AUTORES
Automated Research Project
Purpose
Building an AI that autonomously executes the research cycle.
Policy
Research activities can be divided into the following three stages. AutoRes will build a system that automates each of the phases below.
Problem Formulation Phase
This is the stage of determining what problem to solve. The system will automatically explore valuable problems by understanding the literature.
Research Execution Phase
This is the stage of figuring out how to solve the problem. The system will propose solutions to the problem through automatic theoretical analysis and computational implementation.
Knowledge Sharing Phase
This is the stage of making the obtained results public. The system will widely share the research findings through automatic paper writing and materials preparation.
Features
Research Cycle through Automation
Many existing attempts aim to automate parts of the research phase. We aim to automate all phases of research activities, creating a research cycle by feeding the results generated by the system back into the process. We especially optimize by making the research cycle itself a research subject.
Learning in a Closed Virtual Environment
Many existing attempts incorporate physical robots or experimental plans into the process, which is expected to take a long time for feedback to the model. We perform all learning processes in a virtual space, achieving faster feedback to the system. This is expected to significantly improve the learning speed of the system.
Members / Collaborators
Hideki Aso (AIST)
Ippei Fujisawa (ARAYA)
Yuya Fujisaki (JAIST)
Kazuaki Furumai (KDDI)
Yuki Goto (Keio University)
Shuichiro Haruta (KDDI)
Yuka Hashimoto (NTT)
Ryuichiro Hataya (RIKEN)
Ilya Horiguchi (The University of Tokyo)
Asei Inoue (Proxima Technology)
Yusuke Iwasawa (The University of Tokyo)
Kazumi Kasaura (OMRON SINIC X)
Toshinori Kitamura (The University of Tokyo)
Wataru Kumagai (OMRON SINIC X)
Tadashi Kozuno (OMRON SINIC X)
Kota Matsui (Nagoya University)
Yuma Mizuno (Chiba University)
Shuhei Ogawa (EMOSTA)
Naoto Onda (Freelance)
Kotaro Sakamoto (The University of Tokyo)
Akiyoshi Sannai (Kyoto University)
Sho Sonoda (RIKEN)
Masahiro Suzuki (The University of Tokyo)
Shiro Takagi (Independent Researcher)
Toma Tanaka (BrainPad)
Koichi Taniguchi (Shizuoka University)
Kei Tsukamoto (The University of Tokyo)
Fumiya Uchiyama (The University of Tokyo)
Naoya Umezaki (DWANGO)
Yoshitaka Ushiku (OMRON SINIC X)
Ryutaro Yamauchi (The University of Tokyo)
Hiroshi Yamakawa (The University of Tokyo/WBAI)
Yuki Yoshino (Shimizu Corporation)
Call for Members
We are always looking for individuals to join us in research and development.
We welcome researchers, engineers, students, freelancers, and anyone with an interest.
If you are interested, please contact us at the e-mail address at the bottom of this page.
Let's work together to solve problems related to research automation, not limited to the themes below.
Examples of Ongoing Research Themes
Generating machine learning research topics worthy of submission to top international conferences
Creating feasible research plans considering costs (time, expenses, computational resources)
Developing new machine learning algorithms that are implementable and outperform existing methods
Translating mathematical propositions written in natural language into Lean, either automatically or manually
Generating mathematical theorem proofs through collaboration between proof assistants and large language models
Automatically writing papers that can withstand peer review
Self-improvement of machine learning models executing research