D-Wave Launches Free Quantum Cloud Service | Hacker News Reader

D-Wave Launches Free Quantum Cloud Service

spectrum.ieee.org

102 pointsjonbaer33 comments

Last I knew (a few months ago) we were somewhere less than 20 qubits and getting a straight answer on “does it work” depended on who was being interviewed. How do they have a 2048 qubit chip? And why is it a binary multiple? Did I miss some announcement?

D-wave have spent a lot of effort to show that their machine is quantum, and it seems convincing enough.

However, it's not a computer (i.e. a general-purpose machine, able to run arbitrary programs https://en.wikipedia.org/wiki/Universal_Turing_machine). Instead, their machine is a special-purpose calculator, which finds an approximate minimum value for a certain sort of problem ( https://en.wikipedia.org/wiki/Mathematical_optimization ).

I always take D-wave's claims with a pinch of salt, since their original announcements were incredibly misleading. They claimed their calculator could solve problems orders of magnitude faster than a program running on a classical computer, but they turned out the be comparing apples with oranges: their machine finds an approximate solution, while the program they compared against was guaranteed to find an exact solution (i.e. it would carry on going until all other possibilities had been eliminated).

Other researchers did a more apples to apples comparison, by writing an approximate program (essentially simulating what the D-wave machine does). They found that a normal consumer laptop was actually much faster than the $10,000,000 D-wave machine.

D-wave's newer machines are more powerful, but I don't know if their machines benefit from the scaling properties of quantum computers (their qubits are noisy, which AFAIK prevents them all entangling completely)

What D-Wave is building is not what most other people mean when they say Quantum Computer.

What they build is as far as my lay understanding is something that is much more limited than a "normal" quantum computer. That's why you can't compare qubits from D-Wave with other QC technology.

What D-Wave has is technically called an adiabatic quantum computer and they've yet to show that you can do anything useful with it.

Nitpicking: It is not an adiabatic quantum computer either. Adiabatic quantum computers have as much computing power as the other more usual models of quantum computing (the quantum circuit model for instance). Dwave have a quantum annealing machine, which as far as I know does not do anything better than a classical computer.

Those guys are all building gate based quantum computers which are universal and can run (in theory) any quantum algorithm such as Shor's algorithm. By contrast, D-Wave is pursuing "quantum annealing" which basically solves very specific types of optimization problems and the hope is that it will be economically useful enough in the near future.

See this for a decent comparison: http://moss.csc.ncsu.edu/~mueller/qc/qc18/readings/gottlieb2...

At best, d-wave is making an adiabatic quantum computer, not a gate based quantum computer, at worst (most likely), they're a scam.

It doesn't look like a scam. When you sign up, you get an API key, and you can get an SDK and submit problems to it.

I'm not enough of a mathematician to be able to formulate anything, but there's a growing list of people that have been able to get useful results from it.

No one questions its ability to do some amount of math. The doubt is in how it scales.

Where can one see the results of a useful result.. and by that i mean

Submitted problem X to Dwave cost $Y and returned in Z minutes.

Problem X cost $A and returned in W minutes from AWS...

Nope, at best they have a "quantum annealing machine" which is not more powerful than a classical computer. Adiabatic quantum computers are just as powerful as the gate model of quantum computing.

A universal quantum computer with 20 qubits is kind of like a general purpose 20-core CPU. A 2000-qubir D-WAVE quantum annealing computer is kind of like a 2000 "cores" GPU. It's not a perfect analogy by any means but it should make the difference clearer.

It doesn’t.

Lots of questions about "is this real or a scam", "what's it good for", "what's the difference between this and a true QC". Well...

I have signed up for this, I know very little about the nath, the theory etc. I'm not a computer scientist but apperantly that is what they want, they want people that will come at quantum computing with little to no previous knowledge and come at it with a different point of view..in particular they are hoping for the "killer app". I figure if the Kardashians can get ritch off the internet without knowing one line of code, maybe I have a shot with this!

It will cost me nothing but a bit of time and at the very least I will gain a bit of experience from it. I have not spent much time on it but have watched a few of their videos and they seem to explain things a bit better than the regular "both 0 and 1" crap most people spew.

