D-Wave and Quantum Computing
Posted on February 18, 2007
Filed Under /dev/null/ | 58 views |
Lately a Burnaby company by the name of D-Wave Systems has been receiving a lot of press and publicity over their announcement that they’ve created the world’s first commercially-viable quantum computer.
I’m not even remotely qualified to evaluate that statement but I was on-hand at their Vancouver press conference and if even half of what they’re projecting to be able to accomplish in the next year is viable, we’re in for some interesting times.
I’ll leave describing it all to the likes of Ars Technica, Technology Review, Scientific American and the International Business Times to regale you with the details, definitions, and doubts about D-Wave’s claims and computer. If you’re really interested I recommend subscribing to D-Wave CTO Geordie Rose’s blog rose.blog.
And I’ll just tell you what I heard on Thursday.
First, D-Wave went to great pains to stress a number of limitations to its achievement: when they say “first”, the words “commercially viable” that follow are critically important. Their computer is accessible via the web to those who might want to use it, meaning D-Wave can charge for it, meaning it is commercially viable. They did not claim to have built the first quantum computer.
Second, they were more than willing to laugh at the limitations of their current incarnation. It was stressed that it is 100x slower than the average Future Shop PC and has so far cost them about $22 million to create. No time soon will a D-Wave machine replace your Mac or help speed up your graphics card. For the kinds of computations that standard computers do so well, their machine seems to do remarkably poorly.
Third, it is suited to solving a certain class of computing problem known as “hard” problems, problems in which basically (and forgive me this explanation, mathematicians) the more you solve it, the harder it is to solve. They’re particularly interested in applying their machine to NP-Complete problems, problems for which finding an answer is incredibly hard for traditional computers but for which once an answer is found, verifying the correctness of the answer is “relatively” easy. This is an important point because….
Fourth, the computer doesn’t actually “solve” anything in the finite, absolute answer sort of way. Rather, it computes a range of probable answers, weighted towards the one that is most probably correct. In effect the right answer is the one derived when the machine’s processors settle into their lowest stable quantum state. This is because….
Fifth, when they compute the answer to a problem they do so multiple times, then present the answers as a list of probabilities with the answer that came up most often being most likely the correct one. Remember that bit about NP-complete problems being hard to solve but easy to verify? This is where that comes in handy. The reason the answer needs to be computed numerous times is because the computer doesn’t do any real-time error correction and so instabilities and errors can creep in, the most probable cause being temperature (which greatly affects quantum state), which is why….
Sixth, they cool that puppy to 4 micro-kelvins using a Leiden Cryogenics Dilution Refrigerator. By comparison, the average background temperature of the universe is about 2.7 Kelvin. That’s pretty damned frosty.
Seventh, the machine is programmed using declarative languages, specifically SQL. That’s right, turns out SQL is pretty well suited for telling D-Wave’s machine what sort of answer it expects back. (A declarative program is one in which the limitations and desired result are specified, the “what”, as opposed to an interpretive program which tells the computer “how” to do something, ala Java or C#).
Eighth, yes it could theoretically be used to break to break public-key crypto. No, they haven’t tried to do that. No they don’t have any intention of doing so. Yes, that could all be a public line they feed and they could already be secretly funded by the NSA (who undoubtedly have a whole farm of these things cranking away already). Their first target seems to be the drug companies, specifically the problem of matching up viable drugs with viable drug receptors, a complicated problem indeed. In other words, publicly stated, they plan to follow the cash.
Ninth, it was simply a real pleasure to see a bunch of people obviously completely thrilled to see the fruits of their labours make it to this point. The presentation wasn’t a polished sales job, they made no bones about the doubts that still linger, the question of whether or not the whole thing really is commercially viable, and if the computer can scale the way they need it to. They made no statements of the absolute (save for the ones involving Kelvin) but rather presented it as “our prototype works… now that the hard part’s done, it’s time to get on with the hard part.” There was no mention of an IPO.
Nagging in the back of my mind is an unfortunate, and hopefully unfounded, comparison to Transmeta Corp., the company that exploded onto the microprocessor scene a number of years ago, and was perhaps most notable for hiring Linus Torvalds for a brief stint. I remember the hyper, the “Intel-killer”, the expectation of low-powered processors becoming so ubiquitous as to end up governing power consumption in every electronic device. And then a whole lot of nothing more. I hope that doesn’t happen to D-Wave.
Comments
Leave a Reply