November 9, 2007 11:39 AM PST
D-Wave's quantum computer ready for latest demo
During a session at the SC07 supercomputing conference in Reno, Nev., Hartmut Neven, a Google specialist in image recognition, will show an image recognition algorithm running on a device, made by start-up D-Wave Systems, which is claimed to be the first practical quantum computer.
The system is said to be the first commercially viable quantum computer, implementing the idea of speeding up computation by carrying out multiple calculations simultaneously using different quantum states of a system. In such a system, all possible quantum states exist, so hypothetically, a quantum computer could find the best answer to a problem by testing all possible answers at the same time, in what are sometimes described as "parallel universes."
Proposed by Nobel Prize winner Richard Feynman and elaborated on by David Deutsch in the 1980s, quantum computing has made slow progress because of the difficulty of building systems with more than a few quantum bits, or "qubits," and then maintaining them in a "coherent" state so that the different quantum states can operate at the same time and for long enough to carry out useful work. To get a long "decoherence time," quantum computers have had to be completely isolated from the outside world.
Funded by venture capitalists, D-Wave, based in British Columbia, claims to have solved this problem using "adiabatic quantum computing," in which a device is designed to solve a particular problem and settles on the answer through a process referred to as "annealing." The decoherence time is not a problem because the system can operate with thermal noise, according to D-Wave's chief technology officer, Geordie Rose.
D-Wave is the only commercial quantum computing company, having raised $44 million from partners, including Draper Fisher Jurvetson, GrowthWorks, BDC Venture Capital, Harris & Harris Group, and British Columbia Investment Management. It demonstrated a 16-qubit computer, called Orion, in February, but scientists have been skeptical that D-Wave demonstrated true quantum computing, as no results have been published in peer-reviewed journals.
"Over the last year, rather than answering scientists' questions about what, if anything, they've actually done that's novel, they seem to have descended ever further into the lowest kind of hucksterism," said Scott Aaronson, assistant professor of electrical engineering and computer science at the Massachusetts Institute of Technology.
Orion is probably a classical computer, according to Aaronson. "They apparently built a device with 16 very noisy superconducting quantum bits," he said in a talk given at Google's offices in the summer. Noisy qubits let information into the system and behave like classical bits, said Aaronson. "To make a long story short, it's consistent with the evidence that what D-Wave actually built would best be described as a 16-bit classical computer. I don't mean 16 bits, in terms of the architecture; I mean 16 actual bits. And there's some prior art for that."
If the qubits are actually behaving like classical bits, the result would still be the same, according to Umesh Vazirani, professor of computer science at Berkeley: "This would still be consistent with the results of the demo, since the decohering qubits would act like classical random bits, and the adiabatic computer would act like a classical computer implementing simulated annealing."
Rose has denied this and has claimed that the system has developed further: "Since the demo, we have been developing the support infrastructure for our projects and have used it to design, build, and test seven generations of processor prototypes. Each of these generations has focused on a specific issue related to performance and/or scalability of commercial processors." He claimed that the demo at SC07 will have 28 qubits and will demonstrate an algorithm co-developed by Neven, who has been at Google since the search giant bought his image-processing company, Neven Vision, in 2006.
D-Wave has not had its system externally validated, said Rose, because "there is only one meaningful measure of validation for a technology like this: does it outperform the systems people are using today in a metric that they care about? We are getting very close to achieving this objective."
D-Wave has previously promised to make a 1,000-qubit computer that integrates with conventional database systems by the end of 2008 and to allow the public to use an Orion system made available on the Web.
Peter Judge of ZDNet UK reported from London.
2 commentsJoin the conversation! Add your comment