Webinar: Issues concerning symbiosis, co-creation, and co-evolution of humans and avatars/robots

This event featured discussions on the use of avatar technology, robots, and AI inviting academic researchers working on projects related to such technologies under the JST Moonshot R&D Program.
This event was conducted as a public event with mandatory pre-registration. There were 80 participants from universities and professional firms in Japan, and with simultaneous interpretation, there were 20 participants from the US, UK, German, France, South Korean and other countries. The event proceeded with questions from the participants in Japanese and English.

First, Professor Hideaki Shiroyama, Director of the Institute of Future Initiatives at the University of Tokyo, gave his opening remarks. He mentioned that “the researchers invited to this meeting belong to the Moonshot R&D Program, which aims disruptive innovation while addressing major social issues. However, there are always ELSI issues (ethical, legal, or social challenges) surrounding experimental projects that we need to be aware of.” He further stated it is critical to practice a bottom-up approach to raise voices from the researchers and to have cross-disciplinary, multi-stakeholder discussions to tackle ELSI issues effectively. He closed the remarks by addressing his expectation that this event would become a fruitful opportunity to discuss ELSI issues regarding advanced technologies and their social implications.

For the second part of the event, we had six researchers involved with the JST Moonshot R&D Program providing keynote presentations of their respective projects. Prof. Hiroshi Ishiguro of the Graduate School of Engineering Science, Osaka University was the first presenter. He is the project manager of the ” The Realization of an Avatar-Symbiotic Society where Everyone can Perform Active Roles without Constraint,” which is one of the three projects of Moonshot Goal 1. Prof. Ishiguro has been one of the leading researchers on avatar technology since 1999. Although the popularity of research and development of avatar technology has fluctuated in the past, the covid-19 pandemic has increased the demand for a system that can function remotely. The new era of research has introduced avatars that could be used in a wide variety of ways such as easily manipulating avatars’ appearance. This project aims to achieve a virtualized real world where people are free from the constraints of physical space (similar to metaverse) by the year 2050 and create an avatar-symbiotic society where everyone could actively participate in social life. He envisions a future in which everyone, including the elderly and people with disabilities, can freely participate in a variety of activities, as well as revolutionizing the way we spend our time and the way we live our daily lives. Prof. Ishiguro expressed his enthusiasm for working together with everyone to design a prosperous society that is revolutionized by avatars.

Next, Prof. Kouta Minamizawa of the Graduate School of Media Design, Keio University, gave a presentation on his project. He is working on the second project of the JST Moonshot R&D Program Goal 1 known as “Cybernetic Avatar Technology and Social System Design for Harmonious Co-experience and Collective Ability”. Prof. Minamizawa conducts research on the implementation of Embodied Media: technology involves haptic sense and physical capabilities. For instance, scholarship in this field includes extending individual behavior, increasing individual welfare, and connecting people that are physically distanced. Cybernetic avatar technology is a way to enhance people’s wellbeing and having a cybernetic avatar as another body will allow us to share our actions, experiences, and skills over this digital network in the foreseeable future. He also addressed concerns that the use of avatars also has its aspects that need to be reconsidered from the ELSI perspective. First, we need to evaluate the impact of cognitive augmentation on our communication causing individuals to be able to change their appearance through avatars by their will. Secondly, we need to study the social impact of parallel agency: where one person can work in multiple places at the same time via avatars and robots. Thirdly, we need to assess the emergence of collective ability: multiple people can work in one avatar, potentially resulting in a superhuman avatar that exceeds the limits of a single person’s abilities. The creation of ethically and legally designed “Cybernetic beings” will certainly contribute to diversifying people’s life experiences. Prof. Minamizawa concluded that he is passionate about creating ethically and legally designed cybernetic beings to offer diversified life experiences and deliver an unprecedented way of life.

