Inside IBM Q System One: an archetype for the quantum computing frontier?
- 17 January, 2019 01:37
Last week at the Consumer Electronics Show (CES) in Las Vegas, IBM Quantum displayed a replica of its first "commercial" quantum computer - Q System One. Following years of painstaking attention from a team of engineers and designers, the technology giant has created what it hopes will be an archetype for the next mode of computing.
This week, Computerworld UK talked with the CTO of IBM Quantum Bob Wisnieff, along with industrial designers Will Howe of Map Project Office and Jason Holley of Universal Design Studio (UDS), to find out more about the boundary-pushing project.
Q System One is touted by IBM as a "universal" quantum computing system for scientific and commercial use. Unlike the company's previous efforts in quantum, System One is designed to be operated outside of a laboratory setting.
The two-year project to create an integrated system that brings the individual components of a quantum computer together in one 'box' can, IBM says, deliver a more stable environment than anything previously created - bolstered by intricate cryogenic engineering that maintains a continuous cold and isolated environment to allow quantum bits (qubits) to perpetuate their all-important quantum state.
"About two years ago we had a project to begin looking at how we will get a commercial quality system," says IBM Q CTO Bob Wisnieff. "As we began to look at that problem, we realised that there's an awful lot of systems engineering potential within quantum computing - to begin to optimise not only each of the sub-systems, at a sub-system level, but to look at the synergistic optimisation between all of the sub-systems, to try to get to the overall best performance for the system.
"The core driver of this - quantum computing - is that as the number of qubits increases, the fidelity or error rate of the system has to be decreased at about 1/n^2. So we want to grow over time and we know that we need to continue to do better and better on every aspect of the system in order to achieve that."
IBM has major electronics projects underway to exert better control over the qubits to minimise error rates. "We have developed specialised gate control pulses in order to control the qubits more accurately," Wisnieff says, "we have developed better methods of reducing the vibration within the machine, in order to eliminate one method of decoherance [when quantum behaviour in a system is lost] for the qubits.
"As we began to put all of these projects together, we realised that a little over a year ago now, there was a tremendous potential to begin to think of it as a totality - and at that point we began working very closely with Map and UDS to do the design of the overall systems."
A synergistic dance
Wisnieff and the IBM Q team believed that by bringing together the various quantum computing elements into close proximity, they could create a "distinct break" from the exploratory-oriented systems that the company had previously built - to involve a system design that had the potential to serve as a patient-zero archetype for quantum computing in a more public domain.
Although quantum computing is unlikely to supplant conventional computing, the ability to solve totally different kinds of algorithms could lead to advances in solving problems that have evaded even the world's top supercomputers - such as unlocking the mysteries of the human genome, shedding further light on the structure of the universe, or even human consciousness.
One concrete example of a problem quantum computing might be able to solve in the near future relates to chemistry and the creation of new, more effective materials. A quantum system might be able to unfurl complex chemical systems such as nitrogen fixation in nitrogenase, according to the American National Academy of Sciences. If this problem was solved, more efficient, low-energy materials could be used for better fertiliser - a product that currently accounts for three percent of the world's total energy output.
Quantum computing is very much at a "primordial" stage right now, to borrow a phrase from Wisnieff. But he is clear that for IBM, the resources it is investing into quantum computing is wholly strategic. Whether this system is, as it stands, as commercially viable a product as the company claims is up for debate - but it makes clear the firm's goals for establishing itself as a business leader in this largely uncharted realm.
"If you think about the path of commercialising quantum computing, there are several things that we have to be able to do successfully," he says. "The first is to get to a sufficient number of high-quality qubits that we can solve problems that are relevant.
"The intermediate scale quantum machines we're working on now, what we aspire to is to have qubits in the range of 50-100 qubits - where we'll cross in to a domain where quantum computers will be able to do problems that are not doable, not feasible by conventional means.
"We know several things about that domain. We don't know precisely what the right algorithm is going to be at this point, but we do know that this quantum computer is going to work very closely with a conventional supercomputer - and that the two as a synergistic dance will solve algorithms in the future."
IBM Quantum, then, is aiming for the "close coupling" between a supercomputer and a quantum computer in one machine.
"We also knew that over the course of the next several years that we will continue to rapidly evolve the capabilities of the machines themselves, going through one generation after another, so we knew that we wanted to build a machine we could do a rapid upgrade on," Wisnieff says.
"[Design leads] Will and Jason - what they came up with was an extraordinarily clever way, a beautiful way, to replace all sub-elements of the machine very rapidly. We can now do in a few days what would typically take us two months - more than a 10x reduction in the time it takes to do a major upgrade on the machine."
By bundling all of the relevant components at close proximity into an integrated system, the designers have also drastically lowered the physical footprint of the system. Not only does this mean more quantum compute per square metre, but it also leads to benefits in machine performance.
