Address by the President of UTokyo for the 2024 Spring Graduate Matriculation Ceremony [Translated Version]

Congratulations to all of you on your admission! Those of you gathered here today are at the starting point of your life in graduate school at the University of Tokyo. In our graduate programs, learning specialized knowledge is of course crucial. But even more important is acquiring an approach to learning. As we face a society on the brink of major innovations in information technology, such as the proliferation of artificial intelligence, questions arise: How should we learn? How should we think? And how should we act?

At UTokyo, in addition to our undergraduate and graduate education programs, we are also engaged in research and application development on AI through various research organizations, including the International Research Center for Neurointelligence, the Next Generation Artificial Intelligence Science Research Center, the Collaborative Research Organization of Educational Technology, and the Institute for AI and Beyond.

AI is a convenient tool, but its meaning and value can change drastically depending on how it is used. When signing an executive order on AI safety, U.S. President Joseph Biden stated that it both “expands the boundary of human possibility and tests the bounds of human understanding.” The world views the usefulness of AI with both expectation and concern.

One issue of great relevance to our future is what occupations might be replaced by AI.

In 2013, the economist Carl Benedikt Frey and the machine learning researcher Michael Osborne, both of the University of Oxford, published a paper titled “The Future of Employment.” After surveying 702 occupations, they concluded that 47% of U.S. employees were at high risk of losing their jobs to automation within a decade or two. That paper suggested that counsellors would be difficult to replace, but today counseling AI already exists. Over ten thousand types of counseling software have been developed worldwide, and some are even used in the treatment of mental illness. Until recently, we believed that tasks involving emotions, such as counseling, as well as creative activities like art and science, reflect uniquely human qualities and are sacred and irreplaceable. Today, as we confront AI, we need to re-examine that traditional understanding of what humanity is.

I do not mean to say that AI is now as valuable as humans. Rather, we need to ask whether we have understood the value and qualities of human beings only superficially. It is not a question of which is better, humans or AI. What is important is to first acknowledge that, at least for the time being, humans and AI excel in vastly different ways.

Generative AI raises serious issues now, including copyright infringement, operational safety, the spread of misinformation, and the promotion of bias and discrimination. People have various positions regarding AI’s convenience, privacy issues, and potential. What matters now is for each of us to develop a better sense about AI. We need to acquire the judgment and sensibility to make the best decisions without relying on AI alone.

Consider writing as an example. Generative AI can weave together texts that skillfully summarize existing knowledge. However, the outputs of AI are merely samples, neither the best answers nor divine revelations. Even if AI performs better than expected, its excellence is evaluated only from a human perspective. It is humans who consider, decide, and design what kinds of responses AI should give. As its users, we must become smarter and more sensitive about utilizing AI properly. If you want to produce better writing, you need to make what you input more enriching by exposing yourself to a wide range of expressions and works. Without experiencing the concision of easy-to-understand, well-written texts and being moved by their tone and style, and without struggling to put your own inchoate thoughts into words, you will not be able to develop a sense for how to write well.

It is precisely because of the emergence of AI that we must now, more than ever before, think deeply about intelligence and acquire for ourselves the ability to create.

One thing that AI lacks is embodiment, that is, the various qualities an entity acquires by possessing a physical body that interacts with the environment. Embodiment has many components and mechanisms, including the sensations of sight, hearing, smell, taste, and touch, stimulus responses such as perspiration and changes in body temperature, and movement of the limbs.

The body is an interface that connects the brain to the external environment. In humans, a vast range of information is received through all parts of the body and conveyed to the brain. In response to that input, the brain acts on the body by moving muscles and controlling organs. Specifically, the brain receives information from approximately 10 million sensory nerves throughout the body, and it moves the body by controlling 640 muscles. Such complex embodiment cannot be replicated with today’s robotics technology.

AI is essentially computation. As long as its computed data are not fed to actuators for conversion into motion and action, AI cannot act upon the surrounding world. Its ability to perform tasks involving physical actions is thus limited. In healthcare, for example, AI excels at analyzing vast amounts of medical data, but it cannot hold patients’ hands or soothe them through touch as human doctors and nurses can.

But because they lack embodiment, AI and robots also have advantages. Underwater robots, for example, can keep working for dozens of hours in the dark, cold ocean without running out of breath or getting bored by monotonous tasks.

In other words, while humans possess a rich sensibility through embodiment, we cannot escape the limitations of our physical bodies. We live in the real world together with our bodies and their limitations. Because our real world has gravity, air resistance, thermal conduction, and other physical processes, it is fundamentally different from virtual worlds. The unforeseen effects of gravity and friction mean that the control and motion of robots and machines simulated in virtual reality often cannot be accurately reproduced in the real world.

