Programmable Quantum Computer Created 132
An anonymous reader writes "A team at NIST (the National Institute of Standards and Technology) used berylium ions, lasers and electrodes to develop a quantum system that performed 160 randomly chosen routines. Other quantum systems to date have only been able to perform single, prescribed tasks. Other researchers say the system could be scaled up. 'The researchers ran each program 900 times. On average, the quantum computer operated accurately 79 percent of the time, the team reported in their paper.'"
79% accuracy ... (Score:2)
20% of the time it got it wrong, and 1% of the time, someone looked in the box and it wasn't there. 79% accurate. That's pretty useless. I've got a pair of dice that can do just as badly.
Re: (Score:2, Funny)
My dice are 100% accurate. I ask them for a random number, and every time that is what they return.
Re: (Score:2)
int rand() //Chosen with a fair dice roll
{ return 3; }
(Old joke, yeah.)
Re: (Score:1, Funny)
Re: (Score:1, Funny)
Someone will put linux on it, and it will be able to crack RSA, but people won't be able to figure out how to get their printers to work with ti.
...Or their keyboards.
Re: (Score:2)
I can fix that for you with a bit of sandpaper (dice that are slightly sanded on one or more faces are called"flats [metal-express.net]", and come up non-random).
Re:79% accuracy ... (Score:5, Insightful)
79% accurate. That's pretty useless.
Not useless at all, just have it solve the same problem 5 or 15 times and go with the answer that it gives most often. Plus, for some problems it's much easier to verify an answer than to come up with it -- for those problems, just pair it with a normal computer to check the answers, and keep trying until it says the answer is right.
Re: (Score:3, Informative)
79% accurate. That's pretty useless.
Not useless at all, just have it solve the same problem 5 or 15 times and go with the answer that it gives most often. Plus, for some problems it's much easier to verify an answer than to come up with it -- for those problems, just pair it with a normal computer to check the answers, and keep trying until it says the answer is right.
One of the classic examples of that last one is prime factorization. In general it's very hard to come up with the two primes that were multiplied to create a very large number, but if the quantum computer coughs up a candidate it's downright trivial to check whether that's a solution.
Re: (Score:2)
This means that quantum-based processor will either become useful for a certain niche (something that doesn't require precise results) or we'll find a way to make them useful for everyday stuff... like outfitting classical processor technology with quantum capabilities to solve specific types of problems more efficiently.
Re: (Score:1)
There's still room for error there though, and that is simply unacceptable based upon how we use our computers today.
Well, no. Computers can always be wrong. Repeat the calculation often enough, and you'll be better than your average PC. Besides, as RudeIota notes, sometimes the calculation is hard, but verifying the result is trivial (that is often the case in cryptography, for example). In those cases, you can just repeat the calculation until it's correct.
Re: (Score:2)
There's still room for error there though, and that is simply unacceptable based upon how we use our computers today.
This is why everybody uses only ECC memory in their desktop machines and all filesystems in common use support checksumming for data integrity.
Re: (Score:2)
Or uses floating-point hardware. Regardless of how many bits you use, floating-point computations are always approximate (mostly because the range of numbers that can be represented is larger than the number of possible numbers a given bit length can r
Re: (Score:1)
So I guess in the near future we will be seeing things like:
>>> 1 + 1
2 (p < .001)
Re: (Score:2)
Re: (Score:1)
Re: (Score:2)
Plus, for some problems it's much easier to verify an answer than to come up with it
Yes. They're called "NP Complete" problems.
Re: (Score:2)
Plus, for some problems it's much easier to verify an answer than to come up with it
Yes. They're called "NP Complete" problems.
Maybe [wikipedia.org]:
I'm not certain exactly what this means, but part of it seems to be that "one-way" != "NP-complete" (because that would result in a much shorter explanation). But it does at least seem to me
Re: (Score:1)
Re: (Score:2)
Re:79% accuracy ... (Score:5, Funny)
That's TOTALLY moronic. That's like saying "get 5 or 15 people to guess your birthday and go with the answer that it gives most often."
Are people accurate 79% of the time? In the examples you gave? Then no, it's not like that at all.
How stupid can you get?
Many thanks for demonstrating!
Re: (Score:2)
There were examples where people had the same (lousy) accuracy as this. Try reading the links instead of just looking at the portions I quoted for brevity.
