Generative AI: Pathways to Democratization, Transparency and Sustainability

Professor for Humanoid Robots, University of Bonn
Website / LinkedIn

Maren Bennewitz is professor for Computer Science at the University of Bonn and head of the Humanoid Robots Lab. She is additionally with the Lamarr Institute for Machine Learning and Artificial Intelligence and has been PI in several national and European projects. She is a member of the executive board and steering committee of the Cluster of Excellence PhenoRob – Robotics and Phenotyping for Sustainable Crop Production and founding member as well as steering committee member of the Center for Robotics, University of Bonn.

Researcher at CNRS
Website

I am currently a senior researcher at the LINK lab (CNRS / NIMS / Saint-Gobain), an international French-Japanese CNRS unit located in Tsukuba. Additionally, I serve as the deputy director of the French research group on Material Science by Artificial Intelligence (GDR “”IA-MAT””).

My research focuses on solid-phase chemistry through thermodynamic modeling and electronic structure calculations, with an emphasis on materials such as hydrogen-absorbing intermetallic compounds. I predict phase equilibrium and investigate crystal lattice vibrational properties for mechanical stability. Utilizing cluster expansion theory, I also analyze disordered solutions. Recently, I am exploring methods as high-throughput DFT and machine learning algorithms to predict innovative materials as exemple for thermoelectricity and energy storage.

Professor in the Image, Data and Signal Department, LTCI, Télécom Paris, IP Paris
Website / LinkedIn

Florence d’Alché-Buc is professor at Télécom Paris in the department Image, Data and Signal that she led during more than 2 years. She has served as program co-chair of NeurIPS 2019 and was the scientific director of Digicosme, the Paris-Saclay excellence center in Computer Science from 2017 to 2019. She has held the Télécom Paris chair on Data Science and AI for Industry from 2018 to 2023 and is a member of the Ellis board. Her researches span various areas of Machine Learning including structured output prediction and frugal and trustworthy AI. Her leitmotiv is to develop mathematically grounded AI to solve key challenges in biochemistry, energy and industry.

Executive Director at DFKI GmbH, Kaiserslautern
Website / LinkedIn

Andreas Dengel is Professor at the Department of Computer Science at RPTU Kaiserslautern-Landau and Managing Director of the German Research Centre for Artificial Intelligence (DFKI) in Kaiserslautern as well as Head of the Smart Data & Knowledge Services research department at DFKI. Since 2009, he has also been a professor (kyakuin) at the Department of Computer Science and Intelligent Systems at Osaka Metropolitan University. He has received many awards for his work and scientific endeavours, including being selected by a jury on behalf of the Federal Ministry of Education and Research (BMBF) in 2019 as one of the most influential scientists in 50 years of AI history in Germany for his research in the field of document analysis. He is the recipient of the Order of Merit of the State of Rhineland-Palatinate and was honoured in 2021 in the name of His Majesty Emperor Naruhito with the oldest Japanese order, the ‘Order of the Rising Sun on the Necklace with Golden Rays’.

Senior Fellow, CEA (Digital Systems and Integrated Circuits Division)
Website / LinkedIn

Dr. Marc Duranton is a member of the Research and Technology Department of CEA, where he is involved in realizations for Artificial Intelligence, Edge to Cloud and Cyber Physical Systems. He previously worked in Philips Semiconductors and NXP where, among others, he led the development of the family of L-Neuro chips, digital processors for artificial neural networks. He is participating in the definition of the SRAs of ECS/KDT and of ETP4HPC, and is in charge of the roadmap activity of HiPEAC on High Performance, Edge And Cloud computing.

Associate Professor, The University of Tokyo
Website

Arisa Ema is an Associate Professor at the University of Tokyo and Visiting Researcher at RIKEN Center for Advanced Intelligence Project in Japan. She is a researcher in Science and Technology Studies, and her primary interest is to investigate the benefits and risks of artificial intelligence. She is also a member of the Japanese government’s AI Strategy Council launched in May 2023. Internationally, she is an expert member of the working group on the Future of Work, GPAI (Global Partnership on AI). She is an elected member of the UN Secretary General’s Advisory Body on Artificial Intelligence.

