Quantum computing – exciting and off-putting all at once – is a kaleidoscope of technology and market questions whose shapes and positions are far from settled. Hyperion Research (formerly IDC’s HPC group) is now ramping up efforts to aggressively track this emerging sector. Led by Bob Sorensen, who recently added chief analyst for quantum computing to his title of VP of research and technology, a big chunk of next month’s HPC User Forum agenda (April 16-18), organized and run by Hyperion, is devoted to quantum computing and will serve as a kind of introduction to the new practice.
Hyperion, of course, is hardly alone. Quantum computing is garnering attention from all quarters including market watchers, technology developers, industrial users, and governments alike. Notably, additional funding for quantum computing was included in the Trump administration’s latest budget request for the U.S. exascale computing program (see HPCwire article: And So It Begins…Again – The FY19 Exascale Budget Rollout). That quantum computing is (or will be) important seems agreed upon. Among organizations with quantum presentations at the next HPC User Forum are Google, Microsoft, Intel, Rigetti Computing, NASA, NIST, MIT, IBM, and D-Wave.
In a wide-ranging conversation with HPCwire, Sorensen discussed the triggers for Hyperion’s increased attention to quantum computing, emerging geopolitical rivalries in QC, the need for algorithm and application development, and outlined some of the goals for Hyperion’s quantum computing tracking plans; the breadth of some of the latter may surprise you.
“We have been tracking quantum computing development for a number of years, almost in a maintenance mode kind of a way where if there was an interesting development we kept in touch,” said Sorensen. “What happened in the last few years is we have moved beyond the stage where we are tracking interesting physical qubit development – the idea that Google or IBM or Microsoft or Rigetti has an interesting new design for a qubit piece of hardware. It’s moved now more into what we consider to be the stages of a more complete quantum computing (QC) ecosystem that encompasses hardware, software, application development and perhaps most importantly QC algorithms.”
Fundamentally and for the longest time, said Sorensen, quantum computing was a solution looking for a problem. Now, instead of just chasing qubit technology – and lots of that is still going on – it’s about identifying use cases and building tools and including domain specialists.
“IBM is a company that gets it. It’s not only developing this quantum computing hardware, but also putting out an ecosystem that allows people to play with a quantum computing simulator in a relatively low barrier-to-entry kind of way, accessible through the cloud, that encourages people to start to think about how you develop QC algorithms that matter. To me that was the final tick in the check list that said this is going to take off now,” said Sorensen. “Once we start to see some sophisticated algorithms coming out then the next step is when can we start to pump those it applications.”
Ever hear of Shor’s algorithm? Everyone else has too! That was a problem, said Sorensen. In spite of the hype there really haven’t been many new and interesting quantum computing algorithms. A handful of uses cases have dominated the discussion. Acquiring a better understanding of nitrogen fixation processes is a favorite. And it’s important. But others are needed and the new tools will foster their development. There are many problems whose solutions are largely intractable with classical von Newmann computing architecture but which seem well suited for quantum computing.
“It’s not just about building the hardware anymore. So IBM has a quantum simulator. You’ve got Atos which is making a simulator of a quantum system available – they are going to sell an appliance which in some sense is a nice little supercomputer in a box that has all the outward appearances of a quantum computer. Again, that’s trying to foster the ecosystem around algorithms and applications in addition to the hardware. Look at Rigetti. I love the fact they call themselves a full stack QC company, which means they are going all the way from essentially building the QC hardware all the way up offering programming language and an application development environment.”
“One of the things I am hoping this program draws attention for is the idea of collecting quantum computing grand challenges as we used to have in the HPC world. What were the big problems of the day that we needed HPC for,” said Sorensen. “The compelling issue of grand challenges is they were accessible to someone who didn’t have to know what HPC was or MPI was or multicore programing. They just knew that this machine could solve a compelling use case in pharmaceuticals, in oil and gas, in medical technology and that was enough. What I am hoping for now is to gather counterpart quantum computing grand challenges.”
Sorensen has broad goals for the new practice. He has also reached out to the quantum computing community seeking participation on an expert panel to help guide Hyperion efforts and stimulate the quantum computing conversation more generally. That said, there are the usual market analysis questions.
“How big of a market will there be for quantum computing and how will it break out? How much hardware? How much software? How many of these [quantum] simulators? How will this affect the HPC sector? Where will the revenue streams come from? If we could wave the magic wand and come up with a nice market forecast for the next five or six years out, that was highly credible and based on insight, at this point that would be a Holy Grail. That’s not going to happen I think the very near future,” said Sorensen
“This entire field is in a huge state of flux right now. I wouldn’t even call it the Wild West. It may be another decade before it is the Wild West. All we are trying to do is get a sense of what the experts, the people who are involved in the day to day issues of pushing this technology forward, what do they think about where this is all going. [At least initially] I think what we are going to find is there is no right answer. We are just staking out the landscape boundaries.”
