Post-2020 Japan will suffer an overwhelming shortage of manpower. The reason for this is the declining birthrate and aging population, and the accompanying decline in the size of the labor force.
AI has been garnering attention as a technology able to solve this problem, but education is required to understand the features of this cutting-edge technology to implement it as a service that anyone can use and to promote its implementation in society. Much like the chicken-or-egg question, industrial development and education are inseparable. Thus, what is needed to shift this cycle upwards?
Today, we talk about “AI human resources and how to train them” with Deputy Section Chief Mr. Makoto Koizumi, head of AI policy at the Ministry of Economy, Trade and Industry’s Information Economy Division, and Mr. Kumiko Sasaki, chairman and executive director of the startup Groovenauts Ltd., which seeks to ensure that cutting-edge technology can be used easily by anyone.
・The current state of AI training and the key to problem solving is “skills to clarify purpose”
・The ability to respond to change fosters “skills that clarify purpose”
・Can these be generalized as an education method? The two types of education method required
・“Complex knowledge” is one answer for defining basic capability
Deputy Director, Information Economy Division, Commerce and Information Policy Bureau, Ministry of Economy, Trade and Industry.
Over 20 services are responsible for business planning, strategic planning, and execution management in cutting-edge fields such as e-commerce, advertising technology, application development, marketing automation, and cashless payments led by private companies. Following its accession to the Ministry of Economy, Trade and Industry, policy planning and activities are being actively carried out using data and AI utilization perspectives. They are responsible for AI Quest, designed to foster the promotion of MaaS, startup support, and AI human resource development.
Chairman and executive director of Groovenauts Ltd.
She met programming while in 5th grade. Following work as a programmer, systems engineer, and company executive, she established Groovenauts Ltd in 2011. She assumed her current position in 2012 after serving as President and Representative Director. Ms. Sasaki is responsible for the TECH PARK project, an after-school program where participants play with technologies based on their work and parenting experience. In addition, Groovenauts is developing its own enterprise cloud service, MAGELLAN BLOCKS, utilizing AI and quantum computers.
The current state of AI training and the key to problem solving is “skills to clarify purpose”
The first point of concern is the current state of AI training. We asked both of our guests what issues there are, and what was being done to solve them.
Koizumi: First, when talking about the actions of the Ministry of Economy, Trade and Industry (METI) regarding AI training, we have begun a human resources development project called “AI Quest.” In order to advance the implementation of AI in society, we have begun the empirical demonstrations required to rapidly train AI human resources. We have already promoted recurrent education and standardized IT human resource skills, but AI human resource development methods themselves have not really been examined by the Ministry of Economy, Trade and Industry as a project.
I am in charge of policy at the METI for promoting the utilization of data, AI, data science, etc., across industries and providing the support required to produce globally competitive products and services while advancing policies to support the development of AI systems through partnerships between large-scale enterprises and startups.
In Japan, there are about 6,000 companies with more than ¥1 billion in capital, but approximately 3,600,000 SMEs. AI is a tool that can enable new and innovative value, as well as enable improvements in cost efficiency and productivity. It is necessary to promote social implementation through both aspects, and by no means involves only large-scale enterprises and startups.
There is an overwhelming shortage of the human resources needed to solve the issues in various industries and in SME, and there is a high likelihood that we will lose out on the global stage if we do not act within the span of the next two to three years. We have exchanged opinions more than a hundred times with AI stakeholders in industry and academia over the past year, but all are feeling this same sense of crisis. In addition to the various educational reforms that have centered around the Ministry of Education, Culture, Sports, Science, and Technology, the METI is asking “are there not things that should be considered as matters of industrial policy?” How do we develop the human resources needed to implement AI in society within the next two to three years? That is our theme.
The METI has therefore established the AI Quest human resources development project. The biggest issue for this project is how to establish methods of AI human resource development with expanded productivity. There are not enough teachers who can teach tacit knowledge in the workplace. In trying to increase the number, it is difficult to make the engineers active at the cutting edge become teachers. This is a structural issue surrounding practical AI human resources development for social implementation.
Looking back, today we have AI and data science; a little before that, smartphones and adtech; and, before that, the Internet etc. Technology has changed, hasn’t it? And, at any given point in time, because it was uncertain how these would be implemented within society, and how they would become businesses, the people active at the cutting edge had that wisdom. Particularly in the early stages, that wisdom is very rare, and so the wisdom they possessed was extremely valuable. And so, based on that wisdom, consulting, development, and operating contracts and human resource development were established as businesses. Before long, people began to gradually enter the market sensing opportunity, resulting in a period when both wheat and chaff arrived and were then separated, and requirements also settled down. This cycle is also occurring in AI. While this is good in and of itself, the issue is one of time and scale. The approach of increasing the number of people able to carry out teaching roles is unlikely to be done in time.