Professor for Media and Cultural Theory / Vice-President International Affairs and University Development
Website / LinkedIn

Jiré Emine Gözen is a professor of media and cultural theory at the University of Europe for Applied Sciences in Berlin. She has been vice president of international affairs and university development since October 2023 and Chair of the German Society for Media Studies since September 2020.

Her research focuses on media cultural studies, media aesthetics, and visual culture, emphasizing AI, future technologies, posthumanism, and the visual construction of subversive masculinities. Influenced by critical theory, gender and queer studies, and new materialist perspectives, she explores the epistemologies underpinning human-machine relations. Having lived and worked in Japan at institutions such as the Mori Art Museum and 3331 Arts Chiyoda, she brings a global perspective to her research, engaging with marginalized perspectives and their entanglements with hegemonic understandings.

Senior Researcher, NEC

Florence Ho is a senior researcher at NEC since 2023, where she conducts research on optimization methods for solving various industrial problems and applications. Before joining NEC, she was a researcher at AIST (National Institute of Advanced Industrial Science and Technology). Her research interests include multi-agent systems, optimization, metaheuristics, traffic management.

Professor at CentraleSupélec, ParisSaclay University. Head of the MICS Laboratory.
Website / LinkedIn

Céline Hudelot is Professor in Computer Science at CentraleSupélec. She heads the MICS laboratory (Mathematics and IT for Complex Systems). She is interested in both data-driven and knowledge-driven Artificial Intelligence for the semantic interpretation of non-structured data, with a particular interest in explainable AI (XAI). She also works on foundation models for the medical domain and LLMs.

Senior Research Fellow, Center for Science and Thought, University of Bonn
Website / LinkedIn

Dr. Ana Ilievska is Senior Research Fellow on the joint project “Desirable Digitalisation” between the Universities of Cambridge and Bonn, funded by Stiftung Mercator. Before coming to Bonn, she taught at Stanford University and has studied at the University of Chicago, Yale, Lisbon, Tübingen, and in Sicily. Her teaching and research focus on the critical role of the humanities in the age of digital technologies. Dr. Ilievska’s publications include various peer-reviewed articles, public scholarship, podcast appearances, and translations on literature and ChatGPT, the question of critical thinking and AI, and the ethos of Silicon Valley. Most recently, she translated Maurizio Ferraris’s book Webfare: A Manifesto for Digital Well-Being (transcript 2024) and hosted an international conference on Humanism and Artificial Intelligence (May 2024).

Director, Artificial Intelligence Research Center, AIST
Website

Yasuhiro Katagiri received his Ph.D in Information Engineering from the University of Tokyo in 1981. He worked in NTT Basic Research Labs. and ATR Research Labs. He was director of ATR Media Information Science Laboratories. He is currently president at Future University Hakodate. He is a fellow of Japanese Society of Cognitive Science, and president of The Japanese Association of Sociolinguistic Sciences.

AI Policy Advisor, Organisation for Economic Co-operation and Development (OECD)
Website / LinkedIn

Johannes Leon Kirnberger is a policy advisor for AI and sustainability at the AI Unit of the OECD Division of Science, Technology and Innovation (STI). He previously led the program on climate action and biodiversity preservation at the Global Partnership on AI (GPAI) and the International Centre of Expertise in Montreal on AI (CEIMIA). Johannes is a member of the UNEP Expert Group on Digital Tech for Circular Economy, advises the Intergovernmental Panel on Climate Change (IPCC) on the green and digital “twin transition” and serves as guest lecturer at the Technical University of Munich (TUM) on climate change and AI policy. He holds a Bachelor of Science in Management from ESCP Business School, a Master of International Public Management from Sciences Po, and a Master in International Affairs, Energy and Environment from Columbia University.

Senior researcher, Inria
Website / LinkedIn

Jean-Baptiste Mouret is a senior researcher (“directeur de recherche”) at Inria, the national French institute dedicated to computer science and mathematics. His research work intertwines robotics and machine learning to make robots more adaptive. Before joining Inria, he was an assistant professor at Sorbonne Université, where he earned his Ph.D. in computer science in 2008. Mouret was the principal investigator of the ERC-funded project ResiBots (2015-2020), which aimed to develop robots that adapt to damage, and his work was notably featured on the cover of Nature (Cully et al., 2015).

