At Tohoku University's Graduate School of Information Sciences. Clockwise from the front-left: Masamichi Miyama, Masayuki Ohzeki, Keizo Azuma, Kazuya Takemoto (titles omitted)
Masayuki Ohzeki, Associate Professor, Tohoku University
Masamichi Miyama, Special Assistant Professor, Tohoku University
Keizo Azuma, VP of Platform Services, Fujitsu Intelligence Technology Limited
Kazuya Takemoto, Director, Technology Development Project, Digital Annealer Unit, Fujitsu Laboratories Ltd.
Facilitator: Kentaro Fuji, Head of Strategy Planning Division, Global Marketing Unit, Fujitsu Limited
(Continued from Part 1 of the interview.)
Keizo Azuma, the business promotion lead of Fujitsu's Digital Annealer business, and Kazuya Takemoto, a researcher, sat down with Masayuki Ohzeki and Masamichi Miyama of Tohoku University, who are leading figures in quantum computing research. This discussion brought to light the community and knowledge base required to invigorate the quantum computing industry as well as Japan's significant potential therein.
Physics Formulas Serve as Platforms for Problem-Solving
Ohzeki: "Say a company decides that the impressive technology that they built is not being leveraged to the fullest. This is when they first begin to review their system to see where the bottlenecks are and what restrictions are important. Once the objectives of what really must be optimized are clarified, you can convert such goals into formulas and start solving problems. When a company or society faces its issues and collaborates with formularization experts and then completes the formulas, 99% of the problem is actually solved.
Problems are solved through formularization. Common, explicit formulas that serve as platforms for solving problems uniformly are very powerful."
Takemoto: "When formularizing challenges in entirely different categories, it is fascinating when you still return to the same combinatorial problem and find a solution."
Azuma: "I agree. For example, the same formulas apply to the logic of solving problems in financial portfolios (optimization of holding of safe assets and high-risk assets) and the logic for solving rearrangement issues when reviewing shelving locations in a factory. These problems are in completely different fields, so it is surprising."
Hoping to Expand Innovation with the Second-Generation Digital Annealer
Fuji: "In what ways will Digital Annealer develop in the future? Can you describe this from both business and technological perspectives?"
Azuma: "Starting on December 21 last year, we began providing services for the second-generation Digital Annealer within Japan. The number of bits is 8K, with 64-bit precision."
Azuma: "You may wonder what you will do with 64-bit precision (laughs), but in expanding the business, we are thinking of a three-layer system for Digital Annealer.
For the parts closest to the hardware, we are conducting research and development with the aim of achieving a million-bit system. In the middle layer, in order to strengthen the middleware parts, we are collaborating with Canada's 1QBit (1QB Information Technologies Inc.) to solve various challenges.
For the top layer, where we find solutions to our customers' challenges, we focus our efforts on how to identify and formularize such challenges. Our approaches fall under two categories: areas where we identify common points and apply parameters to some degree to repeatedly solve problems, as in the vehicle routing problem, and areas where we dig for very specific points of application, as in drug discovery and radiation therapy for cancer."
Keys to Technological Development: Scale, Collaboration, and Grouping
Takemoto: "We are concentrating on three objectives in our approach to Digital Annealer's technological development. First and foremost is scale. Among combinatorial optimization problems, some are difficult to solve simply because of their massive scale. Large-scale logistics and product recommendations from big data are examples. The second-generation products we produced are 8K-bit, but customers have requested products that can handle an even larger scale. Digital Annealer is digital, so we feel that we can leverage the fact that there is not much noise and expand on the scale.
Second is compatibility with existing computing technologies. Conventional computing is an uninterrupted accumulation of algorithms, so this is something we must leverage. For processes used in drug discovery such as protein folding, Digital Annealer searches for stable structures, and supercomputers are used to calculate the docking for targets. This cycle is repeated over and over. There are surely many potential cases in which we can leverage this compatibility, so we want to draw more of that out.
To pursue this potential, alongside Digital Annealer we must efficiently use various other computation methods. We believe doing so will enable us to significantly expand the range of problems we can solve. Going forward, these are the three approaches that we want to engage in."
New Ways of Using Technology Discovered by Tinkering
Takemoto: "We sense the strong possibility that as these kinds of technologies are used by a greater number of technicians and users, there will be a point at which their prominence will suddenly increase in the manner of a phase transition. We believe that the increase in use cases and actual users are two halves of a whole, but unless we uncover a certain number of use cases, we will not see what the needs are. In fact, it is often only after we identify the needs that use cases emerge. We cannot be lacking in either component.
To achieve this, it is also very important to have more people seriously engage in these efforts on an engineering level. It does not start with Digital Annealer or D-Wave (D-Wave Systems). First, you must extensively prepare environments where these tools can be utilized and then have many people tinker with them. That is when unexpected usage methods emerge and suddenly proliferate. In this sense, I feel that we are now in circumstances with so much potential."
Hoping to Share Expertise through Communities and Knowledge Bases
Fuji: "If we can create a community of people who share their expertise and samples of what problems there are and what formulas are best for solving them, I imagine the knowledge will become increasingly sophisticated."
Takemoto: "You're referring to knowledge bases."
Ohzeki: "We take all writing published related to use cases of D-Wave machines and translate it into Japanese to create what we call the T-Wave knowledge base. We also release our own use cases and verification examples, so we plan to publish articles that describe Digital Annealer use cases. We hope this trend becomes increasingly widespread."