Re: (Score:1)
Your links is interesting in one sense but completely irrelevant to the issue at hand. And, frankly, the birthday thing was YOUR example, and it was completely wrong, so maybe you shouldn't have made that your example? You can't possibly castigate somebody for not inventing an argument you didn't make, especially a bad one, based on an irrelevant link you provided.
From looking at the link, it's more like asking 6-17 people what country is the world's largest consumer of oil, and going with the most common
Re: (Score:2)
Note the article - hundreds of different functions were tested, with an overall accuracy of 79%. How are you intending to build more than a trivial piece of code when you have 79% accuracy from each function, on average? The fact is, the more function calls you add, and the more functions you add, the more likely that at least one will be wrong - and then the errors will, of course, cascade, since now you already have bad data for the next fuction, and it's highly unlikely you'll get just the right error
Re: (Score:2)
Re:79% accuracy ... (Score:5, Informative)
No, it's actually a perfectly reasonable idea. Consider running the device (n+m) times. The probability of it being right n times and wrong m times is given by:
P(n,m) = (n+m)!/n!/m! 0.79^n 0.21^m
Now consider the probability of it being right (majority has the right answer) out of 2n+1 trials. This is the given by:
S(n) = sum( P(n+1+i,n-i), i=0..n )
This can be simplied to a closed form using Legendre and gamma functions, but that's kind of messy and it's far easier to just plug in values and do the summation. As it turns out, doing the experiment 15 times and taking the majority (plugging 7 into S(n)) will give you the correct answer 99.4% of the time. Doing things 35 times gets you to five nines of accuracy... completely reasonable in my books.
Re: (Score:1, Troll)
I want 1005 accuracy. I don't want 2+2 = 3.9999999999882 on average after 100 runs, then having it fail every time when I do an if(2+2) == 4).
Sure, I could use a delta, and then make sure it's under that, but then you have the accumulation of errors and tolerances.
Take a barrel of shit, add a cup of wine, you still have a barrel of shit.
Take a barrel of wine, add a cup of shit, you have a barrel of shit.
A failure rate of 21% is bad business. Ask Microsoft wrt the XBox RRofD
Re: (Score:2)
You can't ever have 100% accuracy, ever. Nothing you do can ever change that.
This isn't talking about " 2+2 = 3.9999999999882 on average after 100 runs". The given algorithm is 2+2 = 4, with 99.99999999% confidence, determined after 100 runs. Which is what you ALREADY get with non-quantum computers, because nothing is 100%.
Quantum computing is not like floating-point computing. It gets an exact answer, with any individual run having potentially shitty accuracy, but accuracy can be fixed. The errors and
Re: (Score:2)
The problem is you want to look at the accuracy of the entire system, not just one line. In a 100,000 line program, making multiple recursive calls to many functions, 21% errors in each function will mean you'll pretty much never get the same answer twice between runs, whereas conventional hardware is pretty darned deterministic.
It's not like FP arithmetic, but the errors still do accumulate, because bad output from one step becomes bad input for the next - the errors accumulate aggressively.
Re: (Score:2)
You are acting like they will only make one calculation and go with it. You keep talking about this 21%, when the accuracy measurements, (until they find a way to make it more accurate) would require confidence testing on the numbers before input into as an answer into your program. Not that I think they'll be happy with doing it that way. But for now, it'll give you the accuracy necessary to get the job done right.
Re: (Score:2)
Why do things at a level of granularity of 100,000 lines? Why not get the quantum computer to do the 'repeat x times and pick the most common answer' at each instruction? It'll introduce exactly the same slow down factor, and vastly reduce the chance of error propagation.
Conventional computers already have a certain amount of error that creeps in. Suppose during a single tick of the clock cycle there is some chance P of an error occurring. Find the 'repetition factor' n such that the quantum computer
Re: (Score:2)
Transistor radios were out 5 years (1952) after the transistor was invented (1947). Between 1955 and 1957 several companies were selling fully-transistorized calculators and computing devices.
I doubt we'll be seeing fully-quantized computers in 10 years.