Career Researcher, National Institute of Advanced Industrial Science and Technology (AIST)
Website

Dr. Hiroyuki Sawada is a Career Researcher at Industrial Cyber-Physical Systems Research Center in National Institute of Advanced Industrial Science and Technology (AIST) in Japan. He received B.E. and M.E. from the University of Tokyo in 1987 and 1989 respectively. He received Ph.D. from the University of Strathclyde, Scotland, in 2001. He joined Mechanical Engineering Laboratory (MEL) in 1989. From 1990 to 1992, he left MEL and worked at the Institute for New Generation Computer Technology (ICOT). His research interests include end-user development in manufacturing industry for digital transformation.

Global Head of Industrial AI, Siemens Digital Industries
Website / LinkedIn

Dr. Michael Schrapp is the Global Head of Industrial AI Innovations at Siemens Digital Industries. He holds a PhD in Physics from the Technical University Munich, a Master in Theoretical Physics from the Technical University Munich & Utrecht University, and a Master in Stochastic Engineering in Business and Finance from the University of Applied Sciences Munich. Prior to his current position, he was the Director of Data Analytics & Customer Services at Siemens Healthineers.

Professor for Ethics and Information Technologies, University of Hamburg
Website

Judith Simon is Full Professor for Ethics in Information Technologies at the Universität Hamburg. She is interested in ethical, epistemological and political questions arising in the context of digital technologies, in particular in regards to big data and artificial intelligence. Judith Simon is a member of the German Ethics Council, where she was the spokesperson for the report on the ethics of AI. She currently serves on various other committees of scientific policy advice and has been a member of the Data Ethics Commission of the German Federal Government (2018-2019). She is the editor of the Routledge Handbook of Trust and Philosophy(2020).

NICT Fellow/National Institute of Information and Communications Technology (NICT)
Website

Kentaro Torisawa graduated from the University of Tokyo in 1992. After receiving an MSc in Computer Science from the Graduate School of Science of the University of Tokyo, he became an assistant professor in the university in 1995. In 2001, he was appointed associate professor at the Japan Advanced Institute of Science and Technology. Since 2008, he is working at NICT and has been an NICT fellow since 2020. He has been awarded many awards/grants including the JSPS prize and Twitter Data Grants.

Fellow, Artificial Intelligence Research Center, AIST
Website

Professor Junichi Tsujii is a Japanese computer scientist specializing in natural language processing and text mining, particularly in the field of biology and bioinformatics. He graduated from Kyoto University Graduate School in 1973, obtaining a PhD in engineering. He became Associate Professor at Kyoto University in 1973, Professor at Manchester University in 1988, Professor at Tokyo University Graduate School in 1995, and Principal Researcher for Microsoft Research Asia (Beijing) in 2011 before taking on his current position as Director of AIRC in 2015 and Fellow of AIST in 2017. He still holds a Professorship at Manchester University.

Professor, Institute of Science Tokyo, Supercomputing Research Center
Website

Rio Yokota is a Professor at the Global Scientific Information and Computing Center, Tokyo Institute of Technology. His research interests lie at the intersection of high performance computing, linear algebra, and machine learning. He is the developer numerous libraries for fast multipole methods (ExaFMM), hierarchical low-rank algorithms (Hatrix), and information matrices in deep learning (ASDL) that scale to the full system on the largest supercomputers today. He has been optimizing algorithms on GPUs since 2006, and was part of a team that received the Gordon Bell prize in 2009 using the first GPU supercomputer. Rio is a member of ACM, IEEE, and SIAM.

Postdoctoral Researcher
Website / LinkedIn

Legal studies at Heidelberg University (Germany), first state exam, 2016; Ph.D. in private law, 2020; second state exam (German judicial/bar exam), 2021; studies of economics (University of Hagen, Germany); postdoctoral researcher since 2021 and, starting in summer 2024, leader of a junior research group funded by the German Research Foundation at Heidelberg University (Institute for comparative law, conflict of laws and international business law). Current fields of research include private law, law and technology as well as the interplay of public and private international law as governance instruments.

