There are several approaches to building a quantum computing machine. Here's how they're categorized...
Interest in quantum information technologies is growing. Companies and organizations working in this field have been reporting milestones reached. The public is also becoming familiar with its powerful potential. Yet even for engineers, this is a tricky subject to assess.
First, there are several approaches to actually building a quantum computing (QC) machine. Then there are separate use cases, with various requirements. Finally, there are alternative ways to test how effective the different approaches are in solving any given problem.
Several current approaches
Quantum information technologies may seem relatively new, but the application of quantum physics to computer science goes back many years and has led to multiple experimental lines. Here are four leading approaches:
These approaches, in turn, may have variants. Quantum-inspired optimization, for instance, can be based on any number of algorithms. And according to one of the universities that is collaborating with NTT Research PHI Lab, a super-conducting circuit could combine with a CIM.
Some ask whether OPO-based CIM belongs in its own category. Whereas quantum information technologies based on the quantum-circuit mathematical model can always be considered quantum digital, an OPO changes in nature from a quantum to a classical device, making it a kind of hybrid to begin with.
Use cases and tests
Another difference is the kind of problems that QCs can solve. There are factoring problems, which typically require locating a hidden structure, such as found in encryption algorithms. Another group involves combinatorial optimization. The textbook problem is to minimize the total distance in a journey of a traveling salesman to a given number of cities. That type of problem requires looking at every possible combination, the number of which, calculated by factorials, quickly becomes astronomically high.
The search for medically appropriate molecules that can stably attach to a given type of protein is one timely class of combinatorial optimization problems. Others involve compressed sensing (or sparse coding). Efficient answers in the fields of astronomy, magnetic resonance imaging (MRI) or computed tomography (CT), for instance, can result from improving the signal-to-noise (SNR) ratio by discarding large numbers of elements with no useful information.
A final way to assess QC is the test load. Is the machine tackling a 2,000 bit or 1 million bit-class problem? If it is a combinatorial optimization problem, what is the problem size? Those combinations, as noted, can get extremely large, requiring unbelievably long compute cycles. Under the conditions of theoretical quantum information technologies (no decoherence, no gate error, all-to-all connections and 1 ns gate time) for instance, if n = 100, that calculation will take 700 years.
The cutting edge
Meanwhile, over the next five to ten years, the academics and other researchers involved in this field will see advances on multiple fronts. As in any cutting-edge field of technology, there will be some bumps and bruises, if not cuts, as we go forward. When it comes to tracking the progress, keep in mind that there are various approaches to QC. Results, involving different kinds and sizes of problems, will likely vary.
— Kei Karasawa is vice president of strategy, NTT Research