Great site, I've shared it internally to some folks here at D-Wave!
215 karma · joined November 17, 2020
Well, that's what patents are for, right?
This is a wonderful time to be a younger person in the industry; there are tons of job opportunities and every hiring manager is hungry for talented, dedicated people.
That said, we do tend to be hiring for more senior roles, considering the small size of the company. If your goal is to land in QC eventually, make sure to have a really broad skill set, get some experience working on hard projects, and apply every year or so until you find a role in the field. No matter which firm you end up at, it's an exciting industry where you can feel like you're actually helping with a grand human effort to push the state of the art.
That's only true of gate-model machines; for example, Google's recent claims of supremacy seem to boil down to something like that. Premature, to say the least, though the work by their team is nonetheless impressive. The pressure to publish is very real! Investors (even when it's a company-internal effort like Google's) still expect some kind of progress even if ROI is not yet being delivered.
Quantum annealers are already solving optimization problems well. One of the goals here is to democratize access to high-quality, high-performance optimization capabilities.
> Nobody is solving any actual optimization problems with a QC in any meaningful sense
Not quite true. Check out what we did in Lisbon:
https://www.dwavesys.com/media-coverage/volkswagen-optimizes...
Clearly this is early days; but the way that detractors talk about this effort is perhaps best server through a metaphor. Imagine you were interested in Babbage or Turing's work early on, but your question for them was, "but how will I use this device to get dinner delivered?"
It's easy for us with a hundred years of progress to see how a connected world enables those kinds of questions to have very robust answers, but it would have been very difficult to see that when you're standing in front of an Analytical Engine full of inscrutable steampunk complexity. If the applications we enjoy so much on the modern Internet were the actual end goal of the technology, it would have been considered unfit for purpose for a very long time indeed! Intermediate, small-scale applications are what paved the way forward, and so too they will be what enables quantum computing research to continue for the likely decades that it will take before it is integrated into apps that consumers use on a day to day basis.
That said, the sort of applications that we talk about around here are often in the realms of science-fiction - imagine the kinds of scheduling and coordination problems that people and machines find difficult today being solved cheaply using open-source software and widely available cloud QPU capabilities. Imagine traffic routing where every single person's destination is calculated together as a tremendous optimization problem in near-realtime to maximize the throughput of a city's streets, or where large-scale resource distribution problems can be solved in maximally equitable ways to help deal with the kinds of challenges we all know are looming on the horizon.
> the company you work for has a long history of making grandiose claims
The company I work for has a long history of delivering tremendous technological advancements, from fabrication to algorithms to production-grade online systems that anyone can access for free. We haven't let Aaronson's opinion stop us from working on that.
Hydrogen fuel cells make way more sense for semi-trucks and I'd love to see the whole Trans-Canada Highway done up with regularly spaced stations, but whether it makes sense in the city is a harder sell. It's a matter of infrastructure being required before users - "if you build it, they will come". I'd drive a Mirai no problem, but not if I'm worried about running out of H.
Well, great; that's an opinion. The thing is, if a device uses the quantum-mechanical properties of the universe to do calculation, then it is a quantum computer; asserting that it isn't one is a matter of semantics and categorization, since what you're really doing is redefining the term "quantum computer" to inherently include "gate model" as part of it, which is not a foregone conclusion yet.
I believe, from what I've read, that at this point current and projected gate-model quantum computers will not be competitive on optimization problems where quantum annealing will be, so there is definitely utility in continuing to pursue this research and development exercise.
> There has been also very little if any actual research in other fields powered with quantum computing.
Here's our latest:
https://science.sciencemag.org/content/early/2021/07/14/scie...
As a physics playground, annealers are very compelling, and materials research will definitely benefit from these devices.
In my post, I literally said: "I would like to know what the article's author would take as proof of this."
What, honestly, would it take? I have been thinking about what evidence I personally would want if I were in your position of incredulity. It's different when you sit next to the things and see the people passionately building them and keeping them working day in and day out, I suppose; I can see with my own eyes that there is no fraud taking place, but I can't exactly bring everyone in the world in for a lab tour.