Prof. Florian BOUDIN
Associate Professor at Nantes University, CNRS research leave at the JFLI / National Institute of Informatics, JFLI, CNRS, NII, Nantes Université

Poster Title:
Automatic instruction generation using LLMs for scientific paragraph revision

Poster Abstract:
Large Language Models (LLMs) are mainly used for instruction following tasks. However, when gathering real-world data for training or fine-tuning those models, we are often able to collect the desired input-output pairs but lack the corresponding instructions necessary to transition from one to the other. Manually annotating these instructions is both costly and time-consuming, and cross-annotation can be required as several instructions could be valid for a single data-point.
This poster explores the potential of automatically generating these instructions using LLMs. We focus on generating instructions for the specific task of revising paragraphs from scientific articles. Our work includes identifying the most suitable metrics for automatic evaluation and assessing the initial performance of three models: Llama-2-13b-chat, Mistral-7B-Instruct, and Llama-3-8B-Instruct on the instruction generation task.

Biography:
Florian Boudin is an Associate Professor (HDR) of Computer Science at Nantes University, where he co-leads the TALN research group at the Laboratory of Digital Sciences of Nantes (LS2N). His research focuses on Natural Language Processing (NLP) and Information Retrieval (IR), particularly on applications involving scholarly documents. He is currently on a research leave at the Japanese-French Laboratory for Informatics (JFLI) in Tokyo.

The work he will be presenting is by his PhD student, Léane Jourdan, and was carried out in collaboration with colleagues from the Nantes University (Nicolas Hernandez and Richard Dufour) and the National Institute of Informatics (Akiko Aizawa).

Dr. Andreas STIEGLER
Creative Technologist, Strichpunkt Design

Poster Title:
Branding and Artificial Intelligence – Practical Applications

Poster Abstract:
In this poster presentation followed by a roundtable discussion on Nov 13, we offer a tour through some real-world applications of Artificial Intelligence in the Branding and Design world. We will venture through different categories of AI used here (generative approaches and beyond) and illustrate them through concrete examples from the actual applications in production. These will be presented in client-cases, such as the Honda Research Institute or DHL. Herein, we will also highlight some of the intercultural differences when deploying Artificial Intelligence in the rather delicate and subjective field of Design. Each presented case will be a practical application showcasing how Artificial Intelligence is already put from the research labs into production today.

Biography:
As Creative Technologist at Strichpunkt, Dr. Andreas Stiegler is engaged in the fusion of know-how from branding and design with research in AI and practical applications such as Game Development. With a PhD in Game-AI, he is dedicated not only to technological integration, but also to creative topics. In addition, Dr. Andreas Stiegler is a lecturer for Artificial Intelligence and Game Development at the Hochschule der Medien in Stuttgart.
Sabrina Moldenhauer is an Art Director at Strichpunkt with a interest in interactive design and characters, for instance as found in Game Development.

Dr. Jun KATO
Senior Researcher, National Institute of Advanced Industrial Science and Technology (AIST)

Poster Title:
Creativity support research “in the wild” for the development of human-centered AI

Poster Abstract:
Research on creativity support tools (CSTs) has long been conducted in the field of human-computer interaction (HCI), but there is growing criticism that applied computer science, including HCI, has typically been conducted and evaluated in a limited cultural context, such as the fast-paced academic community and the WEIRD (Western, educated, industrialized, rich, and developed) part of the world. This poster presents recent efforts in creativity support research “in the wild”, taking a longitudinal and constructive approach with the help of creators outside the research community to build computational tools for creative activities. Examples include TextAlive and its API for creating lyric-driven visual art [ACM CHI 2015, 2023] and Griffith for creating animation storyboards [ACM CHI 2024], both of which contribute to the discussion of how to develop human-centered AI that nurtures the creativity of people with diverse technical and cultural backgrounds.

