your qubit ie a two-level quantum system (for example, two excited energy levels of an electron or the nucleus of an atom) 2. Some way of getting many of these qubits to interact however you like (you want to be able to manipulate them arbitrarily so you can implement any algorithm and also "read out" the end result) 3. Some way of stopping these qubits interacting with the environment (quantum systems that interact with the environment lose some of their "quantumness" and cease to be useful). I'm sure you can see that the issue is that 2. and 3. sort of contradict, anything that you can easily manipulate with your laser or magnetic field or w/e will also interact with stray electric or magnetic fields in the environment. This is the main issue.
LiveFromKyoto
Member
Just tell me this is going to bring us closer to the singularity, which will solve all our problems and make me into a perfect cyborg person.
SomedayTheFire
Member
I wish I actually understood quantum computing
Can mods check insider credibility
DevilDog
Member
Don't worry, we don't understand it either. It just works. Somehow.
I am going to go over the concept of quantum superposition here since someone asked, and it is quiet a weird subject to understand. Furthermore the explanation that superposition of 0 and 1 is both at the same time is very misleading, so I am hoping to shed some light on it.
So for regular waves, superposition can add or cancel waves. So if you drip water on a water surface, and do the same on another place, the two places will have waves coming out of them, propagating outwards. When the waves intersect, you will see areas where they enhance each other or subtract each other. Quantum superposition is similar, but there is an additional weirdness to it.
So consider than a particle is in a state where it has a probability of 3/4 to be in energy 1 and 1/4 for energy 2, and let's ignore the specific numbers of what those energies are. Just know the probability of which energy the particle will end up (you can do it with position or momentum or other physical descriptions). So the state for energy 1 is denoted as |1> and for energy 2 as |2>, so |n> is some function (you can think of it being in terms of position if you want, but it can be any physical quantity) that has energy n. It gets stranger, those functions are complex functions, meaning they are in the form of a + i*b, where i is sqrt(-1). So how can you obtain actual probability for them? The rule is that you multiply the complex conjugate of the functions, meaning a - i*b form, denoted as <n|, and this gives you a real number. One final basics is that <1|2>=0, while <1|1>=1, the reason being that if you measure energy 1 during the experiment, obviously you couldn't have measured energy 2 at the same time. Remember, once a measurement is done, which is done via some physical interaction, things are on lock, and the whole superposition disappears. Knowing the basic rules, let's see what happens.
So the superposition is denoted as: |state 1> = sqrt(3/4)*|1> + sqrt(1/4)*|2>.
Why square root? Because once multiplied by the complex conjugate, you want the whole thing to have probability one (similar to saying that the probability of getting either heads or tails in a coin flip is 1):
(sqrt(3/4)*<1| + sqrt(1/4)*<2|)*(sqrt(3/4)*|1> + sqrt(1/4)*|2>) = 3/4*<1|1> + 1/4*<2|2> + sqrt(3/16)*<1|2> + sqrt(3/16)*<2|1> = 3/4 + 1/4 = 1
So what if you want to find the probability of finding energy 1? Look above and see where 3/4*<1|1> is? That number tells you the probability. So, you have got such a particle and measure it. Let's say you measure it and say get energy 1. Then you recreate a particle of the same state and measure it again, obtaining energy 2. Do the same thing 1000 times, and you observe that 756 times you observed energy 1 and 244 times you measured energy 2. Notice how while each individual measurement gives you one or the other, if you do it a lot of times, the total numbers becomes closer and closer to the probability number? That is similar to how a coin flip works, where you might get 5 heads in a row at the beginning, but when you do it a thousand times, the number of heads and tails obtained are almost shared equally.
But see, this is where it is not like a coin flip. A particle might have state: |state 2> = sqrt(3/4)*|1> - sqrt(1/4)*|2>.