The human body is a truly valuable tool. It is through our bodies that we perceive our surroundings, affect others and the world, and, through such experiences, acquire our intelligence. We need to comprehend our own embodiment correctly. Just as no amount of staring at a musical score will improve your piano playing, you cannot hone your knowledge and skills without actually moving your body and interacting with your environment. Humans can grow only through the process of using our bodies.

The campus experience that each of you begins in graduate school today will be a once-in-a-lifetime opportunity for you to hone your embodiment. Your activities outside of campus will also be a great chance to enhance your physical presence. Of course I hope that you will work hard on your study and research in classrooms and laboratories. But please also immerse yourself in situations off-campus and experience the real world first-hand.

To understand the meaning of embodiment in relationships with others, we also need to remember that humans value emotions. AI cannot understand or empathize with human emotions the way we do.

Earlier, I mentioned the possibility of AI providing counseling, but AI’s lack of embodiment is a significant limitation. In interactions between people, true dialogue can often be established only when both people are together in the same space. Unlike with online communication via the Internet or telephone, sharing time and space creates a sense of solidarity, a connection that occurs precisely because of the constraints on the body. This point is closely related to why the judgments made by AI do not necessarily align with the ethics of human society.

While humans cannot run as fast as bicycles or cars, the sight of athletes challenging their limits in track-and-field events like marathons and 100-meter sprints still captivates human interest. We sense a unique and complex attraction in such races that cannot be reduced to a mere difference in abilities. The same applies to games like shogi and chess, where AI has surpassed humans. Compared to AI, humans may seem inferior with our weaker memories and slower calculations. But people have not stopped playing those games. They stick with those games even though they cannot defeat AI because they enjoy the process of the game itself, something beyond merely winning or losing. No matter how brilliant a move AI makes, it does not enjoy playing the game. Experiences of joy and pleasure, as well as feelings of suffering and anger, are all important opportunities and driving forces for nurturing human intelligence.

Humans and AI are different. It would be far more dangerous for us to behave mechanically and simplistically like AI than it is for AI to behave creatively like humans. It is precisely because AI and humans act differently that we are able to have a complementary relationship with it.

When I was in graduate school, I did research on autonomous underwater vehicles (AUVs). Those robots can continue working alone in the dark, cold ocean for dozens of hours. In contrast, for remotely operated vehicles (ROVs), human operators are necessary. Those operators must keep paying attention for the many hours of operation. With autonomous vehicles, no one needs to supervise the robot constantly, but we do need to trust in its work as we wait. This is an example of the new and different types of relationships being born between humans and AI or robots. A remote-controlled robot moves and performs tasks mostly in a subordinate manner under a human’s control, while an autonomous robot executes tasks itself as an independent system after receiving general instructions from the operator. Because it is autonomous, interaction is necessary. The autonomous robot I created when I was a doctoral student was equipped with a lighted panel that enabled simple communication so that I could check its work and interact with it while I was diving underwater.

One of the books I read at that time was Mind Children by Hans Moravec, a robotics researcher at Carnegie Mellon University. In that book, which was published in 1988, Moravec posed the question of what he would do if a computer could write his book for him or do his research better than he could. He pointed out that many jobs might be threatened. The book also told a story about a highly integrated meld of human intelligence and machines forming a post-biological world somewhere beyond Earth. That idea is similar to the artificial general intelligence (AGI) that is being discussed now. The book also mentions superintelligence, a conglomeration of humanity’s vast and diverse knowledge. At that time, when the Internet had not yet become widespread, I was much inspired by Moravec’s panoramic and versatile imagination, as it encompassed the latest knowledge and cutting-edge technologies across a wide range of fields beyond robotics, including semiconductors, software systems, space exploration, human evolution, and DNA.

You will all live in an era when you will contribute to the world using the power of new technologies—not just artificial intelligence but also other innovative technologies that will emerge. Strive to deepen your understanding of them and of their relations with humans. Regardless of how innovative a technology may seem, it is only a means, not an end. It is up to us to determine whether any means or purpose is truly good. I hope that, as creative global citizens, you will continue to ask what choices are desirable for yourself, for others, and for the future of the Earth and human society.

Your voyage as graduate students begins now. Learning need not be a tough and lonely journey. Your friends and family, our faculty and staff, and many others will be here to support you. The University of Tokyo promises to be your best partner and supporter to help you create the future. I hope that you will stretch your abilities to the fullest and advance with excited anticipation towards a brighter tomorrow. Welcome to the University of Tokyo’s graduate programs, and congratulations on your admission.