The Latest Trends in Quantum Computing [Part 1]

—Quantum Annealing to Transform Society—

Main visual : The Latest Trends in Quantum Computing [Part 1]

On November 20, 2018, Fujitsu hosted Fujitsu Insight 2018, which took up themes such as "AI and the IoT." In the seminar entitled "Looking into the New Future Opened Up by Digital Annealer—the Latest Trends and Fujitsu's Engagement in Quantum Computing—," the future to be brought about by Digital Annealer was introduced in three sessions: the Latest Trends in Quantum Annealing and Fujitsu's R&D Visions, Expectations for Digital Annealer, and the Evolution and Future of Digital Annealer.
[Fujitsu Insight 2018 "AI and the IoT" Seminar Report]

The Latest Trends in Quantum Annealing and Points to Use It Effectively

Shu Tanaka of Waseda University, the first speaker, introduced the latest trends in quantum annealing and his visions for joint R&D with Fujitsu.

Needs for Quantum Annealing to Increase in the Future IoT Society and Society 5.0

Shu Tanaka
Associate Professor, Green Computing Systems Research Organization, Waseda University
PRESTO Researcher in the field of Quantum state control and functionalization, Japan Science and Technology Agency
Manager, MITOU Target Project, Information-technology Promotion Agency, Japan
Adviser, AI & Robotics Committee, Mobile Computing Promotion Consortium

Quantum computers are attracting much attention. When a new technology emerges, "where to use it" becomes the key focus. For quantum computing as well, many companies are exploring where to apply it.

What we call quantum computing can generally be divided into the gate (quantum circuit) model and the quantum annealing model. In the gate model, calculation methods for prime factorization, data searching, pattern matching, and simulation algorithms have been established logically. By contrast, expectations for quantum annealing relate to executing high-speed, high-accuracy combinatorial optimization.

What can a quantum annealing machine do, and what do we expect from such machines going forward? As a computation technique, quantum annealing is expected to process combinatorial optimization quickly and accurately. Combinatorial optimization refers to finding a preferable choice from among a huge number of choices.

Examples of combinatorial optimization include the traveling salesperson problem and the delivery planning problem, which involve finding the shortest, most efficient route to visit many places. Combinatorial optimization is also used in work shift planning problems for offices with many employees. In work shift planning, it is challenging to find a solution efficiently and to ensure the plan matches each worker's work style.

These problems can be handled easily if the number of elements (e.g., places, people, and objects) is small—for example, if the salesperson needs to visit only a few cities in the traveling salesperson problem, or if the company has only a few employees in the work shift planning problem. But what happens when the number of such elements increases to 100 or 1,000? The number of choices increases, gradually making it difficult to obtain the best answer.

These kind of problems can be found everywhere in business and daily life. It is important to consider how to move beyond the conventional methods, including manual trial and error and listing all the choices; we need to obtain better, more accurate answers quickly. This is where we place our expectations for quantum annealing's power.

Today, combinatorial optimization appears necessary in various situations, such as the Society 5.0 concept proposed in the government of Japan. Society 5.0 is defined as a human-centric society in which cyberspace (virtual spaces) and physical space (real spaces) are fused in a sophisticated way to develop the economy and to resolve social issues.

In Society 5.0, not only visible combinatorial optimization problems but latent problems must be solved to improve society. To this end, development of computation techniques is a more urgent task than ever before.

The Ability to Spot Combinatorial Optimization Problems Is Key to Efficient Use of "Made in Japan" Quantum Annealing

A look at history reveals that quantum annealing originated in Japan. The quantum annealing theory was proposed in 1998 by Dr. Kadowaki and Prof. Nishimori of the Tokyo Institute of Technology. By the 1990s, superconducting electronics were being studied in many parts of Japan.

In 2011, a Canadian venture firm, D-Wave Systems, developed the world's first commercial quantum annealing machine. Next, the bit count, which refers to the index that indicates the size of problem that can be solved, doubled repeatedly in 2013, 2015, and 2017.

