World not possible to explain by “0 or 1” principle

In the world governed by “Classical Physics” before the introduction of quantum physics, the state of matter takes either “present” or “absent.” The “classical computers” currently used set a certain “threshold” on the electricity flow for the purpose of digitalization by which the matter lower than the threshold is defined as “absent = 0” and the matter higher than the threshold is defined as “present = 1.” Then, combining binary calculations (0 or 1) repeatedly, the computer fulfills high-level calculations at a high speed. Classical computers working on classical physical rules have been improved in performance through further integration and miniaturization of the circuits involved, but this is approaching a limit.

The need for computers that function under the principle of quantum physics governing the natural rules began to be pointed out in the 1980s and theoretical research to this end has been conducted since. In the case of a quantum computer, there is an “superposed” state that cannot be definitely expressed by either 0 or 1. The probability of being in either the state of 0 or 1 is described by the parameter “probability amplitude.” The probability amplitude can assume not only a real number, but also a complex number.

The unit of information on a quantum computer is called “quantum bit (qubit).” During a type of computation which requires “parallel processing” of multiple sets of data, an increase in the number of qubits will enable multiple calculations to be conducted simultaneously through utilization of the characteristic of “superposed” state, allowing an expectation that calculation can be done at a speed much higher than that with classical computers.

Start by self-teaching (Autodidacticism)

When Prof. Fujii entered the School of Engineering, Kyoto University in 2002, he had a desire to “study something fundamental but abstract.” When he encountered quantum physics, he was attracted by its simple rules, which can be understood almost completely by the knowledge of linear algebra learned during the first year of university. At that time, he set a focus on “which games are possible if the rules of quantum physics were accepted.” Soon after that, he became aware of the presence of quantum computers in a scientific journal, where Dr. Michael Nielsen, a famous quantum computer researcher, stated quantum information science is like finding a good move in a chess game whose rule is governed by quantum mechanics. He feeled “I got it!” and thus he dived into this field.

In those days, there was no place for conducting research specifically on quantum computers. He thus belonged to the laboratory of theories on particle physics where members were allowed to study anything having the name “quantum.” He obtained books written by Nielsen that served as textbooks and learned the theories in a self-taught manner.

In the fields where a methodology has been established, and social introduction and acceptance have been advanced, the research topics specialized by individual researchers tend to be highly subdivided, resulting in a situation where what is covered by individual researchers is only a part of the entire field. As far as quantum computing is concerned, on the other hand, this research field is still at the stage of “dawn” at present (2020), and researchers worldwide are competing with each other to stand on the frontline earlier than others. Prof. Fujii is accumulating theoretical research, identifying middle- and long-term tasks, and exploring their solutions, while paying close attention to “almost everything except for hardware.”

Serving as the “witness” for breakthrough

2019 was a historically important year. A paper reporting “quantum computational supremacy” (surpassing the classical computers by a quantum computer in terms of calculation speed) was published in October from Google LLC, USA. It reported that a Google 53-qubit machine was capable of completing a calculation that would take 10,000 years by supercomputer in only 200 seconds.

The two computers competed with each other over the capability of undertaking the task of generating a random bit string with the use of quantum computer circuits or simulating this task with a supercomputer. This was a battle in the “home ground” for quantum. Thus, the victory of the quantum computer over the supercomputer was nothing more than a victory in practically meaningless calculation. IBM, which was a rival for Google, expressed a refutation immediately, saying that such a task could be completed in 2.5 days with a supercomputer. It was, however, of significance that a quantum computer was demonstrated to be superior to the classical computer in a specific field. If quantum computers increase in scale from now on, the difference between quantum computers and classical computers will expand in an exponential manner.

Prior to publication of that paper in Nature, Prof. Fujii was nominated as a reviewer (there are only three reviewers). At that time, Prof. Fujii felt: “This is a phase I have to pass through sooner or later, and such a time has just arrived before me.” That feeling accompanied an emotional uplift: “It pleased me that manuscripts everyone would want to read began to be sent to me earlier than others.”

Is the current status at “vacuum tube” level?

Notwithstanding the arguments above, more than 20 years will probably be required until quantum computers are completed as an all-purpose type computer. The system currently available is called NISQ (Noisy Intermediate-Scale Quantum Computer) which is not designed to remove quantum noise and lacks the function of correcting calculation errors that occur at a certain probability. As the number of qubits increases, the number of errors will also increase. Under the current setting, therefore, it seems difficult to use a large-scale quantum computer with more than 1,000 qubits without error correction.

Errors can be committed also by classical computers, but they can yield reliable calculation results if they are fitted with the function of verifying the output and correcting it as needed. At what stage can we position the current quantum computer development in the context of the history of classical computers? Prof. Fujii points out that the quantum computer currently available is probably comparable to the stage of classical computers prevailing in the 1940s when a giant system was being constructed with an assembly of nearly 20,000 vacuum tubes (although durability is not satisfactory) before semiconductors became widely available for use.

Does this mean that NISQ is “useless”? Is there any way of utilizing the quantum computer even if it involves errors?