Dr. Kanako Harada from Graduate School of Medicine/Graduate School of Engineering, University of Tokyo, has delivered the third keynote presentation. She is managing one of the projects for the Moonshot R&D Program Goal 3 “Co-evolution of Human and AI-Robots to Expand Science Frontiers.” The goal of this project is to develop an AI robot system that can think, act, and automatically discover scientific principles and solutions in the field of natural science by 2050. She stated that there is a milestone goal to develop AI robots that will automatically discover scientific principles and solutions to specific problems by 2030. She explained that the core objectives of this project are to enable the creative co-evolution of humans and AI robots to pioneer science and achieve collaboration between AI robot scientists and human scientists for better science. Dr. Harada insists on the necessity to improve AI technology step by step to achieve this big goal through strengthening foundations of AI robotics, AI that automatically explores science, and category theory-driven AI to establish a mathematical foundation. She further states that there are two major barriers to technological innovation. First, AI robots need to perform beyond the limits of scientists’ physical capabilities as well as information processing capabilities. Secondly, AI robots need to attain performance beyond the conventional limits of current AI and robotics technologies. However, there are ethical and legal issues for creating advanced AI and robotics. She believes that people may fear AI robots surpassing human capabilities, and exploring the origin of their fear is critical to resolving the psychological pushback towards the implementation of such advanced technologies. She said that scientists should always be free to explore what they want to do on their own, even though there could be potential debate about whether to allocate human scientists only on tasks where AI scientists underdeliver. Finally, Dr. Harada mentioned that this project aims to make discoveries without the intervention of scientists, though we may need to reconsider whether this is what we demand.

The fourth speaker was Dr. Minao Kukita from the Graduate School of Informatics, Nagoya University. His research specializes in identifying ELSI issues related to “The Realization of an Avatar-Symbiotic Society where Everyone can Perform Active Roles without Constraint” project managed by Prof. Ishiguro. Avatar is a technology that overturns the basic conventional human life condition that an individual owns only one body and can be at a single location. Therefore, the avatar will have a great impact on society, individuals’ lives, and human interaction. While face-to-face communication has tremendous upside in terms of retaining autonomy, mutuality, and multimodality. On the other hand, the number of people you can talk to at one time is limited, and it becomes difficult to talk to people who are physically far away. Dr. Kukita explained that the advantage of avatar technology is that it can reproduce the advantages of face-to-face communication while solving the downside. In the future development of avatar technology, he plans to make it possible to reproduce non-verbal communication and synchronized behavior with the other person that occurs during dialogue. He also pointed out that we should consider the ethical issue of privacy when using avatars, as there is a possibility that naturally exposed non-verbal information such as physiological expressions and gestures may be captured and misused by others. He also raised concerns about the liability and rights of using multiple avatars to act as alter egos and the possibility that digital narcissism and self-enhancement resulting from intentional manipulation and customization of your virtual self could disrupt communication. He expressed that avatar technology has great potential to enrich the world, but that we need to develop a system that can deal with unpredictable impacts on children, conflicts with traditional social practices, and impacts on the economy and industry.

Next, Dr. Arisa Ema of the University of Tokyo Institute for Future Initiatives made a presentation on the “Cybernetic Avatar Technology and Social System Design for Harmonious Co-experience and Collective Ability” project, which is led by Prof. Minamizawa. She claimed that this Moonshot Program aims to attain long-term technological goals in 2030 and 2050 respectively, though the scale of the research deserves to be considered a trans-scientific advancement that requires both scientific and non-scientific contribution to breaking through obstacles. Therefore, she pointed out that the ethical and social aspect of this project needs to be discussed with various stakeholders. She mentioned that it is critical to have a concrete image of the social implementation of the currently developed technology and to share the drive and motivation based on a common vision. She additionally commented that thinking about not only how to create a desirable society, but also with whom to create it from various perspectives to create an ideal vision is important for the success of the program. For example, she addressed that preventing the penetration of socially created conventions and biases to the virtual world is important to create a virtual world where everyone could live without any forms of social constraints. Dr. Ema further claimed concerns about privacy and safety in the avatar-centric virtual world would become a prominent issue, and one of the unique difficulties in the avatar world is the possible ownership issues since advanced skills and valuable experiences can be converted into data for trading. The ability of anyone to acquire advanced skills through data will change the way people are rewarded, incentivized, and motivated for their efforts. In addition, there is a list of potential problems such as monopolization of technology and military use of technology due to the emergence of a large company that connects the real space with the virtual reality space where avatars exist. Therefore, she insisted the international community needs to oversight the development of technologies to share and discuss possible social concerns and implications.