The interesting problems work differently (Score:2)
For most of the problems that you'd want to solve with a quantum computer, the problem is in NP or maybe even P, so if the QC can guess the correct answer, you can verify that it's correct. For instance, for factoring large numbers that are a product of two primes, conventional computers can't guess the factors in usefully short times, but if somebody guesses or steals an answer, you can easily check whether it's correct or not. You don't need to do a best-of-N to figure out what's probably the answer.
On
Re: (Score:2)
That's TOTALLY moronic. That's like saying "get 5 or 15 people to guess your birthday and go with the answer that it gives most often."
That would actually be quite impressive, I don't know of a computer that can guess a random value, such as a birthday, with 79% accuracy. And everything else in your post, though I've heard most of it before, just makes me feel sad for the condition of the USA.
Re: (Score:2)
His entire post was based on a satire article. The fact that so many people took it seriously makes me happy I'm better than those elitists! Wait...
Re: (Score:2)
My point is simple - we laugh at the American citizens who think the United States is a foreign country and can't find it on the map, but we take :quantum computing" seriously when it has equally laughable results. Until it's accurate, the "computing" part should be removed. It might be quantum, but it ain't computing, because it simply doesn't compute (pun intended :-).
Next buzzword bingo article - "Cloud quantum computing at the LHC makes baguette disappear 50% of the time! Latest theory is toast! Ph
Re: (Score:2)
Please, stop embarrassing yourself. There is nothing wrong with being uninformed, but it should not be a point of pride.
Re: (Score:1)
That's TOTALLY moronic.
I've been seeing this a lot lately.
Is responding to objectively correct posts with a combination of insults and an invalid argument the current fashion in troll-land, or has slashdot been invaded by twelve year old non-nerds? In good faith I'll assume the latter, leaving me with a stupid question: how did a twelve year old get a five digit UID?
Re: (Score:2)
Your assumption is TOTALLY moronic. Creating an emotional argument has ALWAYS been the most effective tactic in trolling.
Re: (Score:2)
Think about it - the posters' premise was that take a sample of 5 to 15 times, and go with the majority answer. I pointed out some of the hazards of that. There are many more, all obvious to anyone who wants to think for a few minutes.
Let's say we do a calculation, 2+2. We do it 5 times. If it's only right 79% of the time, every once in a while, we'll get a sample of 5 where the majority is not 4 ... So, if we have a condition like if (2+2 == 4), it will fail once in a while, in a non-predictable way
Re: (Score:2)
And if you have code that is non-trivial, then you're going to have more than one function call. Say you have code that executes a few billion function calls (not too hard to imagine - your computer does it every day just surfing the web and drawing pretty icons on-screen). Your 4/10000 error for each function call after picking the median of 5 iterations means that we can pretty much guarantee that every run of the overall program for the rest of your life will be in error. So, you say, increase the nu
Re: (Score:2)
They never claimed each statement or function call had a 79% chance of success, they said the program had a 79% of success. This is fine for a randomized algorithm. Check out http://en.wikipedia.org/wiki/Monte_Carlo_algorithm [wikipedia.org] for more information.
Re: (Score:2)
Hmmm? The result sets are independent of each other. Just like the probability of throwing a die and saying that "6" is the "wrong" number, you have a 1/6 chance of getting it "wrong" EACH TIME you throw the dice. It doesn't matter how often you've thrown it before, this time, your chance is 1/6. And next time, you chance is 1/6. This is why people lose money in casinos. The dice don't have to listen to you, they don't HAVE to eventually give the number you want, etc.
Re: (Score:2)
how did a twelve year old get a five digit UID?
I've been watching eBay for a low cost 4 digit UID...
Re: (Score:2)
I hope because of that his post is satire.
Re: (Score:1, Flamebait)
Just because it's satire doesn't mean it isn't true - satire is used to expose truth, not hide it.
Re: (Score:2)
I'm not going to stand up for Americans' knowledge of the world beyond their borders (of their country or their county), but remember that the interviews you often see on the net are the result of hours of interviews cooked down to the 4 minutes which are the funniest and most outrageous.
The reality IS bad, don't get me wrong, b
Re: (Score:2)
Problems have a way of falling down a lot quicker than you'd think. When they started sequencing the human genome, they thought it would take 100 years. Gee, how time flies - it sure didn't seem like it took 100 years.
There's almost always a better algorithm if you look at a problem long enough.