The number of research papers on quantum annealing increases every year and has been soaring recently. This shows that quantum annealing is a technology that draws a lot of attention and continues to evolve. When we think about its future growth, it is not too late to get involved in this field; doing so will be a great advantage.

Overview of the history of quantum annealing, which has been rapidly evolving this decade

The simplest way to describe how quantum annealing works is this: if you identify a combinatorial optimization problem in a social issue and then input that problem into a quantum annealing machine, you will get an answer. To make this flow possible, however, you must have knowledge of the domain of application, which is deeply rooted in the relevant business field.

Assume there is aerial video footage of a highway. Without any specialized knowledge on traffic and automobiles, I could come up with combinatorial optimization problems such as searching for delivery routes, gas station launch planning, processing of machine learning by onboard devices, and processing of machine learning by devices embedded in roads, but this is as far as I can get. When someone with expertise in traffic and automobiles watches this video footage, he or she can spot more combinatorial optimization problems that must be addressed immediately. In other words, expert involvement is extremely important.

What I am trying to say here is that when using quantum annealing, it is not knowledge of quantum mechanics or physics that is required. What you need is the ability to spot combinatorial optimization problems. With that ability, you should be able to make ever greater use of quantum annealing.

PoC of computation techniques has begun in many fields, and research is underway on hardware, software, and application search. Even if new hardware is released, it cannot be recognized as practical technology if the software has not been developed. No matter how excellent the hardware, without software to bring out its potential, we can only wonder at the hardware's complexity and how to use it. To address this issue, we are developing software and an easy-to-use Ising machine.

We must also develop a mid- to long-term strategy for realizing annealing technology and transforming society. I am the manager of the MITOU Target Project hosted by IPA to train people who want to develop software for application search using annealing machines as well as other software that facilitates use of annealing machines. In this project, I hope to train future software developers and to help make a wide variety of people interested in annealing.

To further develop these efforts, Waseda University signed a comprehensive collaborative activity agreement with Fujitsu Laboratories Ltd. Waseda University has a large number of experts actively involved in the fields of the humanities, social sciences, and natural sciences. We plan to promote application of Digital Annealer to combinatorial optimization problems in order to resolve real-world security, medical, traffic, logistics, and financial issues.

Visions for Fujitsu Digital Annealer R&D

After Shu Tanaka's speech, Daisuke Iwai of Fujitsu Laboratories Ltd. introduced the present state and vision for Digital Annealer R&D.

Daisuke Iwai
Vice Project Leader
Digital Annealer Project
Fujitsu Laboratories Ltd.

The number of IoT devices that connect various things to the Internet is said to reach 50 billion by 2020, and the volume of data collected from these devices will increase explosively. Such data is unstructured data, not structured data. How should we process it?

Today's silicon semiconductors can do some of it, but naturally there are limits. For this reason, Fujitsu is focusing on domain-oriented computing. While traditional computers are highly versatile and output precise answers via parallel processing, in domain-oriented computing, simple dedicated cores execute massively parallel processing to output imprecise but mostly correct answers. Digital Annealer is included in this latter group.

The number of joint Digital Annealer research and testing case studies has increased. For example, Digital Annealer is used in middle molecule drug discovery. While middle molecules provide the benefit of reducing side effects, searching for them requires an enormous amount of calculations. Digital Annealer can perform such calculations to attempt to find answers quickly.

In medicine, research is underway to apply Digital Annealer to radiation procedures to treat cancer. Ideally, radiation will only hit cancer cells; healthy cells will be protected as much as possible. However, a massive number of calculations is required to obtain irradiation patterns to aim at only cancer cells, which impacts treatment plans. Now, Digital Annealer is being used in a study to develop technology to perform such calculations with high accuracy at high speed.

The key to effectively using Digital Annealer is application development. Currently, we work with research organizations such as universities to develop applications for the future. With respect to the most pressing issues, we want to solve them through co-creation with corporate clients such as yourselves. You have combinatorial optimization problems, and we at Fujitsu cannot solve them on our own, so we hope to work together with you on them.

