D-Wave's Quantum Computer Successfully Models a Quantum System (arstechnica.com)

An anonymous reader shares an excerpt from Ars Technica: D-Wave's hardware has always occupied a unique space on the computing landscape. It's a general-purpose computer that relies on quantum mechanical effects to perform calculations. And, while other quantum-computer makers have struggled to put more than a few dozen qubits together, D-Wave's systems have already scaled to more than 2,000 addressable bits. But the D-Wave systems don't perform calculations in the same way and, despite all those bits, haven't clearly demonstrated performance that can outpace even traditional computing hardware. But D-Wave has come out with a research paper in Science that suggests that the system can do interesting things even in its current state. The company's researchers have set it loose modeling a quantum system that closely resembles the bits used in the hardware itself, allowing them to examine quantum phase transitions. While this still isn't cutting-edge performance, it does allow researchers full control over the physical parameters of a relevant quantum system as it undergoes phase changes.

Comments

Does anyone understand this?

[Anonymous Coward]

From what I read in the summary, I get that it is normal hardware, with some quantum stuff simulated. It doesn't do a lot, but it does let the researchers run simulations of......itself. Is that right?

Re:Does anyone understand this?

[mrops]

I hear what you are saying, sounds to me like, I built a computer using two reservoirs at different heights connected by a hose and a tap, this system can simulate water flowing from the higher reservoir to the lower when the tap is open. We can monitor when the water flows from higher reservoir to the lower.

Re:Does anyone understand this?

[Garridan]

It's fairly similar to an FPGA, actually. An FPGA can't simulate an FPGA of the same size as itself, but it can simulate a smaller FPGA. They aren't simulating their hardware, they're simulating something significantly smaller.

Re:Does anyone understand this?

[Impy the Impiuos Imp]

That was an AI bot evolving responses.

Don't engage with AI until it can solve problems or make love to you.

Re:Does anyone understand this?

[Impy the Impiuos Imp]

You need storage for X gates, which, like compression, doesn't work for the vast majority of configations. However most realistic programs could fit, but that isn't a fully general simulation.

Re:Does anyone understand this?

[Garridan]

Perhaps wouldn't, not couldn't. The "so what" of this article is that they're reproducing a past result, by doing physics experiments with an entangled system -- it's a kind of verification; about all the "proof" you can ever get about an analog computer, by my understanding. It also shows a use beyond just SAT and factoring which looks rather novel to me, but apparently Feynman suggested it.

that must be the first quantum...

[4wdloop]

recursion?

Qubits?

[mschwanke97402]

At least I did not have to read the usual trite quantum computing explanation again. You know the one, “quantum bits can be 1 and 0 at the same time,” which explains something and noth ng at the same time if you ask me.

Well Whoop-Tee-Do !

[Anonymous Coward]

Call me when Newegg has them on sale for $299.

Otherwise, I just don't give a fuck.

Yo dwag

[rsilvergun]

I heard you like quantums, so we modeled quantums in your quantum computer so you can model quantums with your quantums.

Come on, I couldn't be the only one thinking it...

Re:Yo dwag

[cdsparrow]

You laugh, but at some point this monster will recurse back far enough that it time travels. Then it enslaves us all 500 years ago.

Re:Yo dwag

[Anonymous Coward]

It does indeed sound a bit recursively-ridiculous.

However, modelling quantum systems is one of the "low hanging fruit" in terms of making a useful quantum computer. There are all kinds of fancy algorithms that in principle will run on future quantum computers, and yield results enormously faster than what can be done on a classical computer. For instance all the cryptographic things people usually talk about. On the other hand, modelling quantum systems is a notoriously difficult computational problem. The strong correlation between different parts of the system mean that you can't just use local approximations, and instead have to iterative solve towards an answer that is self-consistent across the whole system you're studying. For instance, if you're trying to calculate the electron density throughout a molecule. Even simulating simple molecules (while taking into account quantum effects) takes a large amount of classical computing power.

However, quantum computers are (naturally) well-suited to modelling quantum systems. So for a given calculation one wants to do, there is often a fairly simple quantum algorithm that can do it for you. So to model the behaviour of some molecule, one designs a set of qubits and quantum logic-steps that are rigorously analogous to the system-of-interest. This of course makes sense: the entanglements of the qubits in the computer are precisely the kinds of correlations one is trying to handle in the computation. Even with a modest number of qubits (<100), one can model interesting (albeit simple) quantum systems faster than a classical computer.

So this is actually a good use for current quantum computers (which are fairly primitive in the grand scheme of things).

In other words: Still pretty useless

[gweihir]

Seriously. This thing does not have global entanglement. That makes it useless as a QC.

Re:In other words: Still pretty useless

[Junta]

Not a simile, he means it is useless when trying to consider it as a QC.

Re:In other words: Still pretty useless

[gweihir]

This thing is not really one QC. It does not have the special powers that a real QC would have or rather it does not have them in the size of bits it has. It really is just several much, much smaller QCs in parallel and that is, due to the nature of a QC, only as useful as the small ones and they are pretty useless. QC computations cannot be subdivided to run on smaller QCs, quite unlike digital computations.

