D-Wave 2000Q hands-on: Steep learning curve for quantum computing | Hacker News Reader

D-Wave 2000Q hands-on: Steep learning curve for quantum computing

arstechnica.com

75 pointsfeross27 comments

I've met the founder or D-Wave and have known them casually over the last few years. Few people epitomize scam artist and scientific buffoonery quite as much as this person.

They are now on to their 2nd or 3rd post-D-Wave company and each time they have taken on massive funding rounds at large valuations and then have moved on to form a new company. Call me old fashioned, but something doesn't smell right.

I hope D-Wave, Rigetti and others can move our field forward, but this company more than any other feels like a big shell game.

If that person specializes in very early company stages, getting some knowledgeable team and funding, then moving on when all pieces are in place, it might be a valid strategy. How many investors have any clue about quantum computing or are even capable of understanding anything beyond simple examples/hype in order to finance it? The business world's level of technical expertise is really low even with informed investors and buffoonery about anything is rampant.

Didn't realize Geordie Rose had been hopping around so much.

I love that we have what is essentially part of an intro-level physics course as part of a software review. This sounds like a totally different paradigm of programming, which I guess it would need to be in order to take advantage of quantum entanglement. Neat, though I guess what it does say is that quantum computing is going to basically be its "own separate thing" for quite a while (if not forever).

I don't get what the author is trying to say. It should be obvious to any physicist that his "classical" algorithm for the Bragg grating is complete nonsense (as he now admits in the editor's note). Furthermore, he does not show that the problem the annealer is tasked to solve has anything to do with the physics of a Bragg grating. Essentially he is comparing a wrong physics simulation with the result of "quantum annealing" some random Hamiltonian that looks vaguely similar. I don't see how anything can be learned from that.

I think the author deserves credit for an honest attempt to roll up their sleeves and using this opportunity to encode their own attempt from scratch and reporting the results...

Requiring an actually meaningful scientific finding from a journalist who is merely reporting on a new programming API seems a bit picky.

In this case the author should make it very clear that what he is doing with the API doesn't make much sense. Judging from the comments, most readers seem to believe that he is describing a legitimate use-case for the machine. In fact, the question whether quantum annealers can be at all useful at solving any real problem is still very much open.

The Dunning-Kruger effect is rampant in the comments of "tech news" websites, particularly when it comes to complex subjects that people want to pretend to understand.

Objectively false information is often upvoted to the top, while the people trying to point out why that information is verifiably incorrect are buried in downvotes.

The fact that quantum computing is something that self-proclaimed "futurists" have latched themselves to guarantees that it will be even worse when that particular topic is being discussed.

Q for quantum experts: is quantum computing the Manhattan project of computing, where the effort to harness new physics simply requires unfettered capital or tolerance for few incremental results with real applications? Or is it actually even more theoretical and complex, so that we don't have a good historical example for the effort involved?

The standardized way of defining what a universal quantum computer is through the DiVicenzo criteria [1].

Quantum annealing is a bridge until we have universal QC, it’s a fairly decent one for optimization problems where the solution set is discrete. I personally think D-wave is a good place to start to wrap ones’ mind around QC and superposition states. IBM Q is great, too.

In many ways, I prefer IBM Q if you are getting into QC for the long haul because many QC concepts such as CNOT, Z, X, Hadamard gates, interaction between quantum registers and classic registers are explicit. D-wave’s API hides quite a bit of what QC is all about, and you have to fit your problem (it has to be adiabatic in nature) into their system. I find this to be rather confusing in the long run.

[1] https://en.m.wikipedia.org/wiki/DiVincenzo%27s_criteria

Towards the end, the article suddenly states:

> That is where the real power of annealing lies

From context, it seems he means 'quantum annealing'. Nowhere in the article is it explained what this is. I vaguely recall hearing something about D-wave not building the kind of quantum computer we've heard about (i.e. the kind that can factor numbers). Is this 'quantum annealing', and what makes it different?

Is it better, worse, or plain different from the usual quantum computer?

From what I understood, yeah, it's a different kind of quantum computer, and it's a cause of lots of confusion. I.e. when d-wave says "we already have a xxx bit quantum computer", it sounds impressive, but it really is meaningless compared to qbits in a "real" (and until now hypothetical) quantum computer.

Also until now D-Wave has been unable to show that their QC can do anything at all faster than a classical computer. Which really puts this into perspective: This is all early research. Something interesting may come out of it some day. Or not.

> I.e. when d-wave says "we already have a xxx bit quantum computer", it sounds impressive, but it really is meaningless compared to qbits in a "real" (and until now hypothetical) quantum computer.

Only, gate model quantum computers require error correction. Which doesn't sound bad except that thousands of qubits are required to implement a single error corrected qubit. I.e. when IBM says "we already have a 5 qubit quantum computer", it sounds impressive, but it really is meaningless compared to qubits in the theoretical framework that promises a speedup.

If we're willing to accept unimplemented theory in favor of scalable architecture, then adiabatic and gate model are known to be polynomially equivalent. So far, no quantum computing effort has shown superiority over classical computers, but you don't see this kind of skepticism over IBM and Rigetti's breathless headlines...

Do you mean error correcting a single qubit requires thousands of separate qubits? As a paradox?

Not at all. We have a similar situation with classical computing, where a "bit" of error-corrected memory is comprised of 3 or more bits of physical memory.

> This is all early research.

D-wave was founded in 1999...

Different, and worse, and better. Quantum annealing is basically taking a function and using quantum trickery to find a minimum of that function. Annealing is useful for a certain subset of problems, but in theory a classical quantum computer should be able to solve any problems that a quantum annealing computer can solve.

As a matter of practicality though, a proper quantum computer is much harder to build, so while it may be more powerful, as a cost-benefit analysis it makes sense to specialize.

I have seen it explain as D-wave is comparable to an analog computer that is good at specific problems. A topological quantum computer is comparable to a general purpose digital computer.

> A topological quantum computer is comparable to a general purpose digital computer.

Bold claim. Has anybody made one?

If anyone is looking to try out a quantum circuit, I found the IBM "Q Experience" to be pretty good: https://quantumexperience.ng.bluemix.net/qx/editor. It simulates simple 5-qubit quantum circuits, and even lets you run them on a "real" quantum computer (within limit).

From past articles on Ars as well as some online quantum physics courses from a few years back, there was a lot of healthy skepticism about what the D-Wave was actually doing. I know from the articles here that Chris at least thinks there is some merit to the computer. And in the past few years there seems to be a larger opinion that there is some sort of entanglement working within the machine.

Not to nitpick but a steep learning curve actually implies that progress accrues quickly after starting (learning/time). A flat learning curve implies difficulty.

To nitpick: almost everyone I've heard use this in casual conversation (who isn't nitpicking) uses "steep learning curve" as in physical analogy, e.g. bicycling to a goal, rather than describing a function plot. A "steep" path is harder going than a "flat" or "shallow" one.

A flat learning curve implies slow learning, which could be because there is little (left) to learn, the need for learning is low, or because it's difficult. A steep learning curve could be due to having to start with unusual basics which means you accrue knowledge quickly but you can't do much with it yet.

Quantum computing: eventually we’ll factor numbers bigger than 15 :p

Can we do SAT solving on those?

Looks like DWave can factor 10-bit semiprimes pretty reliably. But you're not wrong about gate model :p

https://www.dwavesys.com/sites/default/files/14-1002A_B_tr_B...

Even though I didn't understand it, I still found the article very interesting. I think I could understand it if I studied for a while. Sadly, I lack the time.

Reply on news.ycombinator.com