News & Events

[School of Interdisciplinary Mathematical Sciences] SANPEI Tomoki and Professor MIYASHITA Homei win the Best Demonstration Award at Entertainment Computing 2023

Sep. 28, 2023

Scene of the demo exhibition 1<br/>
<br/>
<br/>

Scene of the demo exhibition 1


Scene of the demo exhibition 2<br/>
<br/>
<br/>

Scene of the demo exhibition 2


Awards ceremony 1<br/>
<br/>
<br/>

Awards ceremony 1


Awards ceremony 2

Awards ceremony 2
At Entertainment Computing 2023 (EC2023), held by Special Interest Group on Entertainment Computing of the Information Processing Society of Japan at Tokyo University of Technology from August 30 to September 2, 2023, SANPEI Tomoki, fourth-year student and Professor MIYASHITA Homei in Department of Frontier Media Science gave a podium presentation about their research titled "Agent Agora: A system that obtains ideas by collecting and leaving agents of different roles” and won the Best Demonstration Award (an award given to the two most superior demonstrations in the society).
This is an idea generation system with LLM-equipped agents of various roles. All you have to do is to write a theme and gather agents. The brainstorming process will proceed automatically even if you leave it alone.

SANPEI Tomoki, MIYASHITA Homei, Agent Agora: A system that obtains ideas by collecting and leaving agents of different roles, Entertainment Computing 2023 Proceedings, Vol. 2023, pp. 423-425, 2023.

Outline: This paper proposes a system for obtaining ideas by collecting and leaving LLM-equipped agents of various roles. This system models after a square. After collecting the agents in the square, the user writes a theme and leave it. Agents then brainstorm on their own along each role. The user sees and check the situation at the right time, and if there is a good idea, they adopt it. If there is no good ideas, the user kicks out agents which are not contributing to the discussion, brings in new ones which are likely to contribute and leaves them alone again. By repeating management and neglect in this way, it is aimed to obtain high-quality ideas without having to think for ourselves.

Japanese version