Different quantum designs are suited to different workloads. Aliro hopes to provide a common platform to minimize differences and make it easier for businesses to adopt quantum.
Programming for quantum computers is similar in one respect to programming in the late 1980s and early 1990s. An ecosystem of architecture options—including 68k, PowerPC, IA32, ARM, Alpha, MIPS, SuperH, SPARC, and likely others—all had different strengths and weaknesses. Some applications ran quite well on specific platforms, while the process of porting was typically laborious and not quite as performant as the original.
Aliro Technologies, a software company spinning out of Harvard's quantum computing lab, is attempting to bridge that gap for hybrid classical-quantum programs by providing hardware-agnostic solutions for programmers and researchers. This essentially adds an abstraction layer, preventing the need to learn the intricacies of vendor-specific implementations, as Aliro's platform uses validation tools to determine the best available quantum system for the task.
SEE: Quantum computing: An insider's guide (free PDF) (TechRepublic)
Noisy Intermediate-Scale Quantum (NISQ) systems, the currently available type of quantum computers, hold a great deal of potential but additional work is needed to increase the number of qubits in NISQ systems, thereby extending the capabilities of quantum computers. The design of qubits, and how they are connected, is similar in many ways to the architectural diversity in previous decades.
"I think there'll be a lot of competition in the next 10 years or longer with superconducting [qubits] that's popular now, but trapped ion is coming along, there are cold atom plays, there's atomic. I saw your work about topological qubits," Jim Ricotta, CEO of Aliro, told TechRepublic. "There's going to be all these different hardware technologies… and what Aliro wants to do is build a cross platform stack, that can make it easy to get your algorithm running on quantum, at least probably in the quantum-classical accelerator mode, that's where things are starting. And let us figure out what the hardware is, and which qubits to use, and how to allocate them."
Presently, Aliro is working with IBM Q and Rigetti quantum computers, as well as with designs from two other firms that are not yet commercially available. These heterogeneous designs are benchmarked on Aliro's platform, with an eye toward interactivity and workload management. For quantum-classical acceleration, "developers have a choice of how much of their workload they want to shift to the quantum computer, versus keep it in the classical realm," Ricotta said, noting the balance of how much time developers want to commit to using quantum systems.
"Currently, circuits are a very common way of expressing quantum programs. I don't think that'll be the case all the time, but right now circuits are the way, and there are different ways to map circuits onto machines, and onto qubits. Our software lets you try some variations, and seek optimal solutions to those kinds of problems. You do the benchmarking to verify whether [your program is] optimal or not," Ricotta continued.
Quantifying value of a quantum computer
Stages of quantum computing are generally divided into quantum supremacy—the threshold at which quantum computers are theorized to be capable of solving problems, which traditional computers would not (practically) be able to solve—is likely decades away. While quantum volume, a metric that "enables the comparison of hardware with widely different performance characteristics and quantifies the complexity of algorithms that can be run," according to IBM, has gained acceptance from NIST and analyst firm Gartner as a useful metric.
Aliro proposes the idea of "quantum value," as the point at which organizations using high performance computing today can achieve results from using quantum computers to accelerate their workload. "We're dealing with enterprises that want to get business value from these machines…. We think there's going to be a lot of different metrics. What metrics enterprises are interested in helps them decide whether to invest or not," Ricotta said.
How Aliro's stack compares to full virtualization or software containers
Given the relatively meager capabilities of NISQ systems, the prospect of adding an abstraction layer on top of an already low-performing system will likely make some recoil at the prospect of performance regression through the use of middleware. On a scale of one to 10, in which 10 is full hardware virtualization, five is software containers like Docker, and one is tool-assisted porting, Ricotta puts Aliero's platform between three and four.
"We're not ready for many levels of abstraction above the quantum hardware, but we're ready for a little bit. When you get down to the equivalent of the machine language, these things are very, very different, and it's not just what kind of qubits they are. It's noise characteristics, it's connectivity," Ricotta said. "Riggeti and IBM Q machines both use superconducting Josephson junctions around the same number—approximately, the same order of magnitude of qubits—but they are connected in different ways. Therefore, you may take a program, and you might allocate logical physical qubits in a different way for a Rigetti versus an IBM. [Aliro's platform] can make your program work better and give you easier access to this to this powerful new tool."
Aliro is funded in part by Samsung NEXT's Q Fund. Samsung's interest in quantum computing is decidedly long-term.
"On the demand side, there is a need for increasing parallelism on massive problems, like machine learning, or material discovery, or drug discovery. The search space is too massive for any classical [computer]. On the supply side, we have our own things that we do, on the [semiconductor fabrication plant] side of things," Ajay Singh, senior director of Samsung NEXT, told TechRepublic. "In the last 15 years, the cost of a new fab is increased 11 times, to now roughly $10 billion. That tells you the performance per dollar on the supply side—if you were to start crunching on more chips, it is actually becoming less and less efficient for you, from an economic standpoint."
For more on quantum computing, check out "IBM releases quantum computing textbook and video tutorials" and "How helium shortages will impact quantum computer research" on TechRepublic.
- Machine learning: A cheat sheet (TechRepublic)
- Telemedicine, AI, and deep learning are revolutionizing healthcare (TechRepublic download)
- Managing AI and ML in the enterprise 2019: Tech leaders expect more difficulty than previous IT projects (TechRepublic Premium)
- What is AI? Everything you need to know about Artificial Intelligence (ZDNet)
- 6 ways to delete yourself from the internet (CNET)
- Artificial Intelligence: More must-read coverage (TechRepublic on Flipboard)