Quantum computer works best switched off

Captain tycho · Post by **Captain tycho** » 2006-02-23 04:00pm

Gentlment, that is the sound of my brain screaming RAPE RAPE RAPE RAPE RAPE RAPE.

Captain tycho · Post by **Captain tycho** » 2006-02-23 04:01pm

^^^
Bloody fucking hell. That should be Gentlemen.

Darth Raptor · Post by **Darth Raptor** » 2006-02-23 05:59pm

Large scale physics makes my math-impaired brain hurt. Quantum shit makes it implode.

bilateralrope · Post by **bilateralrope** » 2006-02-23 06:26pm

Faram wrote:
bilateralrope wrote:
Faram wrote:I want one of those.

Bring out your encryption, I will hack you with it
Only if you dont have the decryption program running
Well I cant program, so I can do anything with it!

My next trik after cracking and hacking will be squaring the circle.

So my only hope to beat you is to eaither force you to learn computer programing, or hope that quantum computing makes my programing knowledge useless ?

You win

Admiral Valdemar · Post by **Admiral Valdemar** » 2006-02-23 06:41pm

Anyone else getting Infinite Improbability Drive vibes from this? The way the guy in H2G2 makes the drive reminds me of this bizarreness.

Mr. Coffee · Post by **Mr. Coffee** » 2006-02-23 08:16pm

If a quantum computer fell over and no one was there to hear it, would it make a sound?

Molyneux · Post by **Molyneux** » 2006-02-23 09:12pm

Admiral Valdemar wrote:Anyone else getting Infinite Improbability Drive vibes from this? The way the guy in H2G2 makes the drive reminds me of this bizarreness.

There's at least two other H2G2 jokes in that vein on this thread already ^_^

Winston Blake · Post by **Winston Blake** » 2006-02-23 10:13pm

More info. Or is it less info? Negative information? Hell, i don't know any more.

A striking new way to compute the answer of a mathematical problem sounds like a slacker's dream: you turn on your computer, program it to solve the problem, but then don't have to run the program. Provided you are using a quantum computer, you'll have a fair chance of getting the answer anyway.

It sounds absurd. But Onur Hosten of the University of Illinois at Urbana-Champaign and colleagues have shown that it works. The have created a kind of quantum computer using light beams, and find that it can find a particular item in a database without actually looking for it.

This is bizarre even by the weird standards of quantum mechanics, which is notorious for counterintuitive effects. Hosten and colleagues call it 'counterfactual' quantum computation: a way of probing the outcome of an event by looking at situations in which it didn't actually happen.

All together now

A quantum computer is very different from a traditional desktop computer. It uses the laws of quantum mechanics to perform many calculations at once where a conventional computer could do them only one at a time. This drastically cuts the time a quantum computer takes to find the answer.

This is made possible by the fact that quantum objects, such as individual atoms or photons of light, can be placed in 'superposition' states, mixtures of states that are mutually exclusive in everyday objects. A quantum switch, for example, could be simultaneously on and off.

That's the key to quantum computation, because it means that a quantum computer can be placed in a superposition of states where it is running and not running. This leaves an imprint of the 'running' state on the history of the 'not running' state, such that one can look at the latter and determine something about the former.

"Some people like to think of this as two different universes", explains computer scientist Richard Josza of Bristol University in England. In one universe the computer runs, while in a parallel universe it doesn't.

One might say then that the computer does actually run, but in a 'parallel universe'. "So you wouldn't be charged for the cost of running it," says Josza.

Not so simple

But there's a catch. The outcomes of quantum processes can never be predicted precisely, but only in terms of probabilities. So quantum computation doesn't invariably give the right answer.

When the idea was first proposed by Josza eight years ago, some researchers thought that the success rate would actually be so low that one could do just as well by guessing.

Hosten and colleagues have found a way to avoid this limitation, they report in Nature1. It uses another strange quantum phenomenon, known as the quantum Zeno effect. This amounts to changing the probability of a particular outcome simply by looking for it. It is a realization of the adage that a watched kettle never boils: for a 'quantum kettle', you really can prevent 'boiling' forever just by watching it.

The researchers created a simple 'quantum computer' from laser beams, mirrors and light detectors, which encodes information in the quantum states of photons. In effect, they put a single photon in a superposition of states in which it both is and isn't fed into an optical 'black box' that processes its quantum state according to an algorithm. They then look at the photon to see whether the answer to the calculation is encoded in it.

