Confirmed, finally: D-Wave quantum computer is sometimes sluggish
The D-Wave Two quantum computer clocks in no faster than a standard PC, but those already well-known results still leave us scratching our heads over speed testing.
D-Wave Systems, the leading manufacturer of the world's first commercially available quantum computers, is the most well funded and far along player in the quantum chip race, but hasn't yet succeeded in convincing scientists that its machines are successfully achieving quantum speedup. In other words, we're not sure that its product is speedier than traditional, silicon-based machines.
In fact, in certain situations, the $15 million D-Wave Two is still no faster than the computer on your desk right now.
A research team at the Swiss Federal Institute of Technology in Zurich reports that there is still a lack of definitive evidence that the D-Wave Two can perform functions any faster than traditional machines. The results of the test were published in the journal Science Thursday, though the work of head physicist Matthias Troyer has been widely circulated since January because the paper was available in pre-print.
"Using random spin glass instances as a benchmark, we find no evidence of quantum speedup when the entire data set is considered, and obtain inconclusive results when comparing subsets of instances on an instance-by-instance basis," Troyer, a physicist at the Swiss Federal Institute of Technology, wrote.
Quantum speedup is the process by which a computer not based on silicon -- D-Wave uses tiny liquid helium-cooled loops of niobium for its computers -- can bypass traditional computational limits through the quirky, occasionally unexplainable weirdness of quantum mechanics. While binary computers are restricted to math-based computations using bits that flip between 1 and 0, quantum computers hijack properties like entanglement and superposition using quantum bits, or qubits, that exist as 1 and 0 simultaneously, theoretically amping up the calculation speed exponentially.
The Zurich tech is yet another conflicting assessment of D-Wave's hardware -- and runs counter to the results touted by D-Wave itself, as well as Google. The search giant last year established its Quantum Artificial Intelligence Lab in a partnership with the NASA Ames Research Center and Universities Space Research Association. The goal is to use quantum computations to unearth new ways of breaking through the computation walls currently plaguing the progress of algorithmic AI and machine learning. Google is using D-Wave as its primary product.
Google and NASA celebrated the D-Wave Two back in January with test results of its own showcasing the D-Wave's quantum speedup clocking in at 35,500 times faster than off-the-shelf optimization solvers. But Google was quick to point out, as Troyer laid out in the Science article, that there is no clear cut way of testing quantum computation because we don't really know how to calculate, or even detect, quantum speedup, let alone set a standard for benchmarking.
"While this is an interesting baseline, these competitors are general-purpose solvers. You can create much tougher classical competition by writing highly optimized code that accounts for the sparse connectivity structure of the current D-Wave chip," Google admitted in a blog post concerning the results.
Google even cites Troyer and his team's results as reason not to hail the D-Wave as a catch-all quantum computer: "Two world-class teams have done that [wrote highly optimized code]. One is a team at ETH Zurich led by Matthias Troyer, considered to be one of the world's strongest computational physicists. With help from Nvidia, his team managed to write classical simulated annealing code running on GPUs that achieves an incredible 200 spin updates per nanosecond. The other tailor-made classical competitor was written by Alex Selby. You may recall he won £1 million for cracking the Eternity puzzle."
D-Wave was not pleased with the way the Troyer test has been perceived for those very reasons, as co-founder Geordie Rose told Wired's Clive Thompson in May. "[They] had the best algorithm ever developed by a team of the top scientists in the world, finely tuned to compete on what this processor does, running on the fastest processors that humans have ever been able to build," he said. And D-Wave's offering "is now competitive with those things, which is a remarkable step," he added.
Troyer and his team were using the University of Southern California's D-Wave Two and running it against a Cray supercomputer software optimizer tuned specifically to run blazingly fast on an Intel desktop.
"Basically, the paper being released in Science looks at a narrow set of benchmarking problems that aren't expected to show an advantage for quantum annealing," a D-Wave spokesperson told CNET, adding that Troyer's research roused a similar debate back in January.
That brings up a central problem in the nascent quantum computing field right now. A lot of the hype and debunking surrounding D-Wave is about shifting the parameters around which tests are performed and against which kinds of traditional machines optimized for various types of computations. So you can make the D-Wave look 35,000 times faster -- or 100 times slower, as Troyer did -- if you simply have it tackle certain types of problems and race it against something that's really good, or crummy, at solving said problems.
One thing is clear: The D-Wave is no ordinary computer, and even Troyer and the Zurich team will admit it can sometimes achieve speeds up to five times faster than the Intel PC. "Our results do not rule out the possibility of speedup for other classes of problems and illustrate the subtle nature of the quantum speedup question," Troyer concluded.
For now, quantum computing is still a bit of a question mark. Hitting that quantum speedup remains an issue of feeding it the right problems and tuning the D-Wave software to crack them. For now, there's no reason to write the D-Wave off when quantum computing has barely arrived and there's still mounds of work to do when it comes to figuring out how to adequately test it.
Update at 2:15 p.m. PT, Friday, June 20: Clarified that the research published in Science Thursday, June 19 concerned scientific research made first available in pre-print back in January.