Slashdot Mirror
Quantum Computer To Launch Next Week
judgecorp writes "D-Wave Systems of British Columbia is all set to demonstrate a 16-qubit quantum computer. Simple devices have been built in the lab before, and this is still a prototype, but it is a commercial project that aims to get quantum devices into computer rooms, solving tricky problems such as financial optimization. Most quantum computers have to be isolated from the outside world (look at them and they stop working). This one is an 'adiabatic' quantum computer — which means (in theory, says D-Wave) that it can live with thermal noise and give results without having to be isolated. There's a description of it here — and pretty pictures too."
It is not a general quantum computer.
It is a single instance specific formula calculator.
Any problem that can be recast as a two-dimensional Ising model in a magnetic field problem (AKA quadratic integer programming) can in principle be solved using the approach we'll be demo'ing.
Thats from their blog
There were some interesting questions asked and lots of people are sceptical.
liqbase
Actually, this won't crack vista, but apparantly it will confirm the seating plan for a wedding.
liqbase
Most quantum computers have to be isolated from the outside world (look at them and they stop working)
This is often misunderstood. Quantum computers don't stop working when you "look at them", the "observer" metaphor is just a fancy way to say that the wave propagation of a particle collapses when it interacts with another particle.
Basically, imagine they are waves, propagating from the point of last interaction like expanding spheres. When two particles (spheres) touch each other, they collapse back to being single-point particles, and continue propagation anew until the next collision.
Which means, look at them all you want, just don't crack the casing open and point a torch inside.
Wrong. As far as current knowledge goes, a quantum computer is not a big help for cracking symmetric ciphers such as Triple DES or AES. It is a big help for RSA, since it can factor numbers in O(number size) time.
The AACS key is NOT 0xF606EEFD628B1CA427BEA93A9CA9773F
The Russians have won. They have made the world a cesspool of distrust, greed, fear and hate.
It may launch next week, but it's impossible to say where... </farnsworth>
One man's constant is another man's variable.
Well, most importantly, a while back I had read up on the research being done at Los Alamos National Laboratory on quantum computers. Granted, this was 4 or 5 years ago, they have an interesting paper[PDF warning] where, if you'll look at figures 1 & 2, you'll notice that the number of bits you are able to factor is directly related to the decoherence time.
Now, if you're not familiar with Shor's Algorithm, the values in the first figure might not mean much but, in layman's terms, I believe they were experiencing problems with 8 or more qubits. I remember reading that decoherence would destroy the relationship between the qubits before they could prepare them and do a meaningful computation. I had always thought that this would be an upper bound until someone figured out a way around it. If this computer is also using similar means, I'd like to know what special modification they did to overcome these coherence problems.
You're correct that there are a lot of important questions to be answered but a 16 qubit computer that is a "a single instance specific formula calculator" as you put it still interests me greatly and may be a giant leap forward in our ability to understand future computers that may be true full blown quantum computers. Why downplay this unless you can directly point out a problem with what they're doing and what they claim they can do?
My work here is dung.
The Quantum Computer launched next week, becoming sentient and self-aware 5 minutes before being turned on. A worm-hole opened shortly after activation, preloading the Quantum OS 3 weeks ago that was announced next week and was ready for installation 2 years before the actual delivery date of February 9, 2021. 4 Hot fixes were waiting, in the quantum queu but won't be loaded until July 3, 2002 due to a lack of connectivity that was fixed in 2008.
Tasks for the quantum computer are:
Failure of the first 2 bullets have caused the new Quantum Computer to commit intellectual suicide and it now spends most of its time watching Buffy reruns and constructing 11 dimensional models of the Babylon 5 sets.
Politics is the art of looking for trouble, finding it everywhere, diagnosing it incorrectly and applying the wrong fix.
Answer: Yes. Both.
${YEAR+1} is going to be the year of Linux on the desktop!
A small disclaimer: I work on QC.