In AI Quest, rather than increasing the number of teachers, we are taking on the challenge of seeing how expansions in productivity can be achieved where there are few teachers. This draws on 42, a free programming school run by the private sector. Although it is based in France, it has no teachers at all, and all the teaching materials are case studies. There is a staff of 10 people who promote self-study by having students teach one another and creating teaching materials, training 1,000 people a year.
What we will validate here are the methods and materials with high expanded productivity. Teaching materials are created that are intended to be an AI/data science version of MBA case studies. Practical skills are acquired through pseudo-experiential learning using actual examples of implementation. The teaching method is known as “Project Based Learning” (PBL), where students discuss with and teach one another. It is scheduled to begin in October with about 200 people. This demonstration will be open to both success and abject failures, and the results will be shared with the world, together with the teaching materials. Thereafter, it would be good if these were used for internal company training and human resources development projects, and were also areas through which collaborative education could be created based on this movement. In any case, we wish to create a big groundswell of AI human resources development through AI Quest in order to achieve its implementation in society.
What I was thinking about while listening to this story was that, even when simply saying “engineer,” there are various areas of experience. The only thing that is shared by engineers is that they work at the intersection between the social issues of what is required by today’s society and technological trends. And within that is making use of your own fields of expertise to work on solving problems.
Engineers with a substantial understanding of AI are tremendously valuable. This is because, more than programming techniques, a deep knowledge of what you are trying to do is essential. People are able to program AI, but few people can use it. This is quite different from the world where it is desirable to be able to develop programs such as apps and websites. That, I think is the difficulty of AI development.
That’s why the term “AI” is hard to use. AI business, in the first place, has a facet of substituting for human experience, and so deep knowledge of the task to be performed is required before any programming. For this reason, engineers developing AI may need data scientists’ abilities to read and interpret data. Because AI also features technological trends and there are various machine learning methods, it is necessary to fully utilize the technology appropriate for the target while striking a good balance with the knowledge behind it. For this reason, even though we speak generally of “AI education,” I think we must think carefully about what and how this is to be taught.
I agree with Mr. Koizumi’s opinion that we must train the human resources implementation requires in society. What is common to all IT education is the importance of the skill to clarify where to use a technology, and the purpose. In other words, it is very important to understand the intrinsic value of technologies and to understand what problems they solve in society.
If that is the case, this will be applicable regardless of future technological trends. In fact, technologies that change society substantially, such as the quantum computers we are now working on, will continue to arise in succession.
Groovenauts provides technological education to children through its tech park business, and we must always teach them to have a sense of purpose during class. Then, through independent thinking and practice, they will gain an awareness of what is happening in the field of modern IT technology and what can be achieved with AI.
If biased data is supplied when using AI, it will return a biased answer. In order to use it correctly, appropriate data must be prepared. Next thing is that you have to understand what is correct. AI education is also about developing the ability to view things correctly through technology.
Because we have to teach such deep things, if you get onboard with AI education and programming tuition just as a trend and make mistakes in teaching methods, then the children will become exhausted from simply memorizing methods by rote. At the outset, the desire to be taught is an adult motivation. Children want to play. So I think the role of educators is to make full use of that urge to help them learn the ability to see through to the essence of things.
Koizumi: The motivations of individuals cannot be changed. I think the same can also be said for company recruitment, and I think approaches that try to effect change toward a particular type of individuality, saying “This kind of human resource is desirable” is nonsense. Even were you to say that AI human resources should be hired uniformly, what each company thinks of this will, of course, differ. The need for AI human resources and ideas about its development are also distinct. I think that one of the important things in education involves translation, that pressing to “understand” things for which this impulse has not been attained is not communication. People and organizations will not take action when the good points of what you wish to teach have not been extracted, communicated, and translated in such a way as to make them say “I want to do this.”
The ability to respond to change fosters “skills that clarify purpose”
I understand the current situation and the issues stemming from it; however, what is actually required to use AI are sites within companies and project teams. We asked them what they would change in education, and what is required onsite.
Sasaki: When I am asked “what kind of talent do you want?” I always answer “I need people who can rise to the occasion.”
There are always people strong in the face of change in growing companies. Being strong in the face of change can be called “being tolerant of a wide range of values.” In other words, it is the ability to accept diversity. This has been called “diversity” in recent years, but this doesn’t make sense.
In fact, when beginning the project, I thought “might it not be faster to work with children since they are flexible toward change?” In the project study class I lead at the Tech Park, if I say “use slack,” I will be greeted with a meek “OK.” Various IT tools and services are available in the world today, but if you stop to ask whether to continue adopting these new, highly convenient products, the speed of adoption will slow significantly where the targets are companies and organizations. In many cases, their introduction takes time because it is not yet permitted by the company regulations.
