Steptember 15th, 2021
Mikio Nakano, Ph.D.
CEO of C4A Research Institute
Title: Chat to learn: Dialogue systems that can acquire knowledge during chat
In this talk, I will give an overview of the approaches to building dialogue systems that can acquire knowledge during chat-oriented dialogues. Knowledge bases are crucial for intelligent information services but building them requires a lot of effort. Thus, it is desired that enhancing such knowledge bases through dialogues with humans. I first mention previous work on conversational robots for knowledge acquisition and explain why knowledge acquisition during more natural conversations is desired. Then I will talk about two studies on text-based dialogue systems. One is about a text-based interview system that can engage in small talk. It shows that the small talk gives a better impression to the users, increasing the willingness to use the system again. The other is on knowledge acquisition through implicit confirmation processes, that is, without explicitly asking questions to the users. I will also show the user study results that compare the user impression against explicit and implicit confirmation processes. Finally, I will give an outlook on integrating these studies with human virtual-agent interaction.
Mikio Nakano is the Chief Executive Officer at C4A Research Institute, Inc. He received his M.S. degree in Coordinated Sciences and Sc.D. degree in Information Science from the University of Tokyo, respectively, in 1990 and 1998. From 1990 to 2004, he worked for Nippon Telegraph and Telephone Corporation. From 2004 to 2020, he was with Honda Research Institute Japan Co., Ltd. In 2020, he established C4A Research Institute, Inc. He is also a Guest Professor at Osaka University and a Designated Professor at Nagoya University. He has been studying various types of dialogue systems such as conversational robots, spoken dialogue systems, and text-based chatbots.
Steptember 16th, 2021
Youichiro Miyake, Ph.D.
Lead AI Researcher, SQUARE ENIX
Title: Virtual Autonomous Character Agents in Digital Games
Character agent system has been developed and researched in game industry, and also in SQUARE ENIX for 30 years. A character agent is an autonomous, and it sensors the world, makes a decision, and moves its body by itself. Different characters have their own roles and characteristics. Technically, each agent has agent architecture, and uses one or two decision-making algorithm from among seven popular decision making algorithms in digital games such as rule-based, behavior-based, state-based, utility-based, goal-based, task-based, and simulation-based decision-making. In this lecture, some cases in game products, the history, and research results of character agents are shown as movies, and theoretically explained.
A whole AI system in digital game consists of Meta-AI, Character-AI, and Spatial-AI, and they cooperate each other by taking their own roles. Meta-AI controls a whole game situation by giving orders to character agents and controlling all things in game. Character AI is a brain of each agent. An autonomous character agent takes its own role and behaves autonomously, but try to archive the goal ordered by Meta-AI. Spatial-AI supports Meta-AI and Character AI by passing spatial information which it gets by analyzing terrain in game stage. The AI model is called MCS-AI dynamic cooperative model (Meta-AI, Character AI, and Spatial AI dynamic cooperative model). Some cases of MCS-AI dynamic cooperative model n FINAL FANTASY XV are shown in this session.
Youichiro Miyake, the lead AI researcher in SQUARE ENIX, has been developing game titles and researching game AI technologies in these 10 years. He has developed and technically designed AI system for many game titles, and also he teaches students at the University.