Quantum Or Not, New Supercomputer Is Certainly Something Else
npr.org
npr.org
As http://www.scottaaronson.com/blog/?p=1400 makes clear, the current incarnation of D-Wave looks cool and costs a fortune, but is slower than a desktop PC running properly-optimized code. And that optimized code only needed to be written once because it solves the problem that D-Wave asks us to reduce all problems to before we can use D-Wave.
If we lived in a world where nuclear energy was still just a theoretical possibility, and you had a company D-Uranium trying to build the first reactor ever, and someone like Scott Aaronson kept saying for years and years "You guys suck you haven't even demonstraded there are any atoms split inside your device, for all we know it's all heat from chemical reactions", and then, eventually in the face of evidence he was forced to admit that atoms indeed were being split in D-Uranium's pile, so he retreated to saying "But your reactor generates less heat than a coal burning power station so you guys still suck!"--what would that make you think?
The article, in the second paragraph, says ...the computer can solve a certain type of problem much faster than conventional computers... which is a claim that even cursory research into the subject would find false. (Though a PR company could easily deliver a packet of materials and people to talk to which would give a journalists quotes on both sides, and leave them thinking that it was true.) That theme of "this is insanely fast" is repeated throughout the piece in various ways with no challenges. The journalist did not do their homework.
That right there is the evidence upon which I base my assertion that this is a puff piece. This looks and feels in every way like what a journalist might write based on the information given by a PR firm.
Full stop.
As for the rest, you're unfairly misrepresenting what Scott Aaronson has been saying. His statement all along is that D-Wave's marketing exceeds their delivery, and he's afraid that a legitimate field of study - which he's involved with - could suffer backlash as a result. He has repeatedly said, and I believe him on this, that if they can actually deliver then he'll be a huge fan. But as long as they say they have delivered when they haven't, he's going to be a critic.
This seems to me to be a reasonable response to an overhyped implementation of cool technology. There always will be salesmen overselling their products. But for building long-term trust, remember the mantra "under promise and over deliver". D-Wave is doing the opposite.
Now D-Wave has shown 1 but 2 remains a valid criticism. I don't understand why anyone should just assume that the D-Wave architecture can be faster --- for a restricted set of tasks --- than a classical computer.
Second, compare quantum computing with fussion, which is of about the same level of difficulty. Some physicist thought he had cold fussion, and the response from the community was (rightfully!) to be skeptical. But even if it would have worked, no one would whine about how Evil Physicists were Keeping Cold Fussion Down. Instead, it would be understood as the appropiate reaction to very wild claims, and large scale entanglement is pretty out there.
Why, what would it make you think?
What really interests me though as a programmer is whether we’d have to change our tools too when quantum computers become the norm. Would we have to adopt new programming techniques or we’d make it by refining the ones we already use? How would techniques like multithread programming evolve in a machine like this?
I suspect that quantum computers will, at best, be simmilar to how we view GPU's now, good for specialized programming tasks.
Scott Aaronson has agreed that they've reached a new high, which also helps his credibility in my mind. But they have not reached what they claim they've reached. And being able to solve only one highly restricted problem is a long way away from having a quantum computer. Basically, if you look at what D-Wave actually has, it's a great deal less impressive than their marketing would like you to believe.
Does D-Wave have any top secret schematics or diagrams we could look over?
The general approach is to view quantum computers as a series of quantum gates. Specifically, the CC-Not gate takes 3 inputs and, if the first 2 are true, it toggles the third. This is sufficient to build an arbitrary gate, including a binary adder.
I would try explaining how these gates might be constructed, but its been so long since I've studied it that I would probably butcher the explanation.