Second Digital Observatory ForumProgress on Digital Observatory Technologies through Use of Generative AI and Prospects for Supply Chain Resilience

The forum commenced with a welcoming address from the institute head, Masaru Kitsuregawa, distinguished professor at The University of Tokyo who spoke about the analysis of global supply chains that the institute is undertaking in partnership with government agencies and other Japanese research institutions. In addition to explaining how the institute is seeking to improve the accuracy of its risk assessment and identification through a process of trial and error that includes reviewing which data to observe and making use of generative artificial intelligence (AI), he also had the following to say about the nature and significance of this research.

“Understanding what is happening in supply chains and analyzing their risks is an extremely important task for a manufacturing nation like Japan. However, factors including the inter-disciplinary nature of the work and the need to acquire large amounts of data for analysis mean that it is not something that researchers or universities can undertake on their own. We need the help of others, including Hitachi, as well as the participation of academics from a wide range of the university’s faculties. In my role as director of the institute, I am conscious of how the scale of the work is progressively expanding. In keeping with the university’s principles of diversity and inclusion, we are also analyzing data on people with handicaps to determine what makes an organization inclusive of disability. As each of the research teams will be telling you about their work, I welcome you to hear what they have to say.”

This was followed by a greeting from the forum co-host, Jun Abe, Executive Vice President and Executive Officer at Hitachi, Ltd. He had the following to say about the background to the research being done by The University of Tokyo and Hitachi on generative AI and digital observatory technologies.

“Hitachi has a long-standing involvement in the database and storage businesses going back to the commercialization of the first database created in Japan and we have also partnered with The University of Tokyo on joint research in these fields. In addition to the rollout of the Hitachi iQ solution for AI infrastructure that supports the corporate use of generative AI and involves collaborative creation with NVIDIA, our work on generative AI also includes use of the technology to support frontline workers. The research being undertaken at the Institute for Digital Observatory is an industry-academia partnership between Hitachi and The University of Tokyo that also involves the government. My hope is that this will lead to the creation of an ecosystem for the use of data that will foster AI research, and by doing so will strengthen the industrial competitiveness of Japan.”

Itaru Nishizawa, CTO, and head of the Research & Development Group at Hitachi, Ltd., explained how joint research between The University of Tokyo and Hitachi being undertaken through The University of Tokyo Institute for Digital Observatory is developing techniques aimed at making supply chains more resilient. This work consolidates and analyzes a wide range of data from corporate and other sources to visualize risks and identify warning signs. Recognizing that the lack of visibility into second-tier and lower level suppliers is an especially significant risk in the supply chain, the presentation explained how they have been able to highlight vulnerabilities in the supply network, having developed a technique that uses generative AI to identify the manufacturing sites of these second-tier and lower-level suppliers based on part names and open data.

The presentation concluded by reporting on how Hitachi has provided a test environment for large language models (LLMs) to the Komaba Open Laboratory at The University of Tokyo that is capable of running leading Japanese-language LLMs. CTO Nishizawa also expressed the hope that collaborative creation with The University of Tokyo that includes use of generative AI will lead to improvements in supply chain resilience and the creation of new businesses.

Professor Masashi Toyoda of the Institute of Industrial Science at The University of Tokyo gave an overview of The University of Tokyo Institute for Digital Observatory and described the work being done by its core team. He explained that the goal of the institute is the collection and analysis of data and its translation into value to address risks that are becoming increasingly diverse and severe. The way it has been set up allows for research with participation by eight humanities and science faculties at The University of Tokyo and the Research & Development Group at Hitachi. Activities include investigating ways of providing information on the results of its work in partnership with public institutions and relevant government agencies such as the Ministry of Economy, Trade and Industry and the Ministry of Agriculture, Forestry and Fisheries of Japan.

The three pillars of its research are the investigation, collection, and systematization of information on risks; risk detection, identification, and early warning; and risk assessment. The core team led by Professor Toyoda is working on a platform that will be able to observe a wide variety of social activities. He described its two main systems, a supply chain visualization and analysis system that uses data from the customs clearance records of international industry and a system that uses an LLM to detect and describe anomalous activities found in trade statistics data, and the analysis work that the institute is doing to ensure the reliability supply of agricultural produce.

Professor Yusuke Miyao of the Graduate School of Information Science and Technology at The University of Tokyo, who leads the media analysis team, gave a presentation on the joint research they are doing with the team analyzing the risks for global society led by Professor Takuto Sakamoto of the Graduate School of Arts and Sciences, The University of Tokyo. To provide background to their work, he started by explaining how global supply chains are exposed to diverse risks that include armed conflict and that early risk detection and impact analysis are becoming increasingly important. Professor Sakamoto’s group is responsible for defining and collecting data on armed conflicts and other risk events from a social sciences perspective while Professor Miyao’s group, which specializes in natural language processing, is studying ways of automatically detecting and predicting risk events from news and other text and then automatically generating a description of what their consequences will be. They have been investigating the use of generative AI to improve techniques for the automatic generation of explanations. Professor Sakamoto’s group is working on ways of identifying the warning signs of armed conflict and predicting them in advance. By combining their work, they hope to build a comprehensive system that can automate the detection, prediction, and explanation of risks.

Professor Kazuyori Ito of the Graduate Schools for Law and Politics at The University of Tokyo presented a method for early identification of the risks that government regulations on trade pose to corporate supply chains from the analysis of textual information in government publications and the media. The results of an analysis conducted on a past investigation into the trade restrictions imposed by the USA using Section 232 of the Trade Expansion Act under the first Trump administration found a rise in the occurrence of relevant keywords on news sites before and after the measures were introduced, indicating the potential for pre-emptive detection. The goal is to create mechanisms that companies can use for comprehensive monitoring to better prepare themselves for trade risks, scoring risk based on factors such as which keywords are appearing, their frequency, and the risk level in different phases.

Next, Visiting Associate Professor Kotaro Shiojiri took the stage to talk about early-stage research into using generative AI to identify warning signs. This has involved testing that uses the Gemini generative AI and the “bag-of-words” technique for natural language processing to predict when threats are genuine. This demonstrated potential for further development, with a certain level of predictive accuracy obtained using only small quantities of data. They are also working on measures for improving the accuracy of risk identification using generative AI, including the extraction of words indicating whether threats are genuine from keyword analyses of report documents.

Associate Professor Rie Yamaguchi of the Graduate School of Information Science and Technology at The University of Tokyo, leader of the finance and security analysis team, spoke about the development of a user interface (UI) for understanding shipping and other logistics that her team is working on. As most international freight goes by sea, with marine transportation being especially important for Japan, Associate Professor Yamaguchi’s team has been analyzing shipping movements using real-time data from the Automatic Identification System (AIS) for ships. AIS works by combining a wide variety of information, including Global Positioning System (GPS), customs, and weather data. It has become a useful means of keeping track of trading patterns, including, in recent years, being able to provide an indication of what cargo ships are carrying. Analyses that link AIS data to shipping news have shown the ability to identify patterns in shipping movements and anomalous events. Future plans include work on the automatic collection of news, AIS, and other data, and on algorithms for presenting their consequences visually.

Professor Hiroyoshi Iwata of the Graduate School of Agriculture and Life Science at The University of Tokyo appeared for the food security team. He explained how the three main areas they are working on are the analysis of global supply chains and spatial and genetic analyses of risks to agricultural production, with this work taking place against a background that includes Japan’s low level of food self-sufficiency (producing less than 40% of its own food in calorie basis), climate change, and a changing social environment. On the topic of global supply chain analysis, he highlighted how much Japan relies on food imports from places like China and America by Asian international input output table to show the import-dependence of Japan’s food supply. For the spatial analysis of risks to agricultural production, the team is developing a system for visualizing natural disaster risks and presenting actual disaster damage that works by integrating data collated by agricultural cooperatives into a geographic information system (GIS). Their work on genetic analysis, meanwhile, seeks to develop crop strains that are resilient to risks such as climate change by focusing on data-driven selective breeding and by matching data on damage due to natural disasters with strain characteristics and genetic information.