But wait a minute, isn't it redundant since it gives you the same probabilities?! It does give you the same probability, but what if you prepare the following?:
|state 3> = A*|state 1> + B*|state 2> = (A + B)*sqrt(3/4)*|1> + (A - B)*sqrt(1/4)*|2>
Now the fact that there is a minus sign matters. Let's say A = B, just to make a really eye popping example: |state 3> = 2A*sqrt(3/4)*|1>, and A = 1/2*sqrt(4/3) to make the probability equal 1. So |state 3> = |1>, and because of this there is zero chance energy 2 can be measured!
So I know this is going to be quiet difficult to interpret, especially those without a math background, but hopefully it informs what they mean by quantum superposition.
So for regular waves, superposition can add or cancel waves. So if you drip water on a water surface, and do the same on another place, the two places will have waves coming out of them, propagating outwards. When the waves intersect, you will see areas where they enhance each other or subtract each other. Quantum superposition is similar, but there is an additional weirdness to it.
So consider than a particle is in a state where it has a probability of 3/4 to be in energy 1 and 1/4 for energy 2, and let's ignore the specific numbers of what those energies are. Just know the probability of which energy the particle will end up (you can do it with position or momentum or other physical descriptions). So the state for energy 1 is denoted as |1> and for energy 2 as |2>, so |n> is some function (you can think of it being in terms of position if you want, but it can be any physical quantity) that has energy n. It gets stranger, those functions are complex functions, meaning they are in the form of a + i*b, where i is sqrt(-1). So how can you obtain actual probability for them? The rule is that you multiply the complex conjugate of the functions, meaning a - i*b form, denoted as <n|, and this gives you a real number. One final basics is that <1|2>=0, while <1|1>=1, the reason being that if you measure energy 1 during the experiment, obviously you couldn't have measured energy 2 at the same time. Remember, once a measurement is done, which is done via some physical interaction, things are on lock, and the whole superposition disappears. Knowing the basic rules, let's see what happens.
So the superposition is denoted as: |state 1> = sqrt(3/4)*|1> + sqrt(1/4)*|2>.
Why square root? Because once multiplied by the complex conjugate, you want the whole thing to have probability one (similar to saying that the probability of getting either heads or tails in a coin flip is 1):
(sqrt(3/4)*<1| + sqrt(1/4)*<2|)*(sqrt(3/4)*|1> + sqrt(1/4)*|2>) = 3/4*<1|1> + 1/4*<2|2> + sqrt(3/16)*<1|2> + sqrt(3/16)*<2|1> = 3/4 + 1/4 = 1
So what if you want to find the probability of finding energy 1? Look above and see where 3/4*<1|1> is? That number tells you the probability. So, you have got such a particle and measure it. Let's say you measure it and say get energy 1. Then you recreate a particle of the same state and measure it again, obtaining energy 2. Do the same thing 1000 times, and you observe that 756 times you observed energy 1 and 244 times you measured energy 2. Notice how while each individual measurement gives you one or the other, if you do it a lot of times, the total numbers becomes closer and closer to the probability number? That is similar to how a coin flip works, where you might get 5 heads in a row at the beginning, but when you do it a thousand times, the number of heads and tails obtained are almost shared equally.
But see, this is where it is not like a coin flip. A particle might have state: |state 2> = sqrt(3/4)*|1> - sqrt(1/4)*|2>.
But wait a minute, isn't it redundant since it gives you the same probabilities?! It does give you the same probability, but what if you prepare the following?:
|state 3> = A*|state 1> + B*|state 2> = (A + B)*sqrt(3/4)*|1> + (A - B)*sqrt(1/4)*|2>
Now the fact that there is a minus sign matters. Let's say A = B, just to make a really eye popping example: |state 3> = 2A*sqrt(3/4)*|1>, and A = 1/2*sqrt(4/3) to make the probability equal 1. So |state 3> = |1>, and because of this there is zero chance energy 2 can be measured!
So I know this is going to be quiet difficult to interpret, especially those without a math background, but hopefully it informs what they mean by quantum superposition.
- Status
Similar threads
- Poll
- Poll