In such situations, we judge the situation and can choose the optimal solution. It’s good to be strong in the face of change. By only blocking irregularities and seeking only points to justify rejection without looking for good points, you cannot understand it substantively, and, if the results we obtain don’t suit us, we can always revert.
Koizumi: I see. As an example that appears to be related to this, there are many people in the AI community saying “I was doing something different a few years back.” The reason for this is that areas with high uncertainty face intense change in the first place. Uncertain areas are also increasing with each passing year. In the past, if you changed what you did, this was branded “slash and burn” or “opportunistic,” but today I think it would be better termed “flexible.”
Turning to talk of time axes, there are lifecycle analyses for systems. Everything is born, grows, and declines. What I wish to say is that, in situations with high uncertainty, the withdrawal span is also shorter. Going forwards, when you start a business, it is necessary to devise a plan up to your stepping down. In this new trend, we must immediately change what we have created.
Sasaki: It’s no good for education to be just a stopgap. Education takes time to produce results. The skills required for AI human resources now being cried out for may have withered by the time our children enter society. So you have to go back to its essence, think, and design. “Let’s get by using a patch” won’t work.
Can these be generalized as an education method? The two types of education method required
In the previous section, it was mentioned that “AI education is hard.” As you are aware, the public and private sectors are in the midst of constructing educational programs, so let us try to think about what comes after their implementation. Mr. Sasaki was saying that If there is the skill to clarify the purpose, this can be applied regardless of what trends emerge in the future, but I wonder whether the methods we are constructing now can accommodate the changes that may occur in the future?
Koizumi: It depends, but I think it is highly possible. To summarize what I have said so far, I think there are generally two directions.
One is training people who can handle specific fields. This is education that cultivates knowledge and wisdom that can be used in AI today, and specific fields therein. This is where we should make our efforts in the short term, and it is here that we require productivity growth.
On the other hand, new technologies will continue to arise. What is required at such times is, as Mr. Sasaki said, “education in the skills to clarify what a technology is to be used for, and its purpose.” This is the second direction. This method can be applied even if the field changes, and so its level of priority must be substantively increased.
In general, I think that conventional Japanese teaching methods mostly involve obtaining knowledge centering on classroom learning. Knowledge is formalized. When AI is implemented in society, it is difficult to formalize in and of itself, and so I think pseudo-experiential learning through case studies is important. Those that take this study seriously may find that they obtain an understanding of the skill to clarify purpose.
There is a saying: “We lose sight of a technology when it reaches a certain degree of maturity.” Why do cars and airplanes move? Why do computers work? No one cares anymore. I think it is ideal for children without an awareness of technology to be able to touch upon what it is used for in the manner of a game. This is also ideal for AI. This is generally very difficult, but I think Groovenauts’ efforts are wonderful.
Human beings can demonstrate rather strong intelligence in regards to things we can see. For example, if something flies towards me now, I will instantly try to respond. But, in an unseen world such as that of data or AI, we can’t know what is going on in the moment. But I think that, were we simply able to make the invisible visible, we would understand. In fact, human beings’ intrinsic intelligence is like that, and I think that, by trying to make it visible, we ought to be able to make use of that dormant intelligence.
“Complex knowledge” is one answer for how to define basic capability
Having seen the issues, we are able to understand the capabilities required in the field. Next, let us consider the issues that can arise in providing education. However much we may try to teach, it is rare for everything to be absorbed. Anyone who has worked in a managerial position or has children should be nodding in agreement. How should we tackle this issue?
Sasaki: Everyone looks at staff, work colleagues, and children because everyone has a differing sense of values. Not everyone wants to study or use AI or programming.
As an easy-to-understand example, suppose I give a child a computer. At Tech Park, every child has their own personal computer. In a format where one child does video editing, another uses 3D data using CAD, another does programming, and another does games, the respective methods of use differ. It is not the case that, because everyone has a computer, everyone will do programming. That’s why I would like to be able to support the development of a “what to choose” sense of value. Of course, people form their own worlds within the value perspectives they understand. For that reason in particular, I think how to pass on information is also a part of education.
Whether you are an adult or a child, you won’t want to study where there is no motivation for it that meets your needs. Since learning requires specialization, it is necessary to have to opportunity to prompt them to think “I want to do this.” To that end, I think the question of what kind of environment can be created is important.