An important element in understanding quantum computing’s boundaries is figuring out how ubiquitous it will be. Sorensen currently inclines towards the special purpose machine camp.
“I see quantum as an always-somewhat-esoteric branch of computing and I’d liken it to, would you buy a Cray to check your email? Because of competing price performance issues you will always go towards the system that does the job at the best level of price-performance and there are huge quantities of things that quantum computing will never do that will justify that kind of price performance model. You know, simple transaction processing. Making sure that when you order a sweater from Amazon it gets shipped the next day. I don’t see quantum becoming the end-all-be-all when it comes to the computational platform of say 2030 or 2040. At least for the next 20 or 30 years I think it is going to be…I don’t want to say a niche technology, but perhaps a tactical technology, one that has a very specific set of use cases,” said Sorensen.
Those uses cases, however, may turn out to be disproportionately pivotal. Solving the nitrogen fixation problem could be game-changing in fighting world hunger. VW is already experimenting with D-Wave’s adiabatic annealing quantum computer – an admittedly special purpose machine – for understanding traffic management patterns. A fair amount of early global jockeying for the lead in quantum computing has emerged.
Sorensen, a long-time technology analyst in the U.S. government before joining Hyperion, said, “Technology development at the pointy end of the spear can move in fits and starts where country X or Y can have one- or two-year lead. The implications of those technology leads can take longer to play out and require a myriad of other factors to come into play before it can realize either a geopolitical or economic or military sense. I am not deeply concerned about the impact about say a China or perhaps even a pariah nation pulling ahead in quantum computing.”
The U.S., Europe, and China are steaming ahead with vigorous quantum programs. Less is known about what some other rival geopolitical regions are doing in QC.
“Europe and China are putting a lot of interest in this, and the U.S has a long history. To me there’s the issue the balance between centralized government programs and market-based activity. You’ve got an $11 billion quantum project in China right now where they are literally putting all of their eggs in one large facility to develop QC. Then you’ve got the U.S. model in which there certainly is a government element but also much more of an entrepreneurial spirit. You can’t argue that companies like Google and IBM and Rigetti and Microsoft don’t have their own innovative capabilities here that are independent of what is going at U.S. government sites. So there’s a certain vitality there,” said Sorensen.
“The EU has large government programs, which may not have the full breadth of commercial development activities [the U.S. has], but I think they view this as one of the most level playing fields in advanced technologies that they have seen in a long time and that their history and current capabilities stack up with anybody else in the world. I think they want to build on that.”
Clearly much remains unclear about quantum computing’s future. That’s probably good news for Hyperion and other analysts – the murky waters need some clearing. Sorensen has high hope for the growth of Hyperion’s quantum expertise. Moreover, just as the HPC User Forum has provided a platform for stimulating discussion and influencing HPC policy, Sorensen thinks Hyperion can play a role in QC.
“One of the things I’m hoping we can at least play a role in is the idea of thinking about quantum computing benchmarks. Right now, if you read the popular press, and I say ‘IBM’ and the first thing you think of is, yes they have a 50-qubit system. That doesn’t mean much to anybody other than it’s one more qubit than a 49-qubit system. What I am thinking about is asking these people how can we start to characterize across a number of different abstractions and implementations to gain a sense of how we can measure progress,” said Sorensen.
“What is a benchmark that says it’s progressing the state of quantum computing forward based on these [agreed upon] kinds of performance parameters or metrics. As I’ve said to people in the past, I don’t want to end up with a Q-impact where we have a one-size-fits-all benchmark that forces us in some sense or strongly encourages development in a direction that may not be the best and only relies on something the way some people say LINPACK does.”
Grand goals aside near-term progress on many specific fronts, error correction and algorithm development, for example, are needed said Sorensen. He worries quantum computing may fall victim to the “trough of despair” as AI once did.
“There are pitfalls ahead in the next few years. If quantum computing is not done right we could see the trough of despair lead to the AI nuclear winter that we saw in 80s and 90s when there was a huge promise for AI and everyone thought it was going to save the world and it got difficult. It was overpromised and it disappeared for almost 20 years. I know that because I am looking at all of my graduate school text books on AI and many of them mention convolutional neural networks. I’m worried that quantum computing may be oversold from the investment perspective. We want to scope out the landscape instead of just coming at it and saying here we are. We want a good understanding of that and the first step,” said Sorensen.
Link to upcoming HPC User Forum agenda: https://hpcuserforum.com/downloads/agenda.pdf