Biography:
Dr. Jun Kato is a senior researcher at the National Institute of Advanced Industrial Science and Technology (AIST) and the technical advisor at Arch, Inc. He worked for Microsoft and Adobe Research, received his Ph.D. degree from The University of Tokyo in 2014, and has regularly received academic recognition such as Honorable Mentions at ACM CHI (2013, 2015, 2023) and IPSJ/ACM Award (2021). He has researched Human-Computer Interaction (HCI), particularly creativity support for programmers, music video creators, anime directors, DIYers in the maker culture, etc. His work has often gone beyond research papers and been released for public use.

Juliana VRADY / Andrey VRADY
Media Artist, Via Vrady Media Art Lab

Poster Title:
Emotion Blender. Tokyo.

Poster Abstract:
The artwork shows the dialogue between people and technology, how A.I. perceives human beings and how, through the lens of the artists’ computational code, humans and their sentiments might look as art.

Biography:
Juliana and Andrey VRADY are multimedia artists based in Germany. Andrey has been an artistic director of leading print and advertising companies for many years. He is keen on undertaking experiments with photography, making digital collages and finding new aesthetics in the new media. Juliana, his other half, comes from the film industry. The duo has achieved several milestones in interactive works and is still developing further. Their latest two media art installations, running at MOCA Bangkok and as part of Campus Germany (EXPO 2020) in Dubai, were inspired by the dialogue between people and technology.

Dr. Jana BOERGER
Senior Data Scientist, Data Science, Arkestro

Poster Title:
Enhancing Supply Chain Efficiency: Item and Supplier Resolution Using Large Language Models

Poster Abstract:
In supply chain management, item and supplier entity resolution is critical for ensuring seamless operations and optimal decision-making. This study explores the application of Large Language Models (LLMs) in resolving ambiguities and inconsistencies in item and supplier data. Leveraging advanced natural language processing capabilities, LLMs are employed to analyze and match items and suppliers across diverse datasets. Our approach demonstrates improvements in accuracy and efficiency compared to traditional methods. Our evaluation based on real-world datasets from multiple industries, highlights the LLM’s ability to handle complex linguistic variations and contextual nuances. The results indicate that LLM-driven resolution enhances data integrity which will enable more reliable supplier relationships and inventory management. This research underscores the potential of LLMs to revolutionize supply chain data processes, paving the way for more intelligent and automated systems.

Biography:
Dr. Jana Boerger is a Senior Data Scientist at Arkestro. Her research focuses on the application of Artificial Intelligence in Logistics and Supply Chain Management. She is particularly interested in bridging the gap between academia and industry. Her research has been funded by JSPS, the Dr. Friedrich Jungheinrich Foundation, the Material Handling Education Foundation among others. She received her PhD in Machine Learning from the Georgia Institute of Technology. Boerger also holds Master degrees in Computer Science and Industrial Engineering. She is based in San Francisco, California.

David Tsukiyama is a Staff Machine Learning Engineer at Arkestro, a SaaS startup in the Predictive Procurement Orchestration space. Tsukiyama has years of experience in applying Machine Learning in various startups in the United States. He is based in California.

Dr. Kristiina JOKINEN
Senior Researcher, AIST Tokyo Waterfront

Poster Title:
Grounding Words to the Real World: Construction of Shared Context in Generative AI-based Dialogue Models.

Poster Abstract:
Grounding is a collaborative mechanism for establishing mutual knowledge among dialogue participants. It is a necessary skill for language-capable AI-agents, and the recent progress in generative AI has brought the issues into focus. To avoid generation of irrelevant or false information, AI-agents need to ground their utterances into real-world events, and to avoid the statistical parrot effect, they need to construct shared understanding of the context with the partner. Grounding and construction of shared context bridges the gap between words and real-world understanding, thus making AI systems more reliable and effective.
In this poster, I will focus on context-aware AI-agents: how grounding is linked to conversational reasoning and collaboration in Generative AI-based agents, and discuss implications, challenges and opportunities to develop adequately accepted and trusted AI-agents that support friendly interaction and also sustain technical innovation and societal impact in smart environments.

Biography:
Dr. Kristiina Jokinen is a Senior Researcher at AI Research Center AIST Tokyo Waterfront, and Adjunct Professor at University of Helsinki. She is a member of the pan-European AI network of excellence ELLIS, Advisory Board for Japanese AIE (AI in Engineering) Programme, and Steering Committee for International dialogue workshop series IWSDS. Her research concerns human-robot interaction, GenAI-based dialogue modelling and multimodality. She has published widely on these topics and is Editor of the TiiS Journal Interactive Intelligent Systems. She has led numerous national and international research projects and recently led dialogue research in the EU-Japan collaboration project with partners from Germany, France, Italy, and Japan.

Dr. Tamon NAKANO
Research Staff, Institute of AI Safety and Security, DLR – German Aerospace Center

Poster Title:
Hybrid Reservoir Computing – combine knowledge-based information and Reservoir Computing

Poster Abstract:
Reservoir Computing (RC) is a subset of Recurrent Neural Networks (RNNs), excels in handling time-evolutionary, non-linear, and chaotic phenomena in the form of time series, such as stock market. RC comprises a random RNN with fixed weights (referred to as the Reservoir) and a trainable output layer, enabling it to retain the memory of past inputs. The fixed weights significantly enhance training efficiency. In recent years, RC has gained popularity due to its ability to train effectively on small datasets and its low computational cost, making it suitable for edge computing applications. We are developing a hybrid approach to improve RC predictions by integrating knowledge-based information, such as an imperfect physical model, with the reservoir to enhance its predictive capabilities. This combination leverages both data-driven and model-driven methods for more accurate predictions. In this poster, we present our work on this hybrid approach, including preliminary results and future research directions.

Biography:
Tamon Nakano, an AI researcher at DLR (German Aerospace Center), initially specialized in Mechanical Engineering and Computational Fluid Dynamics. Transitioning to AI-related fields lately, particularly AI in mechanical engineering, he now focuses on developing Reservoir Computing for non-linear system predictions. Originally from Japan, with past residences in France, he now lives in Germany since 2022, fluent in JP/FR/DE/EN languages. His current pursuit centers on connecting of these three nations in AI. The three nations have different perspectives and have different visions. He wishes to play the role to make a bridge between the three.

Prof. Dr. Shoya ISHIMARU
Project Proressor, Osaka Metropolitan University

Poster Title:
Integrating Eye-Tracking and Generative AI into an Interactive Digital Textbook

Poster Abstract:
This poster explores the integration of eye-tracking technology and generative AI to create an adaptive digital textbook that dynamically responds to learners’ internal states, such as comprehension and interest levels. By employing eye-tracking methods, we accurately gauge where and how long a learner focuses while reading. In addition, machine learning algorithms analyze these data to infer the learner’s cognitive and affective states. On the basis of these inferences, generative AI customizes the content, offering tailored explanations, examples, and interactive elements to enhance learning experiences. This poster presentation is a part of the outcomes of our trilateral collaborative project between Japan, Germany, and France, summarizing experimental results and future challenges.

Biography:
Shoya Ishimaru is a Project Professor at Osaka Metropolitan University and the CEO of Affectify Inc. He formally worked as a Junior Professor (PI) at the University of Kaiserslautern, a Senior Researcher at the German Research Center for Artificial Intelligence (DFKI), and a Co-Founder CRO at Alphaben. He received his PhD in Engineering at the University of Kaiserslautern, Germany, and the title of MITOU Super Creator from the Ministry of Economy, Trade, and Industry in Japan. His research interest is to invent new technologies augmenting human intellect.

Gauvain ROBERT
Research Assistant, Artificial Intelligence Research Center (AIRC), National Institute of Advanced Industrial Science and Technology (AIST)

Poster Title:
Obstacle avoidance based on diffusion model in narrow indoor environments

Poster Abstract:
We developed a model for navigating narrow indoor environments, enhancing the NoMaD framework, a newly introduced Foundation Model for Navigation with cross-embodiment capabilities using diffusion to generate navigable waypoints. NoMaD struggles in narrow environments, and our hypothesis is that NoMaD’s pretraining implicitly accounts for the robot’s embodiment. By conditioning our model on the robot’s specific embodiment, we can generate trajectories that consider physical constraints, such as a small robot passing through a doorway that would block a larger one. We will show that our approach can reduce collision rates in indoor environments compared to NoMaD. Additionally, we will highlight the benefits of our navigation approach based on generative models over previous deterministic approaches.

