IBM questions the performance of D-Wave's quantum computer
- 04 February, 2014 22:40
Researchers at IBM and the University of California are questioning whether a closely watched experimental computer used by Google actually relies on quantum mechanics as its manufacturer, D-Wave, claims.
At the heart of the battle is a question about the validity of quantum computing, which some predict may offer a road forward after Intel and other chip manufacturers exhaust "and reach the physical limits" of how powerful they can make their processors.
D-Wave claims it does. D-Wave charges that the IBM and university researchers confined their study to only one aspect of what its computer can do.
"A successful theory needs to explain all the existing experimental results, not just a narrowly selected subset of them," said Colin Williams, the business development director for D-Wave, of the researchers' work. "The right theory has to explain all of the data, not just some of the data."
D-Wave is a closely watched company in that it is perhaps the most advanced in terms of commercializing quantum computing, though even its founders acknowledge that they exploit only a subset of quantum mechanics, called quantum annealing.
Quantum computing is the practice of harnessing the laws of quantum mechanics, or how matter behaves at the subatomic level. Advocates claim quantum computing could be more powerful than standard silicon processing in that its small scale of operations can simulate problems too large to be represented in traditional computing systems.
D-Wave markets its machines, which it started selling in 2011, as very large co-processors, handy for solving complex optimization and machine-learning problems that could overwhelm classically designed computers. Google has invested in one of D-Wave's computers and is evaluating the results.
The researchers, from the IBM T.J. Watson Research Center and from the University of California, developed a model showing that, for a given problem solved by a D-Wave computer, a similar level of computation could be achieved through equipment that relied on classical mechanics. They used actual results from a problem solved on Google's D-Wave computer.
Williams said the charge from IBM and University of California is a pretty common one for D-Wave. Researchers will typically try to match the results of a quantum computer to what could be achieved by using classical physics. However, they usually just confine their study to one aspect of quantum computing.
"Whenever there is an experiment claiming to report quantum mechanical -- or any non-classical -- effects, researchers look for classical models that predict the same results. This is a very common practice in the history of science," Williams said.
The researchers examined a problem solved by Google's D-Wave computer that involved finding the ground state of an erratically operating magnet called a spin glass. The researchers came up with a model of a hypothetical system that could solve the problem equally as well just using components that operated by the laws of classical physics.
"Based on these results, we conclude that classical models for the D-Wave machine are not ruled out," the paper stated. In other words, while D-Wave claims its computer is based on quantum mechanics, it is possible that the results it gets can be achieved using only standard classical physics.
Williams said the paper fell short in that it only looked at one case. "There are many other papers that show excellent agreement between our processes and quantum mechanics," Williams said, referring to numerous published studies that D-Wave and other researchers have published. "All those experimental results cannot possibly be explained by a classical model."
D-Wave founder and Chief Technology Officer Geordie Rose goes into this argument in more detail on the company's blog.
More information about the performance of D-Wave's quantum processor has also recently come to light directly from Google.
The Google benchmarks, posted last month, show mixed results, though Google engineers admitted the performance may improve dramatically with future editions of D-Wave's processors, particularly for severely complex problems.
The D-Wave 2 processor was able to complete some problems much faster than traditional computers. But for other problems, the difference between the two was not as great as expected.
"There are problems for which the classical solver wins or at least achieves similar performance," a Google blog post on the benchmarks stated. However, it added, "but the inverse is also true. For each classical solver, there are problems for which the [quantum] hardware does much better."
Google engineers had suggested that performance will improve as D-Wave continues to double the number of qubits on its processors. A qubit, or quantum bit, is the basic unit of information for quantum computing. Unlike a regular binary bit, a qubit is able to hold two states at a single time, an effect called superposition that could be a key element to powerful quantum computers.
Currently Google uses the 512-qubit D-Wave processor, but D-Wave plans to introduce a 1,024-qubit model this year, and a 1,024-qubit processor in 2015.