It seems that they did build a probably quantum but quite likely not computer. They are actually solving some optimization problem. And they do this quite well. However as far as I understood, they did not convincingly demonstrate that they are actually using a quantum version or a classical version of the system. And quite likely they can not run arbitrary computations. So it seems that the jury is still out, if they did build a quantum computer or a very specialized classical hardware.
There have been some calculations that are consistent with entaglement which is a premise for quantum behavior.
However D-Wave isn't a general quantum computer. It's an adiabatic quantum computer. I'm pretty sure it can't solve Shor's algorithm, because it only does function optimizations.
This is absurd. That's a comment on a blog post, proposing a bombastic title, "Better than Shor". I'm pretty sure if they could get anywhere near Shor using their architecture they'd get more than a sympathy Nature paper.
The truth is, D-Wave has done nothing to foster the coherence of their "qubits", and so they have built a very expensive, superconducting, classical computer. There is no quantum trick, actually. You have a system of buckets and you adjust the pipes and pumps between them so that in steady-state the water levels solve some problem you program in (with the pumps). Then you start in a random configuration, turn on the earthquake machine for a bit of randomness, and end up in the "ground state" solution.
The exponential speedup of quantum algorithms is sketchy enough: good luck performing an N-qubit logic gate, or in assuming that it's your favorite black-box that need only be executed a polynomial number of times (every quantum computing textbook uses their own set of universal gates). But these guys aren't even trying.
The truth is, D-Wave has done nothing to foster the coherence of their "qubits", and so they have built a very expensive, superconducting, classical computer. There is no quantum trick, actually.
Scott Aaronson, the self-described "Chief D-Wave Skeptic", disagrees with you:
Now, I’d say, D-Wave finally has cleared the evidence-for-entanglement bar—and, while they’re not the first to do so with superconducting qubits, they’re certainly the first to do so with so many superconducting qubits. So I congratulate D-Wave on this accomplishment. If this had been advertised from the start as a scientific research project—”of course we’re a long way from QC being practical— —but we’ve shown experimentally that can entangle 100 superconducting qubits with controllable couplings”—my reaction would’ve been, “cool!” -- http://www.scottaaronson.com/blog/?p=1400
If they actually do have a factoring algorithm, it's easy for them to prove: factor something we give them that is obviously unfeasible for classical computers and give us the result.
Quantum computers can't solve arbitrarily large factirisation problems. The signal gets weaker the largers the numbers get to the point where factoring a 4096 bit number is not going to happen.
Real quantum computers don't exist. If one did exist it would need to factor numbers much larger than 4096 bits efficiently for it to be considered a real quantum computer.
And all true scotsman wear kilts. We consider 8bit computers real computers even though they can't run windows 8. The same can be said for quantum computers, a 64qubit quibit quantum CPU could be vary useful even if it can't factor 4096bit numbers.
PS: We have solid state 4bit quantum computers right now. They may not be useful for much, but that's in part because we have spent hundreds of billions of dollars perfecting classical computers.
Like you, I was worried about the snake oil aspects of qc, but ...well, I got over it when I looked into the science of the d-wave implementation and talked with folks like Sergio.
Please note,we're talking about a quantum annealer here, not a full on qc and we're okay with that :-) Also, disclaimer: I'm a googler who worked with our quantum folks and nasa to get the machine in place.
Not that I know of, but my most recent information comes from this over-a-year-old post: http://www.scottaaronson.com/blog/?p=954 Google buying one doesn't make me shift much one way or another -- D-Wave has built something interesting that seems to be faster than other hardware, they are able to improve from the marketing buzz around 'quantum', but may or may not be using any quantum effects. People like http://jeffjonas.typepad.com/ have algorithms for problems that do increasingly better than the competition with increasingly "harder" or larger instances, so that behavior isn't really that weird.
It is unclear if they are actually taking advantage or quantum effects or if similar machines could be built using much simpler and cheaper classical effects.
Recently I read an article [1] about this, and it seems that D-Wave has a large advantage in optimization problems that pop up often in machine learning.
I think this is a lot less esoteric than it sounds: a company develops a computer that is a lot faster at optimizing using a quantum trick, we do classical machine learning faster. I'm pretty certain they're not doing anything qualitatively different than before. No (largely debunked) Penrose theories on quantum consciousness, sorry.
But their demo software is to solve the traveling salesman problem - with 4 nodes. I can do that on paper.
If the bloody thing works, why not show us how to solve it for 50 countries, or all the counties in Ireland, or something that shows it actually doing something cool?
I think if there is someone I would not want to own a quantum computer, next to Lockheed Martin, thats Google. They already are at least 10 years ahead of everyone else.
If they are the first to spend the money on seriously making use of quantum computers, who on earth is going to stop them if they decide to make bad use of their power?
I'd really like to see this power in a companies hands that is not against the concept of privacy.
Well, he did include "at least" which suggests he was just coming up with a rough figure. I somewhat agree with him.
There is no denying that Google is (arguably) one of the most powerful companies out there. Are they ten years ahead of everyone else? Maybe, maybe not. But one thing for certain is that they have the computing power, finance,and connections to pretty much pursue whatever they want.
Is that a bad thing? Who knows. It really boils down to whether anyone/anything should ever have that kind of power.