Here are two links to their videos that cover a lot of the questions asked in this thread.

I say keep an open mind and give it a shot, don't let the nay sayers make you miss out on something that could be a great opportunity.

https://m.youtube.com/watch?annotation_id=annotation_3470308...

https://m.youtube.com/watch?annotation_id=annotation_2178338...

> I figure if the Kardashians can get ritch off the internet without knowing one line of code, maybe I have a shot with this!

Promise me that if you DO strike it rich with quantum computing, you'll change your name to "Hadamard Kardashian"!

So is this actually useful for anything practical? Even with my limited understanding of the state of the art, I'm pretty sure that "factoring large numbers in milliseconds" is not actually something that D-Wave machines can do.

There's some applications listed on their website, this one about Volkswagen doing some traffic pattern analysis was kinda interesting even if I can't follow the math: https://www.dwavesys.com/sites/default/files/VW.pdf

It is an interesting paper, and I can see how the D-Wave architecture would be well suited to it, but, as slide 6 says:

> The Question that drove us... Is there a real-world problem that could be addressed with a Quantum Computer?

So it's a classical case of a solution in search of a problem.

I agree. Even more, their slide is a bit misleading as well. They do not have a quantum computer, they have a "quantum annealing machine" which is not more powerful than a classical computer.

This free API seems like a smart idea then. Since they're a solution in search of a problem, letting the crowd figure out what problems might be suited for their architecture could help them determine what industries they can offer services to.

Im pretty sure d wave never claimed to be able to run Shor's algorithm either

The nice thing that the uncertainty of what they do can be resolved now.

For some time DWave hasn't been making their claims about their systems from the perspective of a black box, the contest has been more about whether we're seeing real quantum behavior inside the machine even through it doesn't appear to have a speed advantage at the scale they can produce right now. To do those experiments you need access to a real machine, so I don't see this resolving anything unless a user of this service comes up with a new problem for which there appears to be an asymptotic advantage.

The superposition of D-Wave's states has collapsed?

To my knowledge, there are still no tasks a D-Wave is better at (asymptotically) than a classical machine

Last time I checked they haven't even beaten semi-recent laptops in direct comparisons of execution time, even for very synthetic problems.

Probably true, but in some ways it's not a fair comparison - since the field is only 20-30 years old, it's not surprising the offerings are comparable to something in the earlier ages of computing. Many kinds of mechanisms that seemed esoteric at their origin (e.g. the primitive but groundbreaking neural networking of the Perceptron) have become everyday reality for us decades later.

What about solving QUBOs ( https://en.wikipedia.org/wiki/Quadratic_unconstrained_binary... ) using Quantum Annealing?

OT perhaps, but there's quite a good introduction to quantum computing by one of Rigetti's engineers at the Bay Area Rust meetup online: https://youtu.be/mrJWpQMx2yo?t=119

Here's the sign up link: https://cloud.dwavesys.com/leap/signup/

Things to note:

* You have to scroll to the bottom of the agreement to activate the checkbox, which activates the sign up button

* You have to be (or say you are) in the US

I appreciate their service. But, is quantum computing going to be costly for dedicated use ($2000/hour)? I hope someone is working on providing a quantum computing service that is cheaper.

Keep in mind that that is an hour block of processing time. If a job only takes a few microseconds, that's a lot of jobs, and you don't have to (nor would you be able to) execute them in just one contiguous hour. You'd spread it out over a month as you refine your processes.

If you're in an industry that hopes to be able to get value from these machines one day, $2k is cheap.

If you're just a hobbyist, the minute or two of time they give you is enough to get started, and if you're good at it, you could probably arrange with one of the universities or national laboratories that have their own machine to get some time on there.

Has anyone compared the terms of these quantum cloud services to get the status of who owns the code & their security?

i.e do these companies have un-restricted access to the user's code. Considering any quantum advantage is worth million $, we need clarification on the IP.

A lot of the services say something about owning the code you send to them to run. But at the present time this really does not matter, as these are only tech demonstrations that barely run even toy models. Even worse with dwave, they do not have a quantum computer as usually defined (something supposedly more powerful than a classical computer).

Reply on news.ycombinator.com