The relationship between AI and science now stands at a historic turning point. Conversational AI (generative AI), which is widely used today for the gathering of information and generation of text, has moved beyond the stage of simply being able to generate responses based on learned data (inference) and has steadily improved its ability to engage in “logical reasoning”—that is, to think deeply and systematically on its own.
Generative AI is moving beyond being a mere tool that assists humans through conversation and is evolving into “agentic AI”—a partner that can make autonomous decisions and take action. Furthermore, a discontinuous evolution—one that could be described as a tectonic shift in scientific research—is underway, as multiple agentic AIs begin to form specialized teams within computers. Starting in late 2025 and moving through 2026, we are witnessing these changes unfold at a pace that has no precedent in human history. Individual agents are taking on diverse specialized roles, autonomously carrying out complex research tasks while discussing and issuing instructions to one another. This leap to a new dimension, or paradigm shift that goes far beyond just an extