No.59 Interview with Professor Tshilidzi Marwala
We are living in an artificial intelligence (AI) moment. Around the world, rapid advances in AI technologies have captured the attention of governments, businesses and individuals alike. While there is tremendous anticipation about the benefits AI can bring, concerns remain about ethical issues, risks, and potential adverse impacts on jobs and lives in an AI-augmented world. To discuss the present and future of AI and how it can provide solutions for sustainable development, we were honored to welcome Prof. Tshilidzi Marwala, Rector of the United Nations University (UNU), and a global authority on the theory and application of AI. (Interviewer: Norihiro Suzuki, Chairman of the Board of Hitachi Research Institute)
Table of Contents
Professor Tshilidzi Marwala
Prof. Tshilidzi Marwala is the Rector of the United Nations University and a UN Under-Secretary-General since March 2023, following his leadership roles at the University of Johannesburg. He has collaborated with UN bodies like UNESCO and WHO, holds degrees from the universities of Cambridge, Pretoria, and Case Western Reserve, and is a member of several prestigious scientific academies. Prof. Marwala has authored over 25 books and received honors including South Africa’s Order of Mapungubwe.
For more information about Prof. Marwala’s activities as UNU Rector, see: https://unu.edu/rector
SuzukiSince coming to Japan as Rector of the United Nations University (UNU), is there anything that has left a strong impression on you about Japan or the city of Tokyo? Could you also tell us about the unique role of UNU and why it is located in Tokyo?
MarwalaThe first thing that surprised me the most about Tokyo, and even surprises the people living in Japan, is that it does not feel overcrowded despite the Greater Tokyo area being the largest metropolitan city in the world with about 37 million people.*1 There are almost no traffic problems, and part of the explanation for this is the fantastic public transport system, which in many instances allows faster travel than by using a car. The nationwide transportation system is also definitely the best that I have encountered, and I have lived in many cities around the world.
UNU’s location in Tokyo is especially important considering that relatively few United Nations (UN) agencies are based in Asia. The only other UN agency headquartered in Asia is the UN Economic and Social Commission for Asia and the Pacific (ESCAP), based in Bangkok, Thailand.*2 The presence of UNU in Tokyo is essential for maintaining the UN’s visibility and influence in the region. As the educational and academic wing of the United Nations system, UNU benefits from Japan’s strong academic ecosystem, which includes one of the world’s highest rates of participation in higher education and excellent institutions. Additionally, Japan’s achievements in building livable, organized and clean cities, as well as advancements in sustainable transportation, further reinforce the importance of UNU’s location here.
SuzukiI’m happy to hear you are inspired by the transportation system in Japan, as Hitachi has contributed to the Shinkansen rail network and urban rail infrastructure.*3
MarwalaOne thing I mention when people ask me about the Shinkansen is this: you can place a cup of coffee on a table in the train and it will not spill. This seemingly simple feat relies on vibration control systems that avoid all natural frequencies to eliminate vibrations and spillage.
SuzukiOn the topic of vibration, you have previously mentioned that, as a child, you became interested in the relationship between vibration and structure through your grandmother's pottery making. Could you share how that early experience led you to pursue studies in mechanical engineering and AI?
MarwalaMy grandmother used to weave mats and make clay pots. She was truly a natural engineer and also was my first engineering teacher. Though she never went to school and wouldn't have known the theory of vibration, she understood it intuitively. As an engineer, you need sophisticated optimization programs to select the right materials. She never had any of those, but she would go down to the river to identify the best clay. After forming the pot and hardening it in a furnace, she would cool it slowly and then knock on each pot, listening to the sound. She said, “If the sound is long, it's a good pot. If it’s short, it's a bad pot.” An engineer would explain this using the concept of the damping coefficient, which you learn about when you study vibration. My grandmother never knew about such concepts, but she could still test by ear whether or not the pots were good.
When I pursued my PhD, that was exactly what I studied — structures. We conducted vibration testing by knocking structures and analyzing the sound, using sophisticated instruments such as accelerometers and vibrometers, and inputting data into AI systems. My grandmother never had AI to help her, and that taught me about the power of automation. Human decision-making can vary because people can tire, or their performance can fluctuate. Automating fault detection with AI reduces such risks and is a concrete example of what AI can do for us. Interestingly, in the days of steam engines in the United Kingdom, there was a person whose job it was to go and knock on the train rails, which is similar to what my grandmother was doing with her pots. By knocking, they could tell whether the rail was in good condition or not. Today, AI is combined with sophisticated sensors such as strain gauges, vibrometers and accelerometers, which are vital tools in this field.
SuzukiHitachi is accelerating its transformation into a digital-centric company with digital and AI at its core, creating unique value with the aim of becoming a global leader in the social innovation business. You have often mentioned that with the emergence of generative AI, “we are standing on the brink of another revolution.” *4 How do you assess the current evolution of AI?
MarwalaGenerative AI has been a revolution for many of us. Back in the 1980s people were already engaged in natural language processing using AI systems, but what has changed today is our ability to collect vast amounts of data with better measuring devices, improved data processing technologies, and enhanced data storage capabilities with minimal loss. Computing power has also advanced dramatically. However, at the end of the day, it is important to remember that generative AI cannot create anything truly new; it generates outputs based on existing data used to train its models. If the training data is incomplete or flawed, problems like inaccuracies and hallucinations will arise. Therefore, when we look to the future, we need to consider how the evolution of AI will unfold. I think we will see broader applications of AI. For example, we will be able to give verbal instructions on assembly lines so that processes can be carried out in a better way. Crucially though, AI must become more adaptable as it evolves. Humans have always been able to adjust quickly to new conditions, so self-adaptive AI systems will be important.
Another challenge is reducing computing costs, which will require advancements in semiconductor technology and sustainable management of resources. Currently, semiconductor devices still dissipate large amounts of heat while computing. Manufacturing semiconductors and cooling data centers also consume significant water resources.
Looking to the future, as AI becomes increasingly pervasive, how we govern it will become very important. Governance should not hinder innovation; rather it should promote beneficial uses for AI while eliminating harmful applications.
SuzukiWe continue to evolve Lumada*5 with generative AI, aiming to contribute to sustainable energy and water utilization through our capabilities in energy supply and cooling systems.
Hitachi is applying AI in digital engineering, infrastructure, and manufacturing operations, and we are witnessing how AI is transforming traditional industries. In your view, what kinds of AI use cases hold the most promise for societal implementation going forward?
MarwalaTake healthcare as an example. Life expectancy in Japan is long yet health expenditure per person is lower than the United States, where life expectancy is shorter. That raises the question: how can Japan spend less but achieve better outcomes? This whole idea of bringing down the cost of healthcare is a very important one. AI can play a key role in reducing healthcare costs by improving diagnostics and accuracy. AI is already being applied in medical diagnostics, and can make fewer mistakes than human doctors, which lowers costs and allows more frequent screening without necessarily increasing the cost. AI will also be instrumental in medical devices, many of which are going to be augmented with AI to increase diagnostic accuracy and life expectancy.
Another area is supply chain management, where AI has been in use for some time. We will see increased automation, which will raise productivity but reduce factory employment. The important question here is how to create safety nets for displaced workers. For countries like Japan facing a population decline, automation is a way to boost productivity and economic wealth.
AI’s role in education is also significant. A friend told me that one of the special wards in Tokyo is piloting digital education systems where students interact with AI to support learning, referring to AI if there are certain things that they did not understand well. Though it may sound idealistic, we are going to see much more AI in our education system in future.*6 However, education is more than just knowledge acquisition; so it will be necessary to rethink how we are going to foster cooperation, foster group projects, and foster a human-AI hybrid system in order to enhance productivity.
SuzukiHitachi is boosting productivity by expanding its Lumada business through domain expertise, existing assets, and digital technologies. This includes streamlining software development and advancing AI solutions like HMAX, an all-in-one asset management platform for rail systems powered by NVIDIA AI.*7 HMAX is now being extended to energy and industry sectors including the healthcare area.
SuzukiIt is critical to take a close look at AI governance. In your papers and interviews, you highlight the ethical risks and biases inherent in AI systems.*8 Regarding the challenge of balancing technological innovation with human dignity, what kind of governance frameworks do you believe are necessary to address this “AI balancing problem”?
MarwalaGovernance must begin with understanding what needs to be governed. The first area requiring governance is data. The second is algorithms, including the software and the processes ensuring that the AI performs as intended. The third is computing infrastructure, such as the role of GPUs versus CPUs, and balancing energy demands and efficiencies, etc. The fourth is the AI applications themselves, whose governance will vary by which sector the AI is being deployed in — healthcare, education, manufacturing, etc.
SuzukiHow should these frameworks be implemented?
MarwalaI think the first step is grounding governance in values. As the United Nations, we say our values are derived from the principles of the Universal Declaration of Human Rights. *9 When we apply something, it must protect people, not harm them. The other guiding document is the UN Charter.*10 AI should not be used to support the invasion of another country, for example. That is quite important.
The second step is changing the behavior of those interacting with the AI system — producers, companies and users — primarily through education. One of the things we are doing at UNU is to implement an AI literacy course in Japanese, funded by a grant from Microsoft.
The third step is related to the actual mechanisms. What are the incentives that we have to put into place to ensure that AI is used responsibly? What are the measures that government can take to incentivize companies in Japan and elsewhere to adopt AI technology responsibly and increase productivity?
Next are policies, regulations and standards. Some standards will be global, others local. For example, the late Prime Minister Shinzo Abe first proposed Data Free Flow with Trust (DFFT)*11 in 2019. It promotes the free flow of data while ensuring privacy and security, which is quite instrumental in AI. And then lastly, there is the issue of laws and regulations. What are laws that we need to formulate? The European Union’s AI Act*12 is a leading example. Governance through laws must balance maximizing benefit and minimizing harm without stifling innovation.
SuzukiHitachi published its “Principles guiding the ethical use of AI in Social Innovation Business” in 2021*13, covering planning, implementation, and management of AI in societal and industrial infrastructure. Hitachi provides internal AI ethics education through discussions and lectures, and has established a risk assessment checklist and an AI Ethics Advisory Board with external experts. Development and certification guidelines tailored to AI were also introduced. These principles are essential for Hitachi, especially in balancing governance, human rights, data use, and privacy. Additionally, addressing AI’s energy consumption and advancing carbon neutrality are key to achieving the SDGs.
MarwalaFirstly, I would like to congratulate Hitachi on the comprehensive process that they have established to ensure the ethical use of AI in their business. One aspect I did not mention earlier is the importance of creating governance bodies to oversee AI. I am very glad that Hitachi has an AI Ethics Advisory Board, which is exemplary of responsible AI governance.
SuzukiThe United Nations designed SDGs to tackle global social issues and promote inclusive growth. However, we are now seeing trends such as nationalism and geopolitical conflicts. With the 2030 deadline approaching, there is a view that many SDGs may not be achieved until 2060 or later. How do you see the current state of the world in this context?
MarwalaThe current state of the world is concerning. Rising conflicts, notably in the Middle East and Europe, are troubling. The 2023 SDG Summit issued a disturbing report that basically suggested we are unlikely to achieve the SDGs by 2030.*14 The SDGs were affected by the significant slowdown in economic activity during the time of COVID-19. There also remains much work to be done to ensure that the commitments that were agreed in the Paris Agreement are adhered to.
Speaking from an engineering perspective, we need to think much more vigorously about how technology can be used to advance the SDGs. This includes applying technological advancements in energy, healthcare, agriculture, education, and climate change mitigation.
AI is just one of the technologies for optimization, be it for optimizing energy grids or healthcare systems. We should also consider how AI systems can be used to build better and more cost-effective models, without compromising performance. The responsible application of AI is also important, not forgetting that it has a huge carbon footprint and currently requires vast amounts of water and energy in computing.*15
In summary, I think we need companies like Hitachi to help society deal with issues of sustainable development. Whether it is through the supply of technologies that will be used in food production, energy supply, or control and optimization of our grids, the input and support of companies is very important.
SuzukiMoving on to the next topic, as technology and society become increasingly complex, how do you think the role of human beings will evolve? In an age of human–AI and human–robot collaboration, what kind of mindset or education is needed?
