A Canadian start-up says it will demonstrate a working commercial quantum computer in Mountain View next week, years ahead of many experts’ predictions. Venture capital-funded to the tune of USD 20m, Vancouver-based D-Wave says it has built a quantum computer with 16 qubits – the quantum world’s version of a digital bit, but which simultaneously encodes 1 and 0, so can carry more information and solve problems more quickly.
Imagine a Beowu–
oh, damn, wrong site. Sorry.
good one.
No it isn’t, it’s been a reply to every single article on Slashdot for the past decade at least. The only thing funny about that is imagining them scrambling to be the first to post it, so they they’ll get all the “LOL had to be done!!!!!” replies and the +5 internet points next to their name.
Yea, but it’s still funny… I mean, did you even try to imagine it (atleast while browsing SD)?
I wonder whether it will work. OTOH maybe even wondering whether it will work is enough to prevent it from working at all.
You can say one thing for sure, though: 20 million dollars buys a lot of cats.
I’m always weary of this kind of announcement… if you remember the quantum-quack behind the ‘atom chip’ people…
http://www.atomchip.com/
yeah, okay..
Keep in mind, I’m optimistic that we’ll have this technology, but it’s applications will take another decade or two to finally debug (which will probably just tbe the start of another barrier/step). But if you tell me there’s a quantum laptop (the link I put up used to have one listed!) ok, I’ve got some moon real estate to sell you!
I give this another 5-10 years before I realy take any notice of it. But that doesn’t mean they should stop trying – it’s just hard to filter out the BS unless you see it working in front of you.
I’m always a bit skeptical of these claims as well, but there are some differences which give me hope. These people are only claiming their QC can do a certain type of problem, not like a general purpose cpu. Also, this prototype only has 16 qubits which they claim is going to be quite a bit slower than conventional computers even on the np problems it will be good at. However, they claim to have the ability to ramp this up to 1000 in 2008 – this seems pretty quick to me, so I guess I’ll have to wait and see.
How do you program these things?
It’s really no different than a normal cpu – you give it some arguments, tell it what to do, and then wait for it to spit out the results on the other end. What happens in the middle is different, but that has always been a “black box” anyway. Well, I guess it’s actually a transparent box, but you know what I mean.
The real difference will be in the software algorithms. I imagine they probably came up with their own limited language as well to help out the compiler.
So what makes these quantum computers different from normal processors aside from the technology employed? I understand they’re totally different, means they require special compilers and the instructions will be totally different from the industry standards we have today. Judging by that blog it’s designed to solve very niche problems, not for general computing. And is there any merit in using them instead of normal processors?
Basically, each “bit” can store more than just a 1 or a 0 at the same time. The physics involved are way over my head, but basically it is a non-deterministic machine. It can be in multiple states at the same time. So while a normal 16-bit cpu can have one of 64K different possible values in a register, a 16-qubit cpu can have 64K different values in the register at the same time and operate on all of them simultaneously.
Anyway, the practical result is this: they can do quadratic algorithms in linear time. Currently there are a lot of problems that simply take forever to calculate on a normal cpu, and by giving them a quadratic speedup these problems can actually be done. Check out the NP class of problems for some examples (non-deterministic polynomial time). Breaking encryption keys is one example. The traveling salesman problem is one you might be familiar with from school. Search algorithms can be drastically sped up.
And is there any merit in using them instead of normal processors?
Well, they seem to be very low power and they may end up scaling better than traditional cpus, but then again they may not. For the foreseeable future they’re going to be a lot slower at most computations and probably a heck of a lot more expensive. They should be fantastic at certain things like simulations and AI, so it is possible we may eventually see an add on card like the physics one out now, but that seems like it will be quite a ways in the future.
But most calculations right now are already designed to run in linear time on a normal cpu (or they would be too slow) and so they wouldn’t get any speed up from a quantum computer. One way to think about it is to say this cpu is like a 64K core cpu, with each core doing the same set of calculations at the same time – obviously each core is going to be pretty slow compared to current technology and if you only utilize 1 of them there’s not much point.
Edited 2007-02-10 09:43
Perhaps a concrete example would help. Take a chess game. The basic way to make an AI opponent is to have the cpu run every possible move from the current game state to get all the next states, then repeat over and over again until the end of the game. Moves that eventually lead to victory are taken while those leading to defeat are avoided. Because of the sheer number of possible moves (about 10^40), the AI has to pick and choose which paths to explore – an algorithm picks those that look likely (you just took out their queen) and ignores those that look unlikely (you just lost your queen) and it will eventually run out of time (usually around 20 moves in the future) and have to guess which of those endings is the best. With quantum computing, you could take out all the guesswork and brute force your way through – it could calculate the result of each possible move all at once. This would allow the creation of a perfect chess player, much like you can play a perfect game of tic-tac-toe today.
Some NP problems have similar ways of guessing which paths to take (called heuristics), but many others don’t. The reason current encryption methods are so hard to break is that there is no way to guess which paths to take. You have to individually examine each unique key separately with no idea when you will hit the right one.
Edited 2007-02-10 10:16
1. I think you mean that they can do exponential algorithms in linear time. Polynomial algorithms — quadratic included — are already considered tractable; it’s exponential algorithms (the kind that work on power sets, like a brute force chess solver) that currently aren’t.
2. A large number of algorithms — sorting, for instance — can’t run in linear time. But they still run in polynomial time. We currently don’t have any exact algorithms for NP-complete problems, and so quantum computing would result in a drastic speedup for every single one. But you’re right that most people wouldn’t notice a huge difference in the speed of their computers, since the kinds of applications that most people use don’t depend on NP-complete algorithms operating on large data sets.
Edited 2007-02-13 18:55
“Its speed at 16 qubits should be roughly similar to standard computers.” – It could may be because it’s a standard computer.
But the real question is, can it run pong?
To get funding , you need to lie.
Con artists made a bundle.
http://en.wikipedia.org/wiki/Quantum_computing