The team took advantage of the Zeno effect in probing the 'non-running' states of their computer, in which the photon didn't pass through the black box, therebye increasing the probability of finding the answer.

Strangely beautiful

"It's a very beautiful experiment that probes the strangeness of quantum theory," says Josza.

Hosten and colleagues say that this is more than just an exercise in quantum weirdness; they think it might help in making useful quantum computers. So far, such devices constructed from light beams or trapped atoms are mere toys, which have far too little computing power to solve really hard problems.

One of the obstacles to scaling up such model computers is that it is very hard to prevent quantum information from leaking away into the surroundings. This leakage is largely due to the way that quantum bits interact with their environment as the computation proceeds. But if the computation isn't actually run at all, this leakage might be much smaller, making quantum computation less error-prone.

Now if a horde of white robots in cricket gear suddenly appears, tears this machine apart and steals a small golden component, THEN i'll start freaking out.

The abstract of the paper, with links to figures and tables and supplementary info.
Details of the algorithm, read at risk to own sanity.

aerius · Post by **aerius** » 2006-02-23 10:48pm

And I thought watching Surlethe & Kuroneko discuss math theory was confusing...

Darth Servo · Post by **Darth Servo** » 2006-02-24 12:16am

When I first saw the thread title, I suddenly remembered Bill Gates first demoing Windows 98 and it crashing on him.

Post by **fgalkin** » 2006-02-24 03:06am

Quantum Windows: Works best with the BSOD.

Have a very nice day.
-fgalkin

His Divine Shadow · Post by **His Divine Shadow** » 2006-02-24 04:25am

wolveraptor wrote:Sonuvabitch, I'm not even going to try to understand this shit any more. Quantum natures flirting can produce an effect? What the fuck?

Quantum is such a slut.

His Divine Shadow · Post by **His Divine Shadow** » 2006-02-24 04:34am

Captain tycho wrote:Gentlment, that is the sound of my brain screaming RAPE RAPE RAPE RAPE RAPE RAPE.

Captain tycho wrote:^^^
Bloody fucking hell. That should be Gentlemen.

AHahahaha! And with your avatar!

FUQ!

His Divine Shadow · Post by **His Divine Shadow** » 2006-02-24 04:36am

No wait, screw that, no FUQ, I read wrong. I thought you said you meant to scream "GENTLEMEN" and not "RAPE" not nearly as funny when it's intentional.

Old Peculier · Post by **Old Peculier** » 2006-02-24 08:14am

While I don't really understand much of the physics behind any of this, I guess my first year QM module was enough to protect me from total brain rape when reading this.

Still, I wonder how much teaching/experience is required to legitimately think, "yeah, I have a pretty good idea what is going on here."

Ariphaos · Post by **Ariphaos** » 2006-02-24 08:53am

...reminds me of the "A watched particle never decays" trick.

Post by **Surlethe** » 2006-02-24 02:43pm

I wonder what this would do with algorithms to factor into primes.

Code: Select all

6 = 2*3, with a 2.3% error

drachefly · Post by **drachefly** » 2006-02-24 04:06pm

The crazy part is, I get this and now that it's been pointed out it makes perfect sense.

So, there was an experiment a few years back which was featured in SciAm; they called it 'looking at medusa'.

So, you have this light path which goes in a big horizontal rectangle; each time it ends up a little higher, though, so it isn't just going in a circuit.
Each time it goes around, it has a vanishingly small chance of being deflected onto a different path, which would cause it to pass through a box.

If the box contains nothing, then it keeps going and meets up on the other side and is recombined into the photon stream. If the box contains something, the photon is blocked.

After going around the circuit about a hundred times, they sent the photon against a screen in some way that provided an interference pattern.

So, by looking at the interference pattern, they could tell whether the side-trips were included in the interference pattern. If they were, there was nothing in the box. If they weren't, there was something in the box.

By tuning the probability of going on the side-trips, they were able to make it so that they could get 99% probability of determining whether something was in the box while maintaining a 1% probability of the object in the box (if there was one) would be hit by a photon.

That's because each time the photon went around the loop, it was affected by the side-loop. If the box contained an obstruction, then that cut off a tiny tiny portion of the wavefunction. If the box did not, then that part would come around and create a small deviation in the wavefunction.
This small deviation adds up to create a noticeable phase shift long before it adds up to create a photon on the detector.