Professor Taiji Furusawa of the Graduate School of Economics at The University of Tokyo, who leads the global supply chain economics team described the progress they have been making in their work on global supply chain resilience and impacts on the global environment. Their research practices include building models based on general equilibrium analysis and trade theory, the use of world input-output tables (also known as inter-industry relations tables) and other data to determine the relationships between trade costs and volumes, and counterfactual analyses of public policy. Current work involves the analysis of how supply chain vulnerabilities and policy shocks such as the Trump administration tariffs impact each nation’s production, income, and CO₂ emissions. Future plans include the development of bidirectional models that consider the countervailing effects of factors such as the increased frequency of natural disasters due to climate change. Another topic that the team intends to pursue is research into the resilience of urban design and changes in the movement of people in cities during a disaster.

Professor Akiyuki Kawasaki of the Institute for Future Initiatives at The University of Tokyo, who leads the team working on the analysis of data on social activities, described the work they are doing on a foundational framework for elucidating the long-term flow-on effects that climate change and measures for adapting to climate change will have on socioeconomic activity, especially in the Global South. To enable detailed impact assessments at a regional level, this involves collecting data using a bottom-up approach and visualizing the indirect and long-term economic effects of measures for climate change adaption and investment in disaster prevention in each region. Professor Kawasaki explained how this involves the modeling of mechanisms whereby adaptive measures for dealing with flood damage foster a virtuous cycle in regional socioeconomic activity by driving changes in public behavior and risk awareness. He also described how they are developing a framework for using the predictions of the Intergovernmental Panel on Climate Change (IPCC) to quantify the effects of these measures. As the model has attracted interest from other institutions both in Japan and overseas, they are planning further enhancements in readiness for its deployment.

Professor Shinichiro Kumagaya of the Research Center for Advanced Science and Technology at The University of Tokyo, who leads the diversity team, described how past studies of companies, universities, and public agencies have found that organizations where diversity fosters greater productivity and well-being tend to have high levels of psychological safety, and that leadership humility (the level of accuracy of the leader’s self-awareness) is one of the factors behind this. The University of Tokyo has leveraged these findings in its measures for diversity and inclusion and is working on the development of an inclusion monitoring system and support tools that use digital technology, utilizing resources such as surveys of psychological safety at the university and data from interviews with disabled people who work or study there. They are also working with the Ministry of Justice on surveying prison inmates and staff to improve working conditions for staff, prevent inappropriate treatment in prisons, reintegrate inmates into society after their release, and prevent reoffending. Plans include the use of data analysis to propose improvements. The diversity team intends to continue its evidence-based efforts to advance diversity, equity, and inclusion (DEI) in industry, academia, and government.

The program ended with a closing address by Nobuhito Saito, Executive Vice President of The University of Tokyo. After thanking the presenters, panelists, and audience, he went on to express his hopes for how activities such as data analysis can benefit from generative AI, a field that has made dramatic progress over the two years since the institute was established. He summed up by saying, “The institute has developed various analysis tools and systems by leveraging a vigorous exchange of views between industry, government, and academia. I look forward to your support in putting these to use in practical applications.”

A networking session held after the conclusion of the forum provided an opportunity for people to gather around the posters, each of which summarized the research work of one of the teams, and to explore collaboration between companies and institutions in areas ranging from the identification of the challenges and use cases of medium-to-long-term initiatives to their practical deployment and business creation. In addition to presenting research results, the posters also served as prompts for the people attending to share their views. The Institute for Digital Observatory intends to continue its work of using digital observatory technologies to monitor data and use it as a basis for uncovering risks and warning signs.