Speedy Annealing Essential for Keeping Pace with Fast-Changing Society and the World

The next speaker was Koichi Hasegawa of Deloitte Tohmatsu Consulting. He spoke about the latest trends and methods for using cutting-edge technology from a consulting firm perspective.

Expansion of Business Data Use from Descriptive and Diagnostic to Predictive and Prescriptive

Koichi Hasegawa
Partner, Deloitte Tohmatsu Consulting LLC

From the business consultation perspective, we at Deloitte Tohmatsu support business transformation using "disruptive technology" such as quantum computing and AI. We also study business fields to find out how Digital Annealer can be applied to business.

Where can this kind of technology be used? Why should we use it?

The descriptive and diagnostic business data use of the past is now expanding into predictive and prescriptive use. In the 1990s, we used data to understand "what is happening?" In the early 2000s, data was used to analyze "why did it happen?" Data has been expected to predict "what will happen?" since the late 2000s, and to suggest or to prescribe "what should be done?" since 2011.

For example, in the field of public safety, a crime prediction system shows where crimes occur. A few years ago, this system was used to predict where crimes may occur. Now, it is incorporated into the police dispatch system to send out officers on patrols; officers patrol where crimes are likely to occur. The major change here is to incorporate the prediction system into the police operation system to directly influence on-duty actions. In this case, not only prediction but activity optimization is required; this is where the combinatorial optimization takes place.

When the prediction system was used only to predict crimes, police needed data analysts. Once the prediction system has been incorporated into the operation system, however, such analysts are no longer necessary. Officers simply go to where the system shows. Therefore, end users can manage the system themselves.

What is the change we are seeing? The answer is "real-time responses." Instead of using the system to plan daily activities and to issue daily instructions like "please do this and that today," instructions are issued directly in response to changes occurring in the field. I think this is the largest issue we now face; it is a serious problem that computing capabilities cannot keep up with such real-time processing.

Three Factors to Successfully Introduce Cutting-Edge Technology

We have been examining whether this new quantum computing is useful in business while taking note of which fields it can be used in. One feature in which Digital Annealer is far more advanced than other technologies is the speed of commercial application. It is already in use in many domains.

When applying Digital Annealer to business, it is important to search for new tasks that are suitable for the new technology instead of making it do what traditional technology already does and then to consider whether doing so can generate an impact that matches the ROI.

Note that introducing such cutting-edge technology comes with big problems. In the case of AI, for example, Japan is characterized by delayed introduction of cutting-edge technology and low interest in it. These characteristics suggest three major bottlenecks. First, Japanese companies act only after they see results; they need competitor case studies or benchmarks. Second, they do not have a sufficient number of employees who will tackle this kind of challenge, or their employees' technical capacities are low. Finally, they resist change; employees react immediately by asking whether there is a good reason to make any changes.

These bottlenecks seem to come from the fact that Japanese companies are not fully data driven. How to approach and resolve this issue is very important. In addition to introducing new technology, preparing an organization to incorporate and use such new technology as part of operations is a crucial task. After all, I think that companies must see the future while laying the groundwork at the same time. To this end, the development of a small-start methodology is also an important task for companies.

Three initiatives to successfully introduce cutting-edge technology

Part 2 will discuss the latest trends in quantum computing mainly in the automotive industry.


Shu Tanaka
Associate Professor, Green Computing Systems Research Organization, Waseda University
PRESTO Researcher in the field of Quantum state control and functionalization, Japan Science and Technology Agency
Manager, MITOU Target Project, Information-technology Promotion Agency, Japan
Adviser, AI & Robotics Committee, Mobile Computing Promotion Consortium

Daisuke Iwai
Vice Project Leader
Digital Annealer Project
Fujitsu Laboratories Ltd.

Koichi Hasegawa
Partner, Deloitte Tohmatsu Consulting LLC

The presenters' departments and positions are as of the November 2018 lectures.