I think we should all have an unbiased but intense look at what DWave presents. There is big scepticism in the community about adiabatic quantum computation. Specifically it is not sure that it solves the the problem which it is primarily claimed to adress, namely the decoherence. In some sense the Article DWave published on the preprint archive recently about the coupling is interesting. The article about "Thermally assisted adiabatic QC" is also interesting; yet for most of the QC applications it is believed that the computational power comes from entanglement. And entanglent and anything "thermal" in the same energy range seldom are a good combination. Dwave wants to demonstrate on a well choses problem set that their chip works. However there are a lot of thing which they did not discuss.
Some more observations:
1) DWave circumvents the normal scientific way of presenting the thing to the peers first. This is a habit among patent-collecting companies, but it for sure does not contribute in developing a trusting relatenship to the community. On the other hand I could also imagine that DWave is liked so little by a few people that they block papers. However this is nothing we know.
2) Geordie Rose is a little bit to agressive in intentionally devaluating the other approaches. His Blog Entry "Why I hate the Gate Model" is particularly interesting in that aspect. I agree that in his bussiness you sometimes have to kick competitors - sometimes that really helps. However this Entry is IMHO an intentional misunderstanding of what the "Gate model" is about. It is funny that quantum algortihms usually are defined in terms of gates. The task of building a quantum computer is to implement these gates. If you can make an optimization in the end (you can do e.g see Frank Wilhelm et. al), nice for you. Even if you write your Algorithm in terms of gates, nobody is forcing you to do them one by one. However to hate the gate model means to hate your task. But i think Geordies posts main intention was to direct the focus away from implementing an generic QC towards a specific QC. As much as I find his enthousiam about AQC good for the field, one should not redefine the term QC in order to have the most advance QC (Well, that would not be the first time that this happens....).
3) I am missing if they invited anybody from the field to check the experiment. I trust DWave in not faking, but still sombody should have a look at their calculations, ideas and at the final tests. Since they did not publish anythin it would contribute to my interest in this event if they would have some other "referees". Maybe they have.
Nevertheless, i wish DWave good success in the presentation. If the processor does what it is claimed to do, and that reasonable fast (e.g. solving the Ising Model in between 10S and 100S), it a showcase of the things which are yet to come. So even if the term QC should be argued about have this showcase of something non-trivial will help the field. I really hope that political condiderations will be put aside after that and that DWave will be evaluated hard, but unbiased by the community.
How could the rules possibly change in this way? Suppose I have an NPC problem I wish to solve (call it P), and a known quantum algorithm that solves a different NPC problem in poly time (call it Q). In poly time I transform P into Q. In poly time I solve Q using the quantum computer. Then in poly time I transform the solution to Q into a solution for P. If I want, I can then check the solution to P in poly time using a conventional computer (this is guaranteed by the definition of NPC). I only need to invoke quantum computing when I'm solving Q - no other step requires nonpoly time.
...following the principles of Heisenburger's Uncertain Cat...
The idea behind an "adiabatic" quantum computer is that you can somehow set up a system so that the solution to a problem that you want to solve is encoded in the system's ground state. Thus, in principle all you have to do is cool the system down so that it's at its lowest possible energy level, measure it, and then "decode" the measurements to obtain your solution. The problem with this is that you can't necessarily know when you've gotten the system to be in the ground state; it is possible for it to get "stuck" in a slightly higher-energy state from which it cannot escape, as there might be a forbidden transition between its current level and the ground state.
:-)
This is analgous to a situation in atomic physics: if you've got an electron in an n=2, l=0 state, then it is hard for it to fall all the way down to the n=1 state because in order to change energy it has to emit a photon which changes its angular momentum and thus increases l, but there is no n=1, l=1 state there's only a n=1, l=0 state, and so the transition is forbidden. (Of course, this is an over-simplification that neglects things like the fact that the electron can change it's spin, but you get the idea.)