Biography:
Gauvain Robert is a research assistant at AIST, focusing on navigation systems, planning, and 3D reconstruction of environments. His current research interests lies in the cross-embodiment capabilities of navigation systems to enhance their efficiency and accuracy in complex, constrained environments.

Yuichiro IWASHITA
Student, Graduate School of Informatics, Osaka Metropolitan University

Poster Title:
Quantitative Knowledge Retrieval from Large Language Models

Poster Abstract:
Large Language Models (LLMs) have been extensively studied for their abilities to generate convincing natural language sequences, however their utility for quantitative information retrieval is less well understood. In this poster we explore the feasibility of LLMs as a mechanism for quantitative knowledge retrieval to aid data analysis tasks such as elicitation of prior distributions for Bayesian models and imputation of missing data. We present a prompt engineering framework, treating an LLM as an interface to a latent space of scientific literature, comparing responses in different contexts and domains against more established approaches. Implications and challenges of using LLMs as “experts” are discussed.

Biography:
Yuichiro Iwashita is a master’s student at the Graduate School of Informatics, Osaka Metropolitan University. He received the B.E. from the College of Engineering, Osaka Prefecture University in 2023. His research focuses on augmenting human intelligence and behavior change. Since 2022, he has also been working on applying AI to solve problems in the medical field. From 2023 to 2024, he is an internship student at the German Research Center for Artificial Intelligence GmbH (DFKI).

Wukai ZHOU
Master student, Toyo University

Poster Title:
Secure Data Acquisition and Privacy-Preserving AI Modeling Platform for Research

Poster Abstract:
In today’s environment where large-scale language models are widely available, the need for data is increasing. However, data required for application in healthcare, education, and other fields may often contain a large amount of personal information. Although the development of artificial intelligence cannot be stopped, techniques to prevent the leakage of private information still need to be developed. Therefore, this study aims to propose solutions for data acquisition processes and AI modeling workflow to help researchers acquire data without compromising the privacy of data providers. The poster introduces a general website platform that will allow researchers to access personal information data and conduct research while ensuring that the data cannot be downloaded and used for purposes other than research.

Biography:
Wukai Zhou completed his Bachelor’s degree in Information Management and Information Systems at Shanghai Business School in China. Currently, he is working on his Master’s degree at Toyo University, researching data annotation techniques to allow AI to read and understand qualitative data. His research interests are in various aspects related to AI technologies, including the relationship between AI and people and research methods to improve existing AI data acquisition and modeling.

Dr. Rohan P. SINGH
Postdoctoral researcher, CNRS-AIST JRL, AIST, Tsukuba, Japan

Poster Title:
Sim2real: Robust Humanoid Walking on Compliant and Uneven Terrain with Deep Reinforcement Learning

Poster Abstract:
For the deployment of legged robots in real-world environments, it is essential to develop robust locomotion control methods for challenging terrains that may exhibit unexpected deformability and irregularity. In this poster, we explore the application of sim2real deep reinforcement learning (RL) for the design of locomotion controllers for large-sized humanoid robots on compliant and uneven terrains. Our key contribution is to show that a simple training curriculum for exposing the RL agent to randomized terrains in simulation can achieve robust walking on the real humanoid robot using only proprioceptive feedback. We train an end-to-end omnidirectional locomotion policy using the proposed approach and show extensive real robot demonstrations on the HRP-5P humanoid over several difficult terrains inside and outside the lab environment.

Additionally, we propose a new control policy to enable modification of the observed clock signal, leading to adaptive gait frequencies depending on the terrain and command velocity. In simulation experiments, we show the effectiveness of this policy specifically for walking over challenging terrains by controlling swing and stance durations.

Biography:
Rohan P. Singh is a postdoctoral researcher at the CNRS-AIST JRL (Joint Robotics Lab). He received his Ph.D. degree from the University of Tsukuba, Japan in April 2024 and his Master’s degree in April 2021. Earlier, he worked as a Robotics Engineer at JRL (then known as the Humanoid Robotics Group) from 2017 to 2019. His current research objective focuses on developing reinforcement learning-based locomotion controllers for humanoid robots.