I don't think we know actual problems they solve better than classical computers. In many senses, they "feel" like analog computers, which had niche applications for quite some time before good enough digital computers outperformed and replaced them. Where they seem more promising is at simulating quantum chemistry problems (which is what Richard Feynman imagined them doing well before the first one was demonstrated).
I sat down a couple weeks ago with some experts and what I got is that some key parts that make a general purpose computer as we know them today are missing (bear with me - I'm a total newbie on this), one that struck me is quantum memory, because we just can't really store anything for a significant time in a way that we would be able to input the result of a previous computation into a new one. Also, the notion of a stored program (in memory) is very alien to quantum computers just yet.
That said, D-Wave says they are solving real-world optimization problems on their not quite general quantum computers, so I won't dispute that.
Ability to simulate quantum chemistry will change the world.
We know that self-assembling, self-replicating, energy-generating nanomachines are possible because we have them in our own body. Ability to design, simulate, and optimize these with compute rather than physical experiments would be amazing.
And being able to do that at scale, instead of running a single time-consuming experiment at a time, running thousands of simulations much more quickly, and maybe using AI to assess and guide evolution in those designs, will prove quite interesting.
Interestingly that is exactly what I was thinking about when I typed that comment. Lot's of promise 35 years ago when I was in high school. Every so often we hear about some breakthrough, but nothing ever materializes.
What is "reliable" though? All digital systems will undergo some rate of correctable/uncorrectable errors, it just depends on what you consider to be an acceptable rate of failure for your device.
> So until now no quantum processing was reliable. Fool me once...
At a fundamental level, pretty much yes. The trick is to build virtual reliable qubits out of a bunch of fundamentally temperamental ones. At some point we'll be able to trust the virtual qubits keep their state indefinitely, like memory.
