Google in superposition of using, not using quantum computer

Way back in 2007, a company called D-Wave systems demonstrated... well, no one is really sure what it demonstrated. D-Wave claimed it was a quantum computer, but if that was so, then this tiny company managed to produce something that was way beyond anything that had been seen in a research lab, despite their large budgets and talented scientists. Since then, D-Wave has occasionally released new claims, but, most importantly, no actual information that shows that whatever it has built is a quantum computer.

As a result, I was quite surprised to find that search behemoth Google had started working with D-Wave on quantum computer-based search. Among the questions that puzzled me were why Google would be interested in quantum computing, and, assuming that they are, why would they work with D-Wave?

In a blog post and New Scientist article, Google researchers claim that quantum computation provides a speed-up in image recognition: e.g., a quantum computer can recognize a car quicker than a classical computer. 

The computer was fed a series of images that contained cars; these were initially located manually. The computer then used this information to recognize cars in a different series of images. If this sounds more like a neural network than a quantum computer to you, you are not alone. In fact, once you consider that quantum computers don't provide a speed boost for every type of computation, you have to wonder if image recognition can, even in principle, be sped up by a quantum computer. The information provided by Google doesn't offer any insight into this question.

But that doesn't really matter, because if Google is really using D-Wave's parts, it probably isn't using a quantum computer anyway. At some point in the past, D-Wave finally admitted that it didn't really know if it had a quantum computer or not. The company's claim is that because it has so many bits—I refuse to call them qubits until they show me evidence of the q-ness of the bits—it doesn't have any way of telling if the machine is operating as desired. 

(Actually, this is a clever bit of obfuscation, because the company is essentially saying that its processors are too big and advanced to characterize. But that leaves me wondering why it doesn't run tests on a smaller processor to show at least a smidgen of quantum-ness. This is something that is very possible, and has been done before with other systems.)

What would D-Wave have to do to show that its processors are on the right track? First, they need to show coherence: can one set two-bit values, allow them evolve in response to the environment, and observe that they evolve in synchrony? How long does that sycnchronization last? This has all been measured in similar systems before, but D-Wave appears to be saying that it has chosen not to do this for a couple of the bits on its boards.

Second, it needs to show that the bits can be entangled. Does measuring the value of one bit automatically let you know the value of another bit? How long does such entanglement last? Again, these measurements have been done for similar systems, but they seem to be beyond D-Wave.