In those industrial positions, these people are afforded a place to do scientific research that is outside the university proper, but where they can still publish papers, collaborate, release the results of their research to the world, etc.
I think it's valuable, as not every researcher wants to work in the confines of academia. I worked in academia myself, albeit doing far more HN-poster typical things than a researcher would do, and while it can be an amazing experience, it's also a much lower salary, and you're not at all surrounded by the same energy, sense of purpose, and rapid pace of change. You're also dealing with lower budgets and lower expectations when the goal is purely to publish papers instead of creating actual, functioning devices that are not only workable, but eventually useful (which is a pre-requisite to being profitable in the long term).
Formulating problems into the Ising model is still a difficult task, further into the realm of mathematics than the average HN poster usually goes, but it is indeed proving useful for real-world applications, e.g. with logistics and optimizing the order things are done in.
> It is unclear how exactly one can verify that a "quantum code" actually runs on a quantum computer (instead of a classical node inserted between the cloud and the QC provider), and there is a huge window for fraud there.
I would like to know what the article's author would take as proof of this. All I can offer presently is my personal assurance as a team member helping to keep D-Wave Leap running. Every day, I work with a team of talented scientists, engineers, devops, and developers to help ensure everything from pipeline performance to monitoring cryogenics, and if there's one thing that I am certain of, it's that our end user submissions run on real hardware at a few millikelvin about absolute zero.
I am certain beyond any shadow of a doubt that we are using quantum effects to achieve low-energy solutions to difficult problems using annealing. I'm also certain that we're making huge progress; from our massive lead in terms of raw qubit count (5000+) to our work making each of those qubits connect to more and more of their neighbours with less noise over time. There are exciting things coming....
If other companies are getting away with anything less and promising they're doing real-time quantum computing in the cloud, (1) it would be a huge surprise, and (2) their lives must be a lot cushier than ours, because it is a lot of work keeping something like this running. You want to talk about the woes of having to manage on-prem and hybrid cloud workloads, well, does your datacenter have plumbing for liquid helium? You monitor the temperature on a few server racks, but do you have to measure ten thousand different datapoints about air and fluid temperature and pressure?
Honestly, it's a lot of fun, it's an exciting thing to be working on, and I don't agree with the author when he complains about brain drain. You want brain drain, go and look at the infinitum of startups hawking SaaS grift-ware like it's the next best thing since sliced bread. Sorry if we find it more interesting to work on this than on the next B2B way of slicing off a chunk of someone else's revenue for providing something obvious. "We do these things because they are hard", as the saying goes.
No squinting required.
This is exactly the kind of problems that the D-Wave quantum computers are designed to run, and in fact simulations of physical systems are one place where we've been able to demonstrate solid progress, e.g. https://science.sciencemag.org/content/361/6398/162 and several other papers.
I'd highly encourage anyone interested in the Ising Hamiltonian and its uses for optimization problems to check out what we're doing and sign up for D-Wave Leap at www.dwavesys.com - we have demos, an in-browser IDE, Jupyter notebooks, a comprehensive Python SDK with excellent docs, and we'll give you free QPU time to run your problems in realtime on a quantum annealer.
D-Wave is seeing really promising results with optimization problems and materials simulation. The real barrier to entry right now is not access to the machines, or the middleware needed to use them - it's the ability to actually harness Binary Quadratic Model or Discrete Quadratic Model problems to apply to something that delivers real ROI.
There is a lot of training material available, but there's still a difficult road between systems that mathematicians and computational physicists can work with, and systems that are turn-key enough for a business analyst with a stats background to integrate into a reporting or planning workflow. Of course, anything in software that is a 'difficult road' is also potentially a goldmine for anyone that pulls it off.
It's not actually a problem at this point, because cryogenics technology is really advanced, to some extent commoditized - if you want a fridge, you can buy one, though at D-Wave we've pushed the state of the art on these systems far past what you'd get off the shelf, to the extent where we've had production systems spend years cold with only routine maintenance.