"When done properly, bringing these things close allows us to also improve the fidelity, the accuracy of the overall machine," Wisnieff adds. "That is core to us in continuing to be able to make progress."
The anatomy of a quantum system
Will Howe, director of the London-based Map Project Office, is one of the industrial designers involved in making System One a concrete reality. He works as part of a wider design group that includes Universal Design Studio, whose director, Jason Holley, joined him on the call with Computerworld UK.
"It's very, very rare a project comes along where you get to define a completely new archetype for a product," Howe says. "We're hoping that when people think of quantum computing, they'll actually think of System One."
Before System One, Howe points out, all of the different elements that make up a quantum computing system - the highly sensitive equipment and instruments - were disparate and in isolation from one another.
System One is completely enclosed, making it simpler for engineers and researchers to upgrade the computer or to perform maintenance.
Enclosed by a nine-by-nine foot glass panel designed by Milanese specialists Goppion, the cryostat -- the instrument which contains the chip and maintains the extremely low temperature and isolated environment necessary for the stability of qubits -- is sealed away from external sound, vibrations, electromagnetic waves and other such interferences.
The system comprises a series of complex interwoven structures that are consolidated into a single volume, say the designers, with each structure supporting a custom set of components including the "intricately engineered" cryostat where the chip is suspended.
The 'conventional' four-poster frame design that supported the cryostat was iteratively revised into a cantilevered concept. This was bolted to the concrete slab below the raised floor of the system, which "enabled the cryostat to be foregrounded towards the front of the machine, allowing 360-degree access for the engineers as well as becoming the focal point of the design".
The glass vitrine is fitted with a two-axis hinge that enables it to open up and wrap around the sides of the system, while the back of it is where the electronics and the gas handling system that serves the cryostat sit, which are all attached to an independent frame. All of this had to be designed so that the instruments wouldn't touch each other, nor obstruct the heat exchanges, based underneath, which blow controlled air throughout.
The designers say that there were initially reservations about how the cantilevered frame would maintain absolute stability as well as shielding qubits from interference, but ultimately this approach delivered better results than the original - allowing the cryostat to be cooled down to 0.01 kelvin - just above the lowest possible temperature.
Speaking on the instruments within the IBM Q lab, Holley adds that although they were dispersed, there were elements of them that were "very specific" to quantum computing, such as the suspended cryostat, which he thought conveyed a kind of 'character' from a design perspective.
"There was no precedent for a way to design a quantum computer, and this was one of the challenges we faced, which obviously for us as designers is incredible - you can't really rely on anything, so it's a very exciting moment for us and it opens up lots of questions," he says.
"There was something very particular, very specific about [the cryostat] that already begins to make you think: yeah, that is what you would expect a quantum computer to look like. There was something really reassuring to that sort of thing, there was actually already something here - it already has some really interesting design to it - so I think part of what we felt we wanted to do was try and channel that to bring them together into one volume, but in doing that, to really simplify, to really focus the character of what this thing was.
"It's heralding a new era of computing, so its role as a symbol and as an incredible development in technology - they go hand in hand."
Howe adds that the iterative design wasn't "just about creating something iconic" but that they also worked extremely closely with the engineers who use, tune, and program the machines every day.
So at the top of their considerations were how it could be made serviceable, commercial, and the footprint. Howe says: "How can the doors open so it can fit into a relatively small footprint within a dedicated data centre - which is something that's going to happen this year? Incremental steps to bringing it more into a sort of commercial reality."
Quantum-ready to quantum-advantage
According to Wisnieff, hundreds and hundreds of people have contributed to the project, many of whom continued their day jobs at IBM in parallel - for example, taking "the best people in the corporation" for running modelling of the computational fluid dynamics of the airflow within the machine, or for heat-flow calculations.
"This is something for the IBM corporation that is really deemed completely strategic," Wisnieff repeats. "This is the future we want to create, and we want to help lead the world to do this."
However, he concedes that quantum "will not succeed on the efforts of any one group alone".
The company in 2017 launched the open source quantum framework Qiskit, which since has seen community engagement soar to 90,000 downloads when we interviewed research staff member Ali Javadi last year, following the IBM Q Experience which allows anyone to experiment with quantum computation via the cloud.
Qiskit, says Wisnieff, is a "fundamental part" of System One.
"What we have done inside IBM is to reach out broadly to the community with the Quantum Experience, where we have 100,000 users doing over six million runs on the computer. But also through the IBM Q Network of commercial relationships as well - and they're working with 43 partners. We get to see how where, inside different organisations and industries, quantum computing might fit - and what it would take to actually cross the threshold to get to quantum advantage for these individual groups.