As far as I know, that was their setup. It gets better if the splitter acts differently for the main beam and the deviated beam.
Let the deviation be more likely to be sent down the side-path next time around. This amplifies the signal sent along the side-path IF there is a negative result from the first time, but if there is a positive result, it isn't sent.

And there you go. The probability of being intercepted grows roughly linearly with the number of times around the circuit; the certainty of the measurement grows exponentially.

I figure something similar is going on here. They're keeping almost all of the photon outside the computer, but letting the computer act as a nonlinear 'echo chamber' that would affect the interference pattern at the end.

KrauserKrauser · Post by **KrauserKrauser** » 2006-02-24 04:31pm

drachefly wrote:snip

Yup, that's the sound of my mind melting. I would have thought it would smell more like asparagus, but nope, burned tires. Who knew.

Bounty · Post by **Bounty** » 2006-02-24 04:34pm

KrauserKrauser wrote:
drachefly wrote:snip
Yup, that's the sound of my mind melting. I would have thought it would smell more like asparagus, but nope, burned tires. Who knew.

So the point of the operation is not to determine the actual event, but the relative amount of times the alternate path is not taken by the photon, based on the principle that it has an equal chance of taking either path unless it is blocked ?

Am I getting close...er ?

Pcm979 · Post by **Pcm979** » 2006-02-24 05:29pm

Every time I think I'm starting to get the concept, my fragile understanding flips me the bird and runs off, cackling madly. At this point, I think I'm going to have to settle for "Quantum computers get better results when not looking for the answer" lest I am reduced to a blithering jelly. More of one than I already am, that is.

Molyneux · Post by **Molyneux** » 2006-02-24 06:47pm

I...*almost* got it that time.

I've got to study up on some QM....

Post by **Surlethe** » 2006-02-24 08:38pm

Drachefly, please correct me if I'm wrong, but here's what I get in my head from your post:

So we have several mirrors positioned around a box; we don't know what's inside. We bounce a light beam around the box in a manner which spirals upward; however, we have a clear mirror which will bounce light through the box a certain small percentage of the time. There are two possibilities: something in the box, and something not in the box. IF there is nothing in the box, on the other side, the reflected will reintegrate into the light wave -- but in doing so, it will slightly affect the wavelength of the light. IF there is something in the box, then the main light stream's wavelength won't be affected at all.

However, if the percentage of deflection is small enough, on average only a percent of an actual photon will be deflected, and thus, after enough turns around, no photons will actually have gone through the box, but the wavelength will have been affected by the probability of something going through the box. Then, we bounce the light over the mirror, separate it into a spectrum, and compare it to what we expect for no deflection at all, to find out if there's a probability of anything in the box. Did I get it?

Post by **Surlethe** » 2006-02-24 08:55pm

Now that I think about this, this probabilistic approach to determining whether or not something is in the box reminds me of the Miller-Rabin primality test for whether a large odd integer n is prime or not. I suppose it's the fact that you can check whether or not it's prime to an arbitrary precision (you can be 1 - (1/4)^k certain n is prime for k iterations of the primality test) which scales exponentially to the number of iterations, but you can never be absolutely sure.

Mad · Post by **Mad** » 2006-02-24 09:44pm

The team took advantage of the Zeno effect in probing the 'non-running' states of their computer, in which the photon didn't pass through the black box, therebye increasing the probability of finding the answer.

The scary implication here is that you'll never have an exact answer, only a probability of having the correct answer. Even at .999999999 probability, running at 1 GHz, you're going to have one error per second.

Or I may be looking at this a bit too linearly as to when the probability factors in. (Each cycle? Each output?)

Still, if this implication is true, then for pure number crunching, running a quantum computer with a binary computer to check the answers may be a good idea, at least for applications where verifying an answer is easier than actually coming up with the answer. (Like getting large prime numbers: verifying the list is easier than getting the list.)

On the other hand, simulation work shouldn't need those exact numbers, since there is inherent error in the measurements anyway. The same is true for games, since the physics are often approximated anyway.

I do have a question, though: if you know the answer, then looking at the wrong answers to verify they weren't chosen, then looking at the right answer is one thing. But what happens with this computer when you don't know the answer? (And, thus, don't know where to look to play with probability?)