DepCoS–RELCOMEX 2024

Frank Coolen "Survival Signature for Reliability Quantification of Large Systems and Networks"

The survival signature is a powerful tool for quantification of the reliability of large systems and networks with relatively few different types of components. It was first presented at the DepCoS-RELCOMEX conference in 2012, and it has led to a large development of theory and applications. In this presentation, an overview of this development is provided, briefly discussing survival signatures for specific scenarios and challenges for practical application to systems and networks with hundreds or thousands of components or nodes.

Frank Coolen completed PhD studies in Mathematical Engineering, specializing in Statistics and Operations Research, at Eindhoven University of Technology, The Netherlands, in 1994. He then moved to the University of Durham (UK), Department of Mathematical Sciences, where he still works as Professor of Statistics. Frank is the founder and main developer of Nonparametric Predictive Inference, a statistical methodology based on few modelling assumptions, enabled by the use of imprecise probability to quantify uncertainty. A further main area of research is Reliability Theory, where he introduced the survival signature, which simplifies quantification of reliability for large systems and networks. Most of his works are jointly with his wife and collaborator Tahani Coolen-Maturi. Frank has supervised 28 PhD students to completion. He is on editorial boards of 10 academic journals and frequently serves on committees for international conferences and panels for international research funding organisations. He has (co-)authored over 300 research papers.

---

Jerzy W. Rozenblit "Human-technology frontier: a retrospective and challenges for the future in the era of artificial intelligence"

As intelligent systems technologies mature, a shift in the modality of use is occurring, namely, computer-based systems are no longer an assistive extension of the human operator but are an inherent part of human endeavors. This, on the one hand, leads to augmented human performance, while on the other hand, presents challenges in management of the boundaries between human vs. “machine” control, over- and under-reliance on automation and artificial intelligence. Social impacts such as dramatic changes in work paradigms, outsourcing, and human interactions are more and more pronounced. This presentation will explore the evolution of smart systems from fundamental models to the most advanced technologies of today. We will explore the human impact of “ultra-high” technologies on engineering, medicine, arts, and the society at large. Then, we will debate the underlying question of “how much better off are we?” living in technology-driven universes and our interactions with the ubiquitous, computer-based world.

Dr. Jerzy Rozenblit is University Distinguished Professor, Raymond J. Oglethorpe Endowed Chair in the Electrical and Computer Engineering (ECE) Department. He holds a joint appointment as Professor in the Department of Surgery, College of Medicine at The University of Arizona. He is also an honorary Professor of the Wroclaw University of Science and Technology. Among various honors, he was recognized with Fulbright Scholar awards for research in Europe. Currently, he is developing computer guided training methods and systems for minimally invasive surgery and serves as Director of the Life-Critical Computing Systems Initiative, a research enterprise intended to improve the reliability and safety of technology in healthcare and life-critical applications. He is a Fellow of the Society for Modeling and Simulation (SCS).

---

Bogdan Denny Czejdo "Limitations of Large Language Models and Measures of Progress"

Large Language Models (LLMs) have become integral to a myriad of applications, offering capabilities to generate text that remarkably resembles human writing. Despite their impressive achievements, these models grapple with a range of limitations, each presenting distinct challenges across different applications. As the use and sophistication of LLMs continue to advance, the demand for assured confidence level is critical. Therefore, refined methodologies to thoroughly evaluate LLM's limitations have become paramount. This presentation introduces a taxonomy designed to systematically categorize the core limitations of current LLMs and to get better insight into their possible predictions in the future. The taxonomy differentiates between limitations stemming from the datasets used for training these models, such as outdated information, inaccuracies, and biases, and those arising from an incomplete understanding of appropriate context and nuances. We will explore the dataset-related limitations addressing the need for real-time information updates, the challenge of ensuring accuracy, and the imperative to overcome bias and enhance sensitivity. The discussion will extend to the context understanding, highlighting challenges in the extraction and use of the models' underlying knowledge, e.g., knowledge graphs. Moreover, this conversation will broaden to encompass the limitations of other generative models that operate on natural language prompts, offering a comparative perspective. The case study of the limitations in the generation of summaries by ChatGPT will be provided by analyzing transformations of the knowledge graph for this summary. In the latter part of this session, we aim to cultivate a structured dialogue based on a meta-knowledge graph, that will leverage the collective experiences of participants and will create a framework for cooperative efforts in the field of Large Language Models.