We Cannot Merely Sit and Wait; Now Is the Chance
Miyama: "I agree. When people see collaborative movements like that, others follow suit, and eventually a makeshift community is created. The concern of the moment is the assumption of impending danger to society.
For example, in cases in which a leader manages somehow to operate a business, it is often the case that the leader cannot effectively pass on his or her skills to the next generation. This leads to the danger of powerful technologies being lost due to the inability to pass them down. Moreover, there is the danger that globalization will lead to complete takeovers of the market, so it is important for us to be aggressively proactive. Another problem is the scaling of business conditions.
If we sit idly and wait, it is highly possible that in about five years, solution vendors from overseas, such as the United States or China, will offer solutions so comprehensive that there is no room for anything else. With quantum annealing, we should leverage the fact that the algorithms themselves originated in Japan and make the most of our head-start advantage. Now is the chance. We must move now."
Determined to Dominate the Global Market
Ohzeki: "We also had a moment when we decided to shift gears in our efforts. A user conference is held every six months for international organizations that use D-Wave machines, where use cases for quantum annealing machines are presented. The first time we participated, we felt that we had a strong chance of taking the lead. In Japan, we not only have D-Wave but also Digital Annealer and other methods of solving problems. By applying formularization to many more industries, we think we can lead the world.
Everyone is exploring to find solutions to optimization problems, and problems are being formularized in ways that are based on cultures and industries. This is why we became somewhat frantic in our search for what problem settings are unique to Japan, and what are the unique settings of Japanese companies. We have resolved to help all of Japan rise up; to create a single, large community that includes corporations, universities, and the rest of academia; and to take the lead globally."
Promoting Open Innovation throughout Japan
Fuji: "I really want to push Japan to take the lead and achieve success. To do so, we want to clear away all the small obstacles, and move forward in collaboration with universities, corporations, and venture businesses."
Ohzeki: "This is turning into some kind of nationwide initiative! (Laughs.) Few opportunities like this have emerged until now. But this is our biggest chance, so there is no room to hesitate in collaborating."
Miyama: "We should commit to collaborating in the areas we can, and move forward with open innovation to keep it from becoming a toothless idea."
Fuji: "Do you feel this kind of ambition among venture businesses in Japan?"
Ohzeki: "We are establishing incentives to encourage such ambition. At the risk of engaging in self-flattery, I want to mention that at Tohoku University, we established Jij Inc. as a result of a project supported by START (a program run by the Japan Science and Technology Agency that supports the creation of new industries generated in universities). Professor Miyama came up with the name Jij, which is based on the name of an interaction coefficient between bits in ising model. A graduate student of the Tokyo Institute of Technology is the CEO, who exhibits great and youthful flexibility, capacity, and strength, even if on a small scale.
We hope to see such developments stimulate collaborate among venture businesses that lead to new outcomes and results."
Fuji: "People running venture businesses have remarkable expertise and unique ways of disseminating their messages. We hope to become more light on our own feet and build more networks."
Ohzeki: "We are starting to see more players join the field. I feel that we are entering a very exciting age."
Azuma: "Especially in the past two years or so, I see motivation increasingly greatly in Japan."
Ohzeki: "This trend is rare and unique even on a global scale. It is a unique opportunity. That is, if we succeed before anyone else. (Laughs.)"
Fuji: "To do so, I hope to first kick off this nationwide initiative. (Laughs.) Thank you all so much for participating in this discussion."
Masayuki Ohzeki was born in 1982 in Tokyo. He completed his doctoral course in condensed matter physics at the Department of Physics, Tokyo Institute of Technology in 2008. He assumed his current post in October 2016 after working as an industry-academic-government collaboration researcher at the Tokyo Institute of Technology, a physics researcher at the Sapienza University of Rome, and an assistant professor at the Department of Systems Science, Graduate School of Informatics, Kyoto University. His notable works include "Ryoshi Computer ga Jinko Chino wo Kasoku suru" (Quantum Computing Accelerates AI), co-authored with Hidetoshi Nishimori (Nikkei Business Publications).
Keizo Azuma was born in Osaka in 1963. He joined Fujitsu after graduating from the Faculty of Engineering, Kyoto University. After engaging in server OS R&D, he became Head of AI Services Business Unit, where he supervised Fujitsu's Zinrai, Digital Annealer, and robot-related businesses. In April 2019, he became VP of Fujitsu Intelligence Technology Limited, which was newly established in Vancouver, Canada; in this role, he promotes the planning and development of AI platforms and services.
Masamichi Miyama was born in Hyogo Prefecture in 1981. He completed his doctoral course at the Graduate School of Arts and Sciences, the University of Tokyo in 2011. He assumed his current post in November 2017 after having worked as a special researcher at the Faculty of Engineering, the University of Tokyo, a special researcher at Tohoku University AIMR, and a special researcher at the Graduate School of Arts and Sciences, the University of Tokyo.
Kazuya Takemoto was born in Paris, France in 1973. He joined Fujitsu Laboratories in 2002 after completing his doctoral course at Institute of Physics, University of Tsukuba. After engaging in the development of a long-distance quantum cryptography system using semiconductor quantum dots, he joined the Digital Annealer development team in 2016. He is currently leading R&D as director of the Technology Development Project of Digital Annealer Unit.
Kentaro Fuji was born in Tokyo in 1974. He joined Fujitsu after graduating from the Faculty of Economics, Tohoku University. After engaging in Fujitsu's cloud computing business, he is now in charge of business promotion for Digital Annealer and blockchain as the Head of the Strategy Planning Division in the Global Marketing Unit.