For the foreseeable future, it makes way more sense to interact with a quantum computer remotely, the same way you interact with your Hadoop cluster or your GPU farm; you don't want to work right next to it, it's noisy, and unfortunately unlike a Cray it doesn't provide built-in seating. I should know, there are a large number of these machines a stone's throw from my desk...
Can't wait to fire this up and start re-doing all my alerts.
Admittedly, it's not a gate-model machine that can run Shor's Algorithm, but it is a quantum computer, and at more than double the number of qubits plus far higher inter-qubit connectivity than our previous D-Wave 2000Q, it definitely demonstrates tremendous progress.
If you have a moment, you can sign up to use it for free at https://cloud.dwavesys.com - we have an online IDE, Jupyter notebook training material and tons of docs, a community forum, and of course some shiny demos that submit problems to the live QPU if you want to try them out.
The gate-model systems have shown that they are pursuing a much more difficult path, one that may indeed be fruitless for years or decades before they can approach our raw qubit count. We also have examples of nontrivial, paying-customer use cases that become more compelling with each new announcement. Factoring integers is indeed an interesting hard problem which would have a massive (negative?) impact on the world if realized, but besides Shor's and Grover's algorithms, it's not like gate-model QPUs have a ton of use cases significantly better than what a quantum annealer can accomplish.
I like to think of it this way: we're basically at the point of an ENIAC scale machine, if you liken quantum computing progress to classical computing. Fills up a room, very specific environmental and power requirements, little or no "memory" to speak of, esoteric and hard for anyone without years of training to master. Only a few decades later, the state of the art machine was thousands of times more capable, far cheaper and smaller, more reliable, more accessible in every way. Imagine describing the Internet as we use it today to an ENIAC operator, or a speculative investor considering IBM, Honeywell, etc. - it would sound like an impossible, Asimovesque dream, not something that children would literally be playing with sixty years later.
The only difference is that so far, we don't really seem to have a real exponential Moore's Law effect in quantum computing. Google et al. still have very low numbers of qubits without any real promise that they'll be able to deliver more of them in any consistent timeframe. At D-Wave we've done better on the scaling front, and we've been trying to keep up to our former founder's "Rose's Law" of qubit scale growth, but fabrication is an incredibly expensive, complicated, competitive endeavour that necessitates incredible quality control in order to produce processors that are up to spec. There are also other factors beyond the raw qubit count; the bigger advantage in our latest Advantage chip may actually be the higher connectivity between qubits on the graph, rather than their raw number.
Of course, we expect that we'll continue to push the envelope in this regard, and given enough time and investment, some of the early applications we're seeing now may well eventually be integrated into large scale products people use every day.
It's actively in production now with our new Advantage chip with 5000+ qubits (exact yields vary as fabrication is truly a difficult thing to get perfect, but each chip that's been made available to customers is at least 5000 to my knowledge).
> I hope things are running smoothly over there, it was for sure one of the more promising approaches I saw.
We have a decent number of investors with NEC being our latest, and contributions from the Canadian government's various innovation funds are also helping. It's an expensive business to run, and we've been in it for longer than anyone, so it makes it clear that this is a long term play. I personally like to liken it to investing in a Babbage or Turing sort of project: it's clear the seeds of some amazing future technology are here, but the current scale of the machine is not yet at the level where it can unlock all the things we're dreaming about yet.
That's where the hybrid solvers come in: we can basically present a way larger capability in terms of our working graph for Ising problems than fits on the QPU, and chop it up and get great quality answers that really make use of the underlying quantum hardware while keeping up with the best classical software.
> I suspect the killer application could actually end up being neural networks.
Yes, that's long been one of our expected markets, and something we hope to see increase over the next few years. One issue is that a lot of ML developers approach things from the far other side of the stack - like most hackers, picking up a new toolkit and exploring it from the outside in, rather than academically learning every aspect of the model and how it works inside. Unfortunately, it's that latter skillset that is still required to do the work of mapping an ML model onto the Ising model - way above my own personal ability to even approach.
> You have some robustness to noise and actually you really don't care so much that the answer is not even the final answer, just that some progress towards the final answer is made.
This is exactly the strength of quantum annealing: getting lower energy state solutions in a shorter time than a classical method by using quantum-mechanical properties of the universe, like tunnelling.