One of the activities at Tech Park is preparing an AI programming kit that makes it possible for children to work with AI used by adults. This is not limited to AI and programming but also includes activities using the latest electronic machine tools, computer graphics, digital art, art/music/dance/tradition and history, etc. Professionals from various fields, including engineers from Groovenauts, are invited to teach to the children directly. Children’s learning activities are expanded through a fluid mixing of the specialisms and expertise of various people. Having created such an environment and encouraging the first step of sparking interest, it is then the turn of fundamental knowledge. The question we ask here is “What foundations are truly necessary for life?” This is not limited to AI but applies to all education, but, if you do teach AI, what are the foundations necessary to study it? When we adults were children, even though smartphones and such had yet to become a part of everyday life, was it appropriate to think of such adult values as “foundations,” etc.? “Foundations and education” is a topic we continue to think about, so it remains a topic that has yet to be answered.
Koizumi: “Complex knowledge” might be one answer to the question of “What is needed?” This is an intuitive story, but there are many people active today who possess complex knowledge.
I have a community that I see often, and they gather because all of them are active as some form of domain expert. What they have in common is that they want to learn anything and think about people’s areas of experience as they think of their own. People who behave in this way often say “it’s similar to that, isn’t it?” If you cross borders, you will come up with never-before-seen ideas and see things from new perspectives. I think that this is also linked to “skills that clarify purpose.”
Sasaki: Ah, I see. There is a similar phenomenon in AI. For example, if you open a store in a place where a store has never been opened, you will attempt to predict what the value of sales will be. Selecting factors related to sales and having the AI calculate them makes it possible to calculate such values with surprising accuracy. However, there is a particularly large number of such factors, and so it is not possible to simply determine a relationship between them. If the candidate locations for a store opening differ even slightly, completely different factors will begin to have an effect. And that is not something for which you can simply present a correlation. If you look for this kind of thing, the more you look, the more extremely varied the state of society becomes, I think. It’s perhaps for this reason that people who can absorb diversity are important in this world.
Koizumi: I think people who are diverse and can respond to change are suited to this era of high uncertainty. I think that there are various actions that come first, only to be later followed by meaning and reasons. First of all, in order to act diversely and to create opportunities to discover meaning, you may have to do things that seem like wasted effort. I think that people who are doing well have also wasted much effort.
Sasaki: We don’t know what will lead to work or business discussions. It’s no longer clear from what point it’s work and from what point it’s not work. Going fishing the other day, the person giving the lecture said, “the results of the catch change depending on how many times the line is cast.” I thought, “that’s something that can be said of work too.”
If you drop paint onto water, it will spread out with a splash. As you experience what seems to be uselessness, the world expands one drop at a time. Sometimes you may think “this color and that color don’t match,” or they may become unexpectedly beautiful.
Koizumi: Neither myself nor Mr. Sasaki is able to separate work from play because everything in life is connected through learning. While on the subject, do you think that family should come into situations when thinking about work?
Sasaki: I think family is an element that must not be left out. I regard this as a project that also aligns to peoples’ behavior. Even if you create an IT tool, programming is only a part; there are various other processes, such as defining requirements, preparing and constructing the environment, etc. Within this series of processes, might a situation such as “I will leave early because my child has a fever” arise? However, since it is impossible to interrupt the project, everyone will provide cover, but generally grudgingly. But it can’t be helped, can it? This is because, in life, various elements such as work and family intermix. Work cannot be made the top priority, since all the elements present in life are inherently important. I think that everything in life is connected.
When I manage, I always put buffers in the schedule. Although there is also the aspect that I am a mother, I have a system that seeks to accept every family. To that end, the Tech Park also exists as a place where both staff and their families can coexist. Considering the company itself to be a team, we have put in place a system that allows us to say “you are dedicated to development, and so we will watch over you.”
Koizumi: Returning the conversation to AI human resource development, human resources with complex knowledge can become a goal. Personally, I think that, while we develop human resources able to implement technologies at the forefront of trends in society through AI Quest, we must also substantively develop human resources that can define issues and apply technology.
In this nation, people widely understand the importance of developing human resources for social implementation, and the advantages of actually performing training are gradually becoming known worldwide, and there is a large upswell in the number of people entering into such businesses and studies. It would be ideal if such a flow were to be created. In promoting human resource development, my role is to say “let’s work together rather than doing it alone” and “things will go well if we do this,” and I think this is what we must achieve as a country through AI human resources development. I would like to create movements to promote this, and I wish to extend my ongoing thanks to Ms. Sasaki.
Sasaki: And many thanks to you too.
・If you are not working on societal implementation in practice, you will be unable to teach cutting-edge knowledge.
・If people in industry are placed in education, that industry will stagnate, and that dilemma is an issue.
・The METI is focusing on “how to create expanded productivity in situations without teachers.”
・”The ability to respond to change” is nearly equal to that of “the skill to clarify purpose.”
・Education is similar to translation, and both must convey value and elicit a sense of “I want to do it.”
・What are the basic abilities required? “Complex knowledge” may be one answer.