As many of you know, D-Wave has a nice paper out about some experiments on one of their eight qubit systems. In addition they have sold one of their systems to the military industrial complex, a.k.a. Lockheed Martin.
One of the interesting things about the devices they are building is that no one really knows whether it will provide computational speedup over classical computers. In addition to the questions of whether adiabatic quantum algorithms will provide speedups for useful problems, there is also the question of how this speedup will be affected when working at finite temperature. If I were an investor this would worry me, but as a scientist I find the question fascinating and hope they can continue to push their system in interesting directions. Of course if I were an investor I’d probably be some multimillionaire who probably has an odd risk aversion profile
A fun question to ponder, at least for me, is what will eventually happen to D-wave, in, say, ten years. Of course there are the most obvious futures. They could run out of funding and close their doors as a device maker and sell their patent porfolio. They could succeed and build machines that do outperform classical computers on relevant hard combinatorial problems. Those two are obvious. BORING.
But my favorite scenario is as follows. D-wave continues to build larger and larger devices. At the same time they perform even more exhaustive testing of their system. And in the process they discover that there are “noise” sources that they hadn’t really expected. Not noise sources that violate quantum theory or anything, but instead noise sources that end up turning their stoquastic Hamiltonian into a non-stoquastic Hamilotnian. While no one knows how to use the Hamiltonian of D-wave’s machine to build a universal quantum computer, it is entirely possible that such a machine, plus some crazy extra unwanted terms could end up being universal. So while the company is squarely behind the dream of a combinatorial optimizer, it’s not at all impossible that their machine could accidentally be useful for universal adiabatic quantum computation (and of course whether this can be made fault-tolerant is still a major open question, at least for the models with non-degenerate ground states.) Wouldn’t it be hilarious if the noise which most people believe will destroy D-wave’s computational advantage actually turns their machine into a universal quantum computer? Ha!
So which will it be? And what odds will you give me on each of these possible futures?