Re: (Score:2)
These aren't new problems like the human genome sequencing was, we've been looking at these problems for quite a while now. While we occasionally do get a slightly better algorithm it's still the same order of complexity.
Improving on the 79% accuracy (Score:3, Funny)
79% accurate. That's pretty useless. I've got a pair of dice that can do just as badly.
You may be interested in purchasing this chip I have here. It has a very nice fdiv routine. Since we're so good friends, I'll give you a 100.00001353% discount.
Re: (Score:2)
>>79% accurate. That's pretty useless. I've got a pair of dice that can do just as badly.
79% accurate? That's good enough for government work!
Re: (Score:2)
A factoring algorithm that gives the correct answer with 50% probability (in a short enough time) would be very useful.
Since you can check the answer with a single multiply, you keep trying until you have the correct answer.
This is one of a set of problems labeled "NP" - a characteristic is that you can verify an possible answer in polynomial time.
Any of these problems can be solved with a polynomial time algorithm that gives the correct answer 50% of the time.
Re: (Score:3, Informative)
Re: (Score:2)
Actually quantum computing is, by design, probabilistic. Every specifically quantum algorithm (even Shor's infamous factoring algorithm) gives incorrect results by design for the simple reason that it's really not possible to have quantum algorithms which succeed all the time (unless you forgeo their quantum properties). So long as the probability of a correct answer is strictly greater than 0.5, however, one only has to repeat the computation a constant number of times to get the probability of success arbitrarily close to 1.
Nitpick: though this is true for the most familiar quantum algorithms, i.e. the ones in BQP (Bounded-error Quantum Polynomial-time), there are also some quantum algorithms that give exact answers with 100% certainty, such as EQP [stanford.edu] (Exact Quantum Polynomial-time).
Re: (Score:2)
/me stirs berylium ions and tastes...
/me announces "needs more cats!"
correct and incorrect? (Score:1)
Re:correct and incorrect? (Score:5, Funny)
So is that 21% of the time is was both correct and incorrect ?
That's correct and the other 79% of the time the cat died.
Obligatory Anchorman quote (Score:2)
Loosely paraphrased here
79% of the time it is 100% accurate
Re: (Score:3, Funny)
So in the other universe 21% of the time cats die. I knew this universe was violent!
On the death of cats (Score:2)
I knew this universe was violent!
Blame god. He goes and kills a kitten every night after he visits me to tuck me in.
Err... I mean... Forget I said anything. Yes, it's violent here. Ahem.
Re: (Score:1)
Re: (Score:2)
The actual state of the machine is "all possibilities at once", it is the act of observing the result that actually collapses the waveform and causes the answer to settle into a specific state.
So obviously, in the 21% cases, the operator just looked at the computer "in a funny way".
wait (Score:1)
79% of the time they work every time
Think of the cats! (Score:1)
Re: (Score:2)
$cat BoxInside.sh ...
Oh crap!
Re: (Score:2)
Screw the cats.... I want to know about puppies.
Could be worse... (Score:3, Funny)
On average, the quantum computer operated accurately 79 percent of the time,
Well, its better then anything Microsoft can come up with...I'll take 10!
Re: (Score:1, Flamebait)
And is better than the probability of total Linux hardware support for any random pc/laptop!
Accurate only 79% of the time? (Score:2)
Since it's a Quantum Computer, shouldn't reading the results actually mess up the results? Or at least that's what I understood from that Futurama racing joke.
In Some Alternate Universe (Score:5, Funny)
In some alternate universe, there's a guy who is riding a bus, a thought pops into his head, "Pick a number between 1 and 100. Now, add 3. Now, divide by 13...". 99% of the time, he does the problem in his head, 79% of the time he finishes it. 1% of the time, he says, "Screw it". 100% of the time, he wonders where the hell these things are coming from and decides to check himself into the nearest mental ward.
Quantum computing is screwing up someone's day.
Re: (Score:1)
And that's by having only one quantum computer in a single universe... now imagine what would happen if all the other universes would start inventing such computers!
Re: (Score:1)
By the way, who would be able to claim the invention when the universes merge? Hmm, intriguing concept, prior art in a parallel universe...
Re: (Score:2)
I've already patented "A Quantum Method for Parralel Copyright, Trademark, and Patent Registration".