The research by Prof. Fujii can play a role. A group of researchers at Osaka University, including Prof. Fujii, published a paper titled “Quantum Circuit Learning (QCL)” in 2018. The AI (artificial intelligence) currently available assumes the form of a mathematical model “neural network” (modeling after brain function) into a classical computer. AI is designed to repeat learning (adjusting the parameters) such that it may give a correct answer from a given set of “supervisory data.” Much more efficient learning will be possible if high-dimensional parameters can be described with the use of a quantum computer in which more complex information can be built in compared with classical computers.

The paper on QCL proposed a methodology for describing parameters with the use of the quantum computer’s AI. Several months after publication of that paper, IBM conducted a follow-up experiment using QCL and made public its results in Nature. During the past approximately 6-year period, the paper on QCL has been referenced by more than 1300 papers.

The proposal of QCL by Prof. Fujii was based on what he experienced approximately 6 years before. In those days, Prof. Kohei Nakajima (currently Specially Appointed Associate Professor at University of Tokyo) was conducting research on robots made of silicon, mimicking the octopus legs, in the attempt of developing soft robots capable of swimming in water. Prof. Fujii was inspired by the attempt of Prof. Nakajima to install the function of learning in the octopus legs themselves instead of using an outside device (serving as the brain) for control of octopus legs, and began his research into utilization of this idea in the field of quantum before NISQ became available.

Concerning the perspectives for the future, Prof. Fujii says: “The services enabling social utilization of classical computers have been established. If a practically applicable quantum computer becomes available, it can advance rapidly through application under the existing framework.”

Resolving the mystery of black holes

Where is the path to be followed by quantum computers? According to Prof. Fujii, “it will not be easy for quantum computers to surpass classical computers. For the time being, the application of quantum computers will focus on the fields where the ‘quantum nature’ can work effectively.”

The chemical world has a high affinity for quantum computers. The molecules and electrons that constitute matter follow the principles of quantum physics. Quantum computers are thus expected to be particularly useful in clarifying the mechanism for photosynthesis (yielding oxygen from light, water and carbon dioxide), developing various catalysts for synthesis of chemicals, etc.

Even the “meaningless bit streams” yielded as an output, discussed in Google’s paper on quantum computational supremacy, may become meaningful if a mechanism for their utilization in fields, such as security reinforcement, is developed. In the field of AI, which has a high affinity for quantum computers, the scope of utilization may be expanded if quantum computers are combined with the quantum-related infrastructure such as quantum communication and quantum sensors.

Google and IBM are competing with each other over the technology to improve the performance of NISQ. The development of general-purpose quantum computers with error correction may be facilitated as an extension of such competition. If a quantum computer compatible with the rules of the natural world is available, it will be possible to perform simulation on the macroscopic world issues (e.g., mystery of black holes) at the submicroscopic level governed by quantum mechanics, possibly allowing further steps towards elucidation of the mysteries.

Setting targets of efforts while looking beyond the near future

The evolution of quantum computers has markedly accelerated during the past approximately 10-year period. A machine called “quantum annealer” specializing in combinational optimization was developed by D-Wave Systems Inc. (Canada). Inspired by this development, Google organized the Quantum Artificial Intelligence Lab in 2013 in partnership with NASA (National Aeronautics and Space Administration), inviting front-line researchers from across the world. Intense competition for development in this field has already begun, involving the USA, European countries and China.

Which perspectives for the future are prevailing in Japan under such circumstances? In Japan, there are researchers at the global front-line of this field, including the team at the Tokyo Institute of Technology who proposed the concept “quantum annealing” in 1998, but Japan is slightly behind the latest research trends such as NISQ development.

Prof. Fujii, however, is not pessimistic about this situation, saying: “Since 2016, Japan has been conducting research along a corrected orbit, focusing on the collection of information about preceding cases and designing systems based on future perspectives, thereby making use of the advantages for late comers.” At Osaka University, the Center for Quantum Information and Quantum Biology was organized as the largest scale institute of this kind, involving more than 70 researchers specializing in “quantum.” Diverse research, ranging from theoretical to practical aspects, has been carried out at this center. Bearing in mind the prospect that an general-purpose quantum computer based on integration of one million to 100 million qubits will become available by 2050, they are considering the topics of research on which emphasis should be laid. Thus, a solid base for research from now on to the next decade is now being established.

The perspectives for the future will become clear if new technologies having the potential of flourishing in the era coming after the near future are identified.

● Professor Keisuke Fujii
Osaka University School/Graduate School of Engineering Science/ Institute for Open and Transdisciplinary Research Initiatives/ Center for Quantum Information and Quantum Biology

• 2011: Completion of doctoral course at Kyoto University Graduate School of Engineering. Ph.D. (Engineering). Specially appointed researcher at Osaka University Graduate School of Engineering Science.
• 2013: Specially appointed teaching assistant at Kyoto University Hakubi Center.
• 2016: Teaching assistant at Photon Science Center of University of Tokyo.
• 2017: Specially appointed associate professor at Kyoto University Graduate School of Science.
• Since April 2019: Current position.
• Majors: quantum information, quantum computing

(Interviewed in March 2020)