"That insight is invaluable - I am firmly convinced that is the kind of insight that will lead to success in quantum computing - being open source, being able to have organisations easily pick it up, allows us to have something that is easily learned - that people can build upon the work of others, and that rapidly becomes a very stable and growing set of software."
Qiskit, then, is "core to everything we are doing".
"That's how we are going to get the insights that will allow us to really be successful," says Wisnieff. "And tying Qiskit down into the higher and higher performance systems over time is what we're about in terms of the IBM Q System one."
What IBM wants to have is not just the archetype for quantum computing but also to be on the ground floor of the evolving Qiskit framework, to improve performance and run "more and more complex algorithms, and become more and more precise in our answers".
Although businesses and researchers tend to be approaching IBM for collaboration, there are certain misconceptions about quantum computing, Wisnieff says, that need to be addressed.
One of the biggest is that quantum is a sped-up version of conventional computing.
"The reality is that quantum computing is computing in a completely different way," he explains. "We are going to use the laws of quantum mechanics to perform algorithms. It requires re-imagining how you're going to re-do your algorithms in fundamental ways. It is not a solution for everything, it will never replace computing as we know it.
"What it will do is extend the range of what can be computed in certain domains that are really important. As we work with partners, they internalise that, and they look at the problems that if they could solve would allow them to do more or do better. Then we can also begin to look at how we can develop those algorithms that work on quantum computers."
In this "early, sort-of primordial state" considerable efforts are underway to understand which of these algorithms will be viable on the machines that will be realised in the coming years.
"What we are really looking for are those first few applications that will push us into quantum advantage," Wisnieff says.
So IBM is 'quantum-ready' to perform quantum algorithms but is not yet at a point where the calculations can exceed the performance of conventional computing.
"But we've made dramatic improvements in the last several years," he adds, "and you can now see that getting to the quantum advantage range is not so many years away as people might have thought years ago.
"What this machine is about is setting up the architecture that will allow us to grow rapidly, and help us fix the problems that we know we will need to fix to get from where we are to where we need to be.
"It is all about building machines that can help us lower the overall error of the machine, building machines that allow us to rapidly evolve the capability of the machine, add new features, upgrade the electronics, things like that. And also to make machines that fit into the machine-room spaces, and are maintainable and reliable in those spaces, so we can move to the point where the machine can be depended on to work on a daily basis."
What could that look like?
"The highest one people think about is quantum chemistry," says Wisnieff. "It goes back to [famed physicist] Richard Feynman in the 80s. The quantum computer would be ideal for looking at inherently quantum problems, and quantum chemistry is one of that class. There's certainly others as well.
"That's something where there's just this wonderful fit at the fundamental level. But that's a no-brainer if you will. Above that I think optimisation and machine learning both have aspects where quantum computers can provide algorithmic advantage that you can't achieve by regular means."
He adds that in these cases quantum will "never supplant completely" what can be achieved by conventional computing, but instead augment it. Quantum augmentation will allow researchers to gain answers more quickly and accurately.
"Those are areas that I think would have a broad application and drive a rapid evolution in the marketplace as well," Wisnieff says.
Form, freedom, future
The design of System One was, of course, informed by complex practicalities, with all these instruments interwoven together - but not touching one another. How much freedom did Map and UDS have to create something that looked aesthetically pleasing - and what informed those decisions?
"We tried to develop a set of principles that would help guide us," explains Howe. "What does a quantum computer look like from the outside? We really wanted to reduce it down to very clean, basic geometry, in the way that you'd think about atoms - and the way that they move around each other - electrons and photons... we wanted to use cubes, and cylinders, and very pure geometry.
"It had to be scalable," he adds. "We wanted to reduce the time it took to upgrade the system, so juts creating these very defined sub-volumes inside the main volume, which could be allocated to certain different systems."
The main thing, Howe says, was to "explore materiality". The cryostat is this "super-polished stainless steel," and the designers wanted to "evoke the idea that inside the cryostat - at nought-degrees kelvin or roundabout there - it's as cold as you can get".
"But how do you invoke the idea that it's super-cold inside? We wanted to do that through material, through stainless steel, people have a tacit understanding that stainless steel is hard and cold to touch."
Holley adds that because of the extreme state that the quantum chip must be held in - 0.01 kelvin, which is colder than outer space at 2.7 kelvin - also informed the final design.
"We felt that it was important to support that sort of understanding here," says Holley. "Trying to make sure there was a stripping down of the essence - certainly form and materially - and help with that kind of storytelling."
Instead of the literal and metaphorical 'black box' hiding the inner workings of a classical computer, Howe writes to Computerworld UK, the language of quantum is conveyed through the unique visual dialogue between the cryostat and vitrine.
This expresses the notion, he says, of "an unwritten future and of territories yet to be explored".