The Pulse
RIKEN starts AI-for-science initiative with shared platform plan
RIKEN has started an initiative to develop scientific AI alongside research that uses it. The institute plans a shared platform connecting supercomputers, quantum technologies and experimental facilities.

RIKEN has started an AI-for-science initiative that will pair the development of scientific AI with research using it, and build an open platform for researchers in Japan and abroad. In a November 9 message, President Makoto Gonokami set out the institute’s plans for the RIKEN AI for X-disciplinary Initiative in Science, known as RIKEN-AXIS.
The institute plans to use insights and findings from research to improve AI technology and refine the platform with researchers internationally. Its approach combines AI development, scientific research, shared infrastructure, international collaboration and training.
From hypotheses to verification
RIKEN intends to develop AI designed for science and foundational technologies for AI agents. The planned work will integrate language and non-language data, mathematical models, simulations and automated laboratories, with the aim of connecting hypothesis generation to verification.
The initiative has five pillars, including scientific AI development and research collaboration between people and AI. The institute also intends to pursue discoveries across its research fields and present a new research model for the AI era.
A shared research platform
RIKEN’s existing resources include researchers across natural science disciplines, the Fugaku supercomputer, scientific data and measurement facilities such as SPring-8 and cryo-electron microscopes. The planned platform will connect Fugaku and the AI-for-science development supercomputer RIKYU with quantum computational and measurement technologies and experimental facilities.
The work is to be coordinated with Japan’s national ARiSE initiative. RIKEN plans to make the platform available domestically and internationally and incorporate findings from research into its development.
Cooperation and training
Another pillar combines AI sovereignty with international collaboration. The institute plans to pursue data security and autonomy under an open-closed strategy; the international cooperation named in the message includes the U.S. Genesis Mission.
A further pillar focuses on creating an environment where researchers, particularly those early in their careers, can use AI to investigate questions across disciplinary boundaries. Gonokami said the institute takes seriously societal concerns about unforeseen risks as AI develops.
Source
Related stories
- Microsoft Open-Sources RetroChimera for Molecule Synthesis Planning
- Endless Frontier releases weights for 180B bioinformatics model
- EUROPA reports faster 1T-scale MoE training on public supercomputer
- AMD-powered Lux supercomputer opens to researchers at Oak Ridge
- JUPITER ranks fifth as AI-focused LISA enters TOP500