Dr. Bogdan Denny Czejdo is a Belk Distinguished Professor of Science and Technology in the Department of Mathematics and Computer Science, College of Arts and Sciences, Fayetteville State University, North Carolina, USA. Dr. Czejdo received a Ph.D. degree from the Warsaw University of Technology, Poland. His research areas are databases and visual languages, knowledge-based systems, robotics, geospatial intelligence, and, more recently, machine learning and artificial intelligence. Throughout his academic career, he was also heavily involved in computer science education research. He has more than 200 publications in all these areas. He has been engaged in cooperation with many researchers from various universities and US national laboratories such as NASA, ORNL, and LBNL.

---

Atsushi Ito "ICT for sightseeing"

Since 2014, we have been conducting research and development on ICT for sightseeing in the Nikko area, one of the World Heritage Sites in Japan. After COVID-19, the number of foreign visitors to Japan is rapidly returning to the previous level. In the inbound market, there is a growing need to shift from "mono: products consumption" in which visiting famous shrines and buying Japanese products, to "koto: experience and activity" in which value is placed on the experience gained from the activity or service. One of the reasons for this shift from “mono” to "koto" consumption is the increase in the number of repeat visitors to Japan. In other words, the desire for goods consumption is being satisfied online, while the desire for koto (experiences), which cannot be satisfied without actually visiting Japan, is on the rise. Therefore, it is necessary to provide information such as on "Japanese food" and "souvenirs unique to Japan" that appeals more to the emotions and desires of visitors, rather than the conventional list of tourist attractions and souvenirs. Against this backdrop, the content that should be provided differs depending on the target of tourism support. Therefore, we have been designing, developing, and providing smartphone applications for inbound tourists, Japanese who are not familiar with the area, and tourists who tour the area on foot rather than by car. (1) While applications and related contents provide a wide range of information, this presentation will focus on the technology for providing tourist information in conjunction with BLE beacons. Here, we focus on information that supports the improvement of safety and security and the satisfaction of people's desire for knowledge, and we also provide information on various products. We are also investigating information related to wellness tourism, as typified by forest bathing, and plan to add it to the content in the future. (2) In addition, we are also conducting research and making recommendations on how to express the language of cultural communication with an awareness of both inbound tourists and permanent residents. In addition to accuracy issues such as mistranslations into English, we found in specific descriptions that cultural elements are greatly reflected in the messages from the local community to tourists developed in daily postings. In addition, changes in the number of stores in the area and the number of tourists can be observed through changes in the content of the postings over time, and an attempt is being made to understand local tourism through the postings. (3) In addition, we are attempting to convert local information into compact and attractive text by using an interactive AI for the purpose of communicating the unique food culture of the region, and we will show that the concept of narrative is effective in this context.

Atsushi Ito received B.S. and M.S. degrees from Nagoya University in 1981 and 1983, respectively. He also received a Ph.D. degree from Hiroshima City University in 2007. From 1983 to 2014, he was with the Research and Development Laboratories of KDDI Corporation, and he was a Professor at the Graduate School of Engineering of Utsunomiya University from 2014 to 2020. He has been a Professor at the Faculty of Economics of Chuo University since 2020. During 1991-1992, he was a visiting scholar at the Center for the Study of Language and Information (CSLI) at Stanford University. His current research interests include ad hoc networking, bio-signal applications, and ICT-based agriculture.