The sixth and final speaker was Dr. Yoshihiro Maruyama from the Research School of Computer Science of Australian National University. He supervises the category theory AI and ELSI aspects of the “Co-evolution of Human and AI-Robots to Expand Science Frontiers” project managed by Dr. Harada. He claimed when we contemplate the human-robot relationship in general, we may think of a relationship where humans mechanically use robots. Although he mentioned the relationship with humans will shift as AI robots could perform more tasks autonomously thanks to the progress in AI robotics. Hence, he pointed out that we need to consider how to reconstruct the relationship between humans and robots to adapt to the evolution of AI robot scientists as well. If robots could explore science efficiently on their own, it will be important for human scientists to redesign and reevaluate the purpose of their research. He introduced two major problematic scenarios that will arise with the emergence of AI robot scientists. First, a very bad AI scientist could emerge and practice dangerous science with perception, mobility, and intelligence that is well beyond human capability. The second question is whether humans will be able to detect the presence of a malicious AI scientist. A truly smart AI might not appear to be evil, but whose research goals also appear to be benevolent to attain human trust. He introduced the challenge of using an abstract mathematical framework called category theory to build a new AI called fusion AI, which combines statistical AI like machine learning with symbolic AI that has been studied for a long time to control the risk of unwanted scenarios.

After the keynote presentations, we moved to a panel discussion having Dr. Kukita and Dr. Maruyama as panelists and Dr. Ema as a moderator. First, Prof. Shiroyama made an opening speech based on the topics presented by the six researchers involved in the Moonshot R&D Program. He claimed that the key concept to consider for ELSI managing issues would be co-existence, co-creation, or co-evolution. He also stated the importance to establish a social relationship between humans and technology upon implementation of new advanced technologies.

The panelists started with a discussion on whether it is possible to remove biases that AI has. First, Dr. Ema posed the question of whether AI can detect and point out natural human biases or biases that researchers include in their hypotheses. In response to this question, Dr. Maruyama answered that it is true that human has biases, but we must acknowledge that such biases contribute to thinking efficiently and discovering new things. Thus, we cannot simply assume eliminating biases is the right thing to do. However, there are certain biases and prejudice that we do not want AI to learn, and the tendency of machine learning to learn based on empirical data makes it difficult to eliminate human-generated biases. As AI could potentially learn implicit biases from our society, it is necessary to incorporate a correction mechanism that is based on top-down ethical principles using symbolic AI. Dr. Kukita introduced the efforts to achieve a moral machine and machine ethics on top of logical mechanical systems by applying category theory.
The theme of the discussion then shifted to the ideal relationship and environment between humans and robots for coexistence and co-evolution from the context of Moonshot R&D Program. Dr. Ema raised the question of whether our society demands a future where everything is mechanized and the connection between humans has decreased. In addition, she addressed whether being able to perform a variety of jobs due to more human activities carried out through robots and living in an avatar-centered form of communication with the cost of a newly transformed busy life will create happiness? Dr. Kukita replied that changes in our lifestyle, the form of labor, and methodologies of scientific research are inevitable as the use of AI becomes more socially widespread. He suggested that we need to think about how to harmonize with current social life and how to mitigate the possibility of technology catastrophically destroying human life and happiness. In terms of the human-AI relationship, Dr. Maruyama pointed out that we tend to embrace data-driven science and utilize AI to make predictions. These habits forced us to drift away from the fun of understanding the causal relationships and dynamics of things. However, scientific research with a division of labor system that relies on machines for prediction and humans for understanding the background structure is lacking in the true essence of scientific research, hence society needs to build a common understanding of the appropriate relationship between AI-based robots and humans.

The relationship between people and robots/avatars is diverse and cannot be considered separately from the way we live and work. The workshop concluded with a reaffirmation of the importance of discussing the social implication of the Moonshot R&D Program and the relationship between people and robots/avatars in our society as we make further efforts to realize the projects.