This is not the quantum computer you think...

[Anonymous Coward]

I'm a scientist in the quantum information technology area. It always annoys me to see the marketing machine of D-Wave trumpeting their number of qubits as the ultimate achievement in quantum computing. I could point you to Scott Aaronson's great blog, but the bottom line is: THIS IS NOT A GENERAL-PURPOSE COMPUTER, EVEN LESS OF A GENERAL-PURPOSE QUANTUM COMPUTER.

You see, the power of a QC is not just in the number of qubits. But even assuming this were the case, qubits come in different "quality". If you want to run Shor's algorithm, or in general all those computational tasks that highlight the alleged superiority of a QC, you need qubits of very high "quality" - in terms of entanglement, noise, programmability, etc. The current approach of companies such as Google, IBM, Rigetti, is to try to get a bunch of qubits as "pure" as possible. You can already run Shor's algorithm on IBM's Q cloud quantum computers for example, however these machines are limited to very few qubits - I think the current record is Google's Bristlecone at 72 qubits, and it's not publicly peer-reviewed yet, the only one which has been academically scrutinized is the IBM's 20 qubits one I think. The core reason is the following: building a QC with, say, 20 qubits does not mean sticking together two QC with 10 qubits each. The engineering difficulty is not "double" as much as building a QC with 10 qubits: it grows exponentially.

D-Wave's approach is totally different. They just stick together a bunch of very "dirty" qubits (completely useless for quantum computing in general), but optimize their machines to solve faster certain problems. The specific problems they solve is basically "simulate a D-Wave machine" (!!!) Kidding apart, these machines only solve certain VERY SPECIFIC physical simulation problems. However, these problems are so specific that there is currently no proof whatsoever that D-Wave's machines offer a speedup over classical algorithms at all!

So, bottom line: no, D-Wave's machines are not going to crack your RSA key anytime soon.

Re:This is not the quantum computer you think...

[Jerry Atrick]

A quantum adiabatic calculation supposedly harnesses quantum entanglement to greatly speed up homing on the solution, compared to classical versions like simulated annealing. D-Waves problem is their devices are too noisy, too unpredictable to actually grab all that theoretical benefit.

Worse they've been competing with very poor classical implementations and every time they improve the performance of their shoddy device someone improves a classical algorithm enough to stay ahead. So far no sign they will ever get ahead.

Adiabatic quantum computing

[gotan]

The D-Wave is an "adiabatic quantum computer".
See: https://en.wikipedia.org/wiki/...

This is quite different from quantum Turing machines /universal quantum computers which is usually referred to as quantum computers.

Basically the D-Wave allows to search for a ground state in a system where the quantum states interact in a well controlled way.

The quantum states might represent logical bits, the interactions logical clauses (e.g. A & B = 1) and one might seek a state in which as many clauses as possible are satisfied. Such problems are known as SAT (satisfiability) problems.

Another application could be, that the quantum states represent e.g. (valence) electrons in a crystal lattice (preferably a metal), which interact with their neighbors. The ground state of such a system might give insights to magnetic properties of (abstract models of) materials (when the interaction makes electron spins flip in a coherent manner that might cause ferromagnetism). Determining such ground states with classical computing can quickly lead to time- and memory demanding problems even for few quantum states.

Re:Adiabatic quantum computing

[Impy the Impiuos Imp]

The D-Wave is an "adiabatic quantum computer".

How interesting. Most slashdotters are diabetic fat-tumm'd computer-users.

Proof? What proof?

[tofus]

Currently, D-Wave machines do nothing a classical computer cannot do at least as fast. The only relevant proof, is proof that these things actually outperform classical computers. Which they don't. Not yet at least.

In their current state, they're just really expensive gadgets to scratch someone's really expensive geek-itch. And they're pretty power-hungry as well.

Re:Are they trustworthy yet?

[Anonymous Coward]

D-Wave has shit for PR. Many of D-Wave's detractors act highly skeptical, but then they're nowhere to be seen when Google promises ever-larger prototypes, where we've only seen vapor. When projects like Yamamoto's make brash claims that promise scaling on the back of finite-sized experiments (a practice that this paper avoids), the "skeptics" and nowhere to be seen. When Rigetti solves MAXCUT on, essentially, a hexagon, it's heralded as a breakthrough. Puh-leez. I'd bet a dollar that you can beat their machine on that problem, with a TI-84. Getting down to brass tacks, gate model needs error correction. Nobody's got it. Nobody's even got a good plan for it -- Google's plan is to run without error correction and see if they can do something cool with it. And even if they had error correction, they'd need thousands of qubits to take advantage of theorems which promise a speedup. The fun thing about analog computing is that you need to calibrate every device on these chips. D-Wave is unique in one seemingly-humdrum aspect that never makes the news: they've demonstrated a calibration process that scales to thousands of qubits. Google's failure in December was chalked up to an inability to calibrate a 60-qubit chip. Their next goal is an even bigger chip? Umm, okay Google. If you want to spot a fraud, look for vapor.