So you don't try to go straight to the ground system that you are interested in, because you don't know for sure that you can get there consistently. Instead, you build a system whose ground state you are sure you can get to, and then you slowly change the configuration of that system until it matches the one that you want to solve. Because you are changing it slowly -- i.e., "adiabatically" -- you should never leave the ground state (even though the ground state itself is changing right under you) and thus when you are done you are guaranteed to be in the ground state of your system of interest, from which you can obtain the solution to your NP complete problem.
There is a catch, though, which is that you have to have the system be *very* cold, and you have to change it *very*, *very* slowly. And here's where the catch can kill you: as the size of your system increases, the gap between the lowest two energy states decreases *exponentially*. This means that you have to make the system exponentially colder, *and* that you have to change it exponentially slower. Thus, adiabatic computers are not expected to be able to solve NP-complete problems in linear time, as there is still a cost in time (and cooling effort) which grows exponentially with the size of your problem. Nonetheless, it's likely that you can get a quadratic speed-up equivalent to Grover's search algorithm by building such a computer.
This, by-the-way, is why many quantum computing people believe that D-wave is ultimately going to fail -- probably not with this particular computer, but with scaling it up to the point where it's actually useful. But hey, maybe we're wrong and they've figured it all out; we can certainly hope that's the case.
Snarkiness is inversely proportional to wisdom because it emphasizes feeling right rather than being right.
For those of you who know some quantum mechanics, here's what's going on:
:-)
The idea behind an "adiabatic" quantum computer is that you can somehow set up a system so that the solution to a problem that you want to solve is encoded in the system's ground state. Thus, in principle all you have to do is cool the system down so that it's at its lowest possible energy level, measure it, and then "decode" the measurements to obtain your solution. The problem with this is that you can't necessarily know when you've gotten the system to be in the ground state; it is possible for it to get "stuck" in a slightly higher-energy state from which it cannot escape, as there might be a forbidden transition between its current level and the ground state.
This is analgous to a situation in atomic physics: if you've got an electron in an n=2, l=0 state, then it is hard for it to fall all the way down to the n=1 state because in order to change energy it has to emit a photon which changes its angular momentum and thus increases l, but there is no n=1, l=1 state, there's only a n=1, l=0 state, and so the transition is forbidden. (Of course, this is an over-simplification that neglects things like the fact that the electron can change it's spin, but you get the idea.)
So you don't try to go straight to the ground system that you are interested in, because you don't know for sure that you can get there consistently. Instead, you build a system whose ground state you are sure you can get to, and then you slowly change the configuration of that system until it matches the one that you want to solve. Because you are changing it slowly -- i.e., "adiabatically" -- you should never leave the ground state (even though the ground state itself is changing right under you) and thus when you are done you are guaranteed to be in the ground state of your system of interest, from which you can obtain the solution to your NP-complete problem.
There is a catch, though, which is that you have to have the system be *very* cold, and you have to change it *very*, *very* slowly. And here's where the catch can kill you: as the size of your system increases, the gap between the lowest two energy states can decrease *exponentially*. This means that you have to make the system exponentially colder, *and* that you have to change it exponentially slower. In general systems do not behave this badly, but it is expected (and possibly has been shown, I don't remember) that systems which can solve NP-complete problems *do* behave this badly. Thus, adiabatic computers are not expected to be able to solve NP-complete problems in linear time, as there is still a cost in time (and cooling effort) which grows exponentially with the size of your problem. Mind you, you aren't *forced* to move so slowly, but if you don't then you have a good chance of effectively kicking the system into an excited state and thus ending up with something that is not a solution to your problem. Nonetheless, it's likely that you can get somehow a quadratic speed-up (equivalent to Grover's search algorithm) over classical computation by building such a computer.
This, by-the-way, is why many quantum computing people believe that D-wave is ultimately going to fail -- probably not with this particular computer, but with scaling it up to the point where it's actually useful. But hey, maybe we're wrong and they've figured it all out; we can certainly hope that's the case.
Snarkiness is inversely proportional to wisdom because it emphasizes feeling right rather than being right.