> I took a very quick crash course in quantum computing some 5 years ago (maybe more) and worked on some toy problems. I should check it out again. I guess you're now at the stage where you have some re-programmable setup? Or is this a simulation QC machine?
No simulation at all! You fire up the IDE from your account at https://cloud.dwavesys.com/leap/ and you will have an API token that works with our Ocean Tools SDK to submit problems to the live QPU. We also have Jupyter Notebooks in our online training session that will submit problems live. No simulation, and it's totally "reprogrammable" in the sense that you can submit whatever QUBO / Ising problem you want, within limits, and it will be sampled and a solution will be returned to you within seconds.
> I would love to hear from you about what some people have been able to do in terms of applications.
Our big win recently was Volkswagen using the system to route busses in Lisbon live with the QPU as a back end. https://www.dwavesys.com/media-coverage/volkswagen-optimizes...
We also just got a paper in Nature with simulation of magnets: https://www.nature.com/articles/s41467-021-20901-5.epdf and have done other work with spinglasses in the past.
Current customer efforts revolve around anything from physics simulations, to logistics work, and anything else where optimization functions come into play. There's actually a really wide range of optimization work people are already doing that can be tied into our system; for now, it's a matter of creating the right customer middleware to bridge the gap.
You're not far off on some of your assumptions for how it ends up working in reality. We do see it as an accelerator for specific operations, specifically for optimization problems; however, much as people originally saw GPUs as being for gaming, there's a lot of room for creative individuals to explore the capabilities of the system and see what other uses it can be put to.
We have a cloud IDE now with excellent visualization tools, and it's free to get started. You should definitely go check it out, considering how interested you sound; it should be very educational. Any developer with some basic Python skill can learn how to fire off toy problems to it, the hard part where you need more of a statistics and hard math background is in mapping real-world optimization problems to the Ising model we use.
For the most part, there aren't enough qubits yet even on our 5000+ qubit machines that anyone would want to share the QPU with someone else during an execution cycle. In fact, this is the reason why we have a Hybrid Solver that combines classical and quantum technologies - people want to work on much, much bigger problems than fit onto the working graph of the QPU.
Does anyone know whether there's some expectation that at some point we'd see first-party apps for common social networks on PureOS? Like, are they going after it at all, or is it intended as a web-first play?
---
As a couple of people have already pointed out, the machine you're looking at isn't really the quantum computer, it's the apparatus that's used to keep the processor in a nice cold, quiet environment (a very necessary and complex endeavour when working with delicate quantum systems). Superconducting processors require dilution refrigeration, a process that uses the two common isotopes of helium to extract heat from the fridge and its payload (although other QC implementations may also require some cryogenics, the specific apparatuses may look quite different); Oxford Instruments (one of many commercial fridge manufacturers) has some relatively accessible documentation here: http://home.agh.edu.pl/~kozlow/fizyka/otrzymywanie%20niskich...
Gold is a commonly used material in cryogenics for a number of reasons: 1) it's soft and conducts heat really well, so by coating pieces made of other materials like copper with it and pressing them together, we can get good thermal contact; 2) it's shiny and doesn't oxidize so it reflects light really well since radiation (light) is another form of heat transfer, this again reduces heat absorption. Overall, material choice/design in cryogenics is complicated because many materials' properties change drastically at low temperatures (fun fact: dental floss is often used for this reason).
Depending on which images you're looking at, the "tubes" may indeed be coaxial cabling to carry microwave signals to the processor, but there are also tubes that act to circulate helium and as structural support of the various stages.
D-Wave's quantum annealing processor requires very few microwave signal lines, one of the reasons we've been able to scale it to over 5000 qubits.
For further exploration of quantum computing systems, I recommend this tour of D-Wave's lab: https://www.youtube.com/watch?v=VfxNdBTH8wY and this WebGL interactive 3-D tour showcasing the D-Wave Advantage processor: https://www.dwavesys.com/d-wave-advantage-showcase
I believe a new lab tour video is being produced right now and should be available soon, so check back on the youtube channel.