The Journal The Authority on Global Business in Japan

The role of technology in our work and personal lives is growing at a rapid pace. It is largely thanks to the rich tools now at the core of business that life has been able to continue during the Covid-19 pandemic.

Powering many of these tools is artificial intelligence (AI), and this was the focus of the first in a new series of online events covering technology, hosted by the American Chamber of Commerce in Japan Kansai Business Programs Committee.

The guest speaker for the June 5 event, presented in Japanese, was Dr. Shotaro Funai, a scientist in the Physics and Biology Unit at the Okinawa Institute of Science and Technology Graduate University. The ACCJ Journal talked to Funai to find out more about his research and how AI can be applied to business, as well as to finding a coronavirus vaccine.

How did you become interested in AI?
Since I was a child, I have been interested in the mechanism of the world. According to the French philosopher Descartes, this world consists of matter and mind, and I wanted to know about both. But when I was a high school student, I realized that research on matter is much more developed than that on mind, and the most fundamental mechanism of matter is studied by physicists. This is why I decided to study physics at university.

When I was a university student, I read the book The Elegant Universe by American theoretical physicist Brian Greene. It is about string theory and suggests that all mysteries of matter might be solved by the existence of higher-dimensional worlds beyond our three-dimensional one.

I was so impressed, and then started my research on string theory in graduate school. After I earned my Ph.D., I continued to research string theory as a postdoc researcher. But I also became interested in AI—especially the neural network—in that it is an attempt to make a simple model of the human brain, which might be related to a mechanism of the mind. I then began reading many articles and papers about the development of machine learning and found that recent discussions of AI use many ways of thinking that are similar to how we approach physics. So, I started researching machine learning from the viewpoint of a physicist.

What is the focus of your AI research?
I focus on how machine learning works. Machine learning is a simple model of the human brain in that its structure—the neural network—is an imitation of the human brain.

These days, there are many types of machine learning, but all extract features of input data and describe them as the long sequences of numbers; that is, high-dimensional vectors.

These sequences are, in general, hard to understand in terms of human languages, but we can interpret some of them in terms of physics. Therefore, using various systems in physics, I train machines and correlate the number sequence of extracted features with various physical quantities.

What is the biggest current challenge in AI?
I think that image recognition has already been sufficiently developed—we have many useful and suitable techniques and can use them for our daily life without any problems. So the current challenge should be linguistic recognition, including translation, summarization, or classification into genres. Recent developments such as BERT, short for bidirectional encoder representations from transformers, seem to enable machines to understand the context of a sentence. BERT is an AI language model used by Google to generate search results.

The next stage may be to guess nuances, common sense, and cultural background from input sentences. I am also curious about this development. Collaborating with linguists, brain scientists, and statisticians, I started to study whether a machine can grasp some emotional aspects of tanka (a Japanese poem of 31 syllables) by comparing the machine results with the reactions of human brains.

What are three ways in which AI can transform business?
Almost all procedures with fixed rules can be done using AI. Companies don’t need to employ people for such procedures. AI can forecast how many products will be sold tomorrow, next week, or next month. This is helpful not only for com­panies but for society as a whole, as we look for solutions to sustainability problems.

AI can also propose the most efficient matching of humans and jobs. Companies could use this proposal as a reference when determining the roles employees might play in a project to make it more successful.

Can data science and machine learning play a role in business?
At least for me, recent AI is almost equal to machine learning. Of course, other types of AI could be developed in the future, but still, I believe, machine learning should remain the main part of AI.

Data science has its own history and now contributes to the development of machine learning—in particular, it provides many methods of data processing applied to machine learning and analyzing how machine learning works. But, from now on, perhaps the two should be developed together. Therefore, as far as I understand, both data science and machine learning play the same roles as AI in business operations.

What role can AI play in developing a Covid-19 vaccine?
In my understanding, the strongest point of AI is to find unex­pected relations and, in terms of medicine, combinations that are effective on humans. AI has already succeeded in proposing ingredients for new, useful materials in several fields of science and medicine. Therefore, to develop a Covid-19 vaccine, we can also expect that AI will propose ways to induce the antibody. I know that several universities and institutes have already begun such research.

What role do you see AI playing in our lives in 20 years?
AI can make our daily lives more convenient and efficient. It always gives us useful information that affects our various decisions. But, I would also like to say that it is my hope that AI will also be a tool for broadening our mental horizon. On the internet, many people share their thoughts and opinions—through writing, images, movies, and so on. Often, this creates conflict with others who hold different views.

AI can learn from the things that are shared, extract features from them, and then describe them as number sequences. This can allow us to easily visualize the relationships between different opinions—which are quite similar or quite different, which things many people agree on and which they don’t, and so on. AI can also forecast how these opinions could change in the near future through our communications and exchange of ideas.

If such a system is widely used, each of us can look over many kinds of opinions objectively and then we may pay more attention to how we can overcome differences of opinion to reach consensus.

That’s just my hope, but I think AI can play such a role soon, and I believe this could be the only way to make our world peaceful.

Christopher Bryan Jones is Editor-in-Chief of The ACCJ Journal. Originally from Birmingham, Alabama, he has lived in Japan since 1997.
AI can forecast how many products will be sold tomorrow, next week, or next month. This is helpful not only for companies but for society as a whole, as we look for solutions to sustainability problems.