(Note: I assert that this comment is now prior art in all Universes and proof of my ownership of said patent, including Universes that contain no processes or concepts of copyright, trademark, or patent.)
How do they know? (Score:3, Insightful)
I'm curious how they could possibly know that it operated correctly 79% of the time, since the underlying quantum state isn't observable. You could say it produced the 'correct' results 79% of the time, but that's not the same as saying it operated correctly 79% of the time; it's very possible for a quantum computer to operate incorrectly and still produce the right result, through sheer random chance.
I suppose I could read the paper.
Re:How do they know? (Score:5, Funny)
I suppose I could read the paper.
I think you might be on to something here.
Re: (Score:1)
Re: (Score:2)
The underlying quantum state *is* observable. Why wouldn't it be?
If you RTFA (and not even the paper is necessary for this), you will see that they are limited by the fidelity of their setup, ie. signal to noise. Hence, when they improve their apparatus, they will get more accurate results.
Ha ha (Score:2, Insightful)
Can you imagine the accuracy of a Beowulf cluster of that?
Re: (Score:1)
Why is the result Right or Wrong? (Score:2)
Or, you get more combinations of right, wrong, and other as answers. Now, what happens when one unit in the cluster suddenly starts throw the right answer 100%?
Or, goes 100% wrong?
Or, goes 100% OTHER?
What if it taps something we cannot comprehend?
What if it hits "other" just once. And as a result, somewhere in the timeless Eternity, God freezes, bends over, and monkeys fly out of His ass?
Re: (Score:1, Informative)
not if they are parallel. computations that are done on each node where the majority of the answer is the right answer (as long as it's accurate past 50%). won't increase speed but will increase accuracy.
Re: (Score:3, Funny)
Again 79%. size does not matter
That's what she sighed as she patted him on the knee and then walked out of his life never to be seen again.
Re: (Score:1)
21% of the time Beowulf accidentally shakes Grendel's hand.
Bye-bye encryption? (Score:1)
How long until they get it to factor huge numbers?
Re: (Score:2)
Bring it on, save the nation (Score:2)
Experimental physicist Boris Blinov says that one of the most exciting things about the new study is that the quantum computer may be scaled up. “What’s most impressive and important is that they did it in the way that can be applied to a larger-scale system,” says Blinov, of the University of Washington in Seattle. “The very same techniques they’ve used for two qubits can be applied to much larger systems.”
Pretty soon they will be able to calculate the US budget with accuracy heretofore unmatched by any recent administration.
Re: (Score:2)
You don't need two quantum states for that, one will suffice.
In The Red, until the end of the universe (where I hear there's a very nice restaurant).
Re: (Score:2)
You don't need two quantum states for that, one will suffice.
Funny +1
In The Red, until the end of the universe
Insightful +1
Too bad I cannot moderate in the same thread I post in.
First infinite loop.... (Score:5, Funny)
do {
solveProblem();
} until (getPhotonPosition() && getPhotonVelocity());
Re: (Score:1)
the result of getPhotonPosition() and getPhotonVelocity() just has to be a function, not a single number...
So you could get for example two gaussian functions as result. And then you use some fuzzy logic.
PS: the velocity of a photon is often the speed of light (in vacuüm)
Re: (Score:1)
Re: (Score:2)
In vacuum Photon Velocity will be a rather famous constant.
The photon momentum, on the other hand, depends on the photon's energy ( i.e the colour ). Thus you can't break the uncertainty principle using photons, even though their velocity is always the same.
Now in before some smartass points out it is gluons that have colour.
Ahh... (Score:2)
Well (Score:2)
You know what they say... 79% of the time, it's correct every time.
Who cares? (Score:1)
79% is the best you can expect in the Slow Zone. (Score:2)
Get out into the Beyond [wikipedia.org], and you can reasonably expect 100% efficiency out of your quantum computers. Keep going into the Transcend, and you can reasonably expect better than 100% efficiency -- or at least that's what it looks like to merely-human minds.
Just don't open any unsigned JAR files.
Re: (Score:2)
Hexapodia is the key insight.
Re: (Score:2)
Yeah, but... (Score:2)
Re: (Score:1)
I think a machine like that is better suited for Windows, because it can offer an improvement. Users found that it would only crash reliably about 79% as often as expected.
Re: (Score:1)