MarwalaWell, I think we're moving into the so-called hybrid model where a human being is always augmented by technology. A manager in a factory, for example, no longer manages just human employees but also AI systems. In this situation, what we need to do is to ask ourselves what kind of educational infrastructure is needed to prepare people so that they are able to understand both humans and machines.
My solution would be multidisciplinary education.*16 Students in science, engineering and technology should also study social sciences and humanities; likewise, those in the social sciences and humanities should also study STEM (Science, Technology, Engineering, Mathematics) subjects.
SuzukiYou have emphasized that leadership today requires not only technological literacy, but also inclusiveness and thoughtfulness. What qualities are essential for top- and mid-level leaders who will drive social innovation in the years to come?
MarwalaI think the first essential quality is that leaders must be data-driven. They need to understand data and recognize that decisions today must ultimately come from observation. They need to invest in the infrastructure to gather the data to be able to do that.
The second aspect is leaders must be able to understand technology and how to use it. Some technologies, like generative AI, have user-friendly interfaces and natural language interaction, so they are not complicated to use. I always say that we need to make sure that the languages that are not in these datasets are not left behind.
The third aspect is that leaders and people must be responsible for how technology is applied, including its limits and financial requirements.
SuzukiIn Hitachi’s new management plan Inspire 2027*17 we aim to contribute to the realization of a “harmonized society where environment, well-being, and economic growth are in balance.” Based on today's discussion, I would like to hear your expectations for Hitachi in the fields of AI and sustainability.
MarwalaOne point I can make is that oversight of AI implementation is crucial, and I commend Hitachi’s AI Ethics Advisory Board for its role. I think in such implementation, the use of more inclusive and ethical AI systems is imperative. The issues of energy efficiency and environmental impact are also very important.
Education and capacity building are also essential. Education is a lifelong journey; there is always something on which we can further educate ourselves. This mindset to keep learning is something that should define every employee of Hitachi.
Being a learning organization means that the organization’s systems, including AI, must also continue to learn, adapting continuously by collecting data and relearning. Rules and regulations, including ethics guidelines, must also evolve, which means regularly revisiting them to incorporate new considerations.
SuzukiWe would also like to hear your expectations for Hitachi Research Institute (HRI) and the researchers.
MarwalaAs a strategic think tank supporting Hitachi’s global growth, HRI is well-positioned to lead impactful research and innovation. I encourage Hitachi and HRI to actively collaborate with universities such as UNU, as well as with institutions both within and outside Japan, to integrate global perspectives.
I also commend Hitachi on its investment in energy and environmental responsibility — these efforts deserve recognition and should be further strengthened, so please build on those commitments. As the Hitachi group develops and implements technological systems, transparency and accountability should underpin technological development with human-centered design at the core of your research, incorporating behavioral science to understand motivation, fear, and how technology can address these human factors. We need more teams than individuals, and I must encourage the culture of working as a team; a diverse and inclusive team is more important now than ever before. I believe HRI will continue to contribute to this area.
Finally, a suggestion. I suggest promoting sabbatical exchanges for researchers with other institutions to broaden innovation and foster fresh ideas.
Engaging with Professor Tshilidzi Marwala reaffirmed my belief that we are at a defining juncture in human progress. The intersection of artificial intelligence and sustainable development presents both a technological imperative and an ethical responsibility — principles that align closely with Hitachi’s social innovation.
Professor Marwala’s perspectives on adaptive AI, the value of interdisciplinary education, and the importance of inclusive governance frameworks offer a strategic blueprint for navigating an AI-enhanced future. His emphasis on human-centric design and hybrid human-machine collaboration resonates strongly with our approach at Hitachi.
As we continue to evolve our Lumada platform and expand our global presence, it is essential to recognize that technology alone cannot address the world’s most pressing challenges. Progress demands collaboration — between industry and academia, between developed and emerging economies, and between humans and intelligent systems — anchored in shared values and a commitment to inclusivity.
Norihiro Suzuki, Chairman, Hitachi Research Institute
We provide you with the latest information on HRI‘s periodicals, such as our journal and economic forecasts, as well as reports, interviews, columns, and other information based on our research activities.
Hitachi Research Institute welcomes questions, consultations, and inquiries related to articles published in the "Hitachi Souken" Journal through our contact form.
