Telescope Magazine: What led you to develop the novel CMOS annealing machine?

Masanao Yamaoka: With the goal of achieving a smart society, we’ve been focusing on advancing computing technology that utilizes cyberspace known as the cloud. However, remotely accessing the cloud system is often impractical, especially where instant response and real time processing are required. To overcome this obstacle, IoT devices and connecting devices on the cloud edge (such as network gateways, routers, local mobile base stations, etc.) must also be equipped with reasonable computing power. Which means that technologies that utilize both the cloud core and the edge have to be developed, as well as the means for optimizing such a system. In an urban traffic control system, for example, resolving traffic congestion requires balancing the traffic volume and the cost of mobility. By using such input parameters as the traffic conditions and the destination of each car, the optimal traffic signal operations can be determined. In fact, efficient solution of an optimization problem like this is essential in achieving a smart society.

The “traveling salesman problem” is an often-quoted example of combinatorial optimization. Given a fixed number of cities to visit, what is the most efficient route for a salesman to visit them all before coming back (Figure 1)? It’s fairly easy to solve when only a small number of cities are involved, but the calculation becomes increasingly difficult as the number of cities gets larger.

Figure 1. Solving a traveling salesman problem
Courtesy of Research & Development Group, Hitachi, Ltd.

Other types of combinatorial optimization problem include minimizing the cost of supply chain logistics, and optimization of power generation volume and transmission routes to stabilize the supply of electricity. Also, a venture capital firm may want to employ this technique to choose which ventures to invest in and in what order.

The general consensus is that the long reign of Moore’s Law in the semiconductor industry is finally coming to an end, and the conventional computer architecture also appears to be reaching its limits. I felt a new approach was needed to solve this problem, and decided to develop a computer that was totally different from conventional ones.

To sum it up, I realized that combinatorial optimization problems were about to become critical, and so I chose to tackle those problems by developing a new computing method based on an unconventional architecture.

Telescope Magazine: That reminds me of Rebooting Computing, an American initiative to develop post-von Neumann computing models.

Masanao Yamaoka: That’s exactly the kind of thing that we are after. At the IEEE International Conference on Rebooting Computing (ICRC) held in the U.S. last November, we gave a presentation on the technology based on simulated quantum annealing (Reference 1). Our approach is in line with the efforts that are going on in the U.S. Our concept of post-von Neumann computing includes diverse architectures, such as quantum computers and AI.