"Quantum" is a bit of buzzword in a sense that people understand it in a very narrow sense. D-Wave's system is NOT a "classical" qubit-based quantum computer. What it does is essentially global optimization of multimode functions using quantum effects.
True, but significantly faster global optimization of multimode functions by any means turns out to be pretty useful for exactly the types of ML problems that are stretching the limits of current hardware. This seems like a promising first step towards technology that could be really useful if it comes anywhere close to doing what it claims.
Interesting, I thought D-Wave was still a bit ropey and quantum computing was a bit of a fussy problem to get working consistently. Although it comes as no surprise Google is getting on the quantum computing bandwagon nice and early, but the idea of the pinnacle of algorithmic research with quantum computing being done by a company that likes to optimise data processing to help advertise to you more effectively is...well, frankly somewhat ridiculous.
While I loathe Ayn Rand I have to give her some credit for capturing accurate human behavior in Atlas Shrugged: the D-Wave skeptics seem to behave exactly like Rearden's detractors.
But in the face of it all, D-Wave keeps plugging along, landing larger contracts, and hitting more impressive benchmarks. They may fizzle out in a flame of failure in the end, but as it stands, they seem like they're doing a great job.
Whenever I hear about Google doing some genuine technology, it just makes me a million times more fucking angry to compare what Google must be like if you're a Real Googler (machine learning projects, basic research, quantum computing, meaningful work) to the misery that the other 90% endure.
I feel like this should be much more of an HR problem than it seems to be. Whenever it comes out that Google's real members are working on a neat AI or robotics problem, everyone who's stuck working on some dead-end maintenance project for some terminal middle manager ought to go to a VP and say, "I want on this cutting-edge technology now or I will fucking quit." That 18-months bullshit should be cause for shit to get blown up.
Right... And someone needs to maintain the boring systems so Google can operate their core services (search and ad sales, say) to pay for the neat stuff that the (ahem) 'real' Googlers do.
Some employees may be more easily replaced than others, but those dull jobs are often equally important.
If you have an open allocation system where people are free to work directly for the company, then people spend some time on the research work that interests them, and some on the maintenance and upkeep of critical systems. Why do they volunteer for less attractive projects? Because they actually care, because the company is (imagine this) worth caring about. Once the maintenance work is done by volunteers rather than extortion victims, it starts getting a lot of respect and people to do it to be "heroes". So yes, it still gets done.
On the other hand, if you work in a careerist closed-allocation hellhole where your boss has unilateral authority to whip it out and flow a shot on your performance reviews/"calibration score" (i.e. managerial extortion) then, obviously people are going to look out only for themselves and the maintenance won't get done unless someone's forced to do it (in which case, it's done very badly). People will not freely participate in a closed allocation company's upkeep because the work culture of a closed-allocation firm is not worth caring about or protecting.
They're both equilibrium states. In one (open allocation) people see themselves as citizens and do the full mix of the work: the important and fun stuff, but also the less-fun but critical "hero" work. In the other (closed allocation/extortion economy) people realize that the company has no consideration for them and that no one's looking out for their careers, so they take anything that's not nailed down and the critical upkeep work doesn't get done.
The way I see is, there is always going to be a 'somebody has to do this dirty work' mentality everywhere. No matter where you work at.
What work you get is very similar to being born. A baby born could have easily been somebody else. If you look at this way, you are almost definitely not going to get those big projects unless you are into some kind of a political game or are simply good enough and get your due chance.
One thing I've learned the hard way is, a person must not 'wait'. He must not wait for chances to come, or the ecosystem to help him, or somebody else to do it for them. Couple of days back I was reading Aaron Swartz's essay on productivity. In which he mentions a very important point- Always choose the most important problem to work on. And if you follow that advice seriously. You will someday find yourself working on most important problems in the world.
In the end, its just comes down to you. In large corporation project allocations are random at best, completely devoid of merit.
If you look at this way, you are almost definitely not going to get those big projects unless you are into some kind of a political game or are simply good enough and get your due chance.
You would have had a more accurate statement if you truncated it before the "or". Closed-allocation companies always have the most important work done by the politicians, not the people most adept at doing it. There are no exceptions.
In which he mentions a very important point- Always choose the most important problem to work on.
This will get you fired. It means that you are pursuing your own career goals (and benefiting the company, but not your immediate managers) rather than your boss's and you will be shit-canned as soon as he finds out.
You will someday find yourself working on most important problems in the world.
No, you will be long-term unemployed if you follow that strategy.
I have a lot of respect for Aaron Swartz, but look at what the fuckers in power did to him.
Ex-googler: While I think open allocation is toxic and the reason I left, the best way to get into one of these projects is to get Acquihired into it.
That said, obviously, many of those there are more than willing to do the dull work in exchange for the very nice perks and compensation. I just wasn't one of those people.
Ex-googler: While I think open allocation is toxic and the reason I left
I think you mean "closed allocation".
the best way to get into one of these projects is to get Acquihired into it.
Isn't that fucked up? What does it say about a company when it can't trust the people who've been there for years to do any of the real work and it, instead, buys mediocre talent at a panic price (typically $3-6 million per head) on the acq-hire market?
google has done a great job so far of serving their customers and shareholders. do you have any evidence that forcing open-allocation on google as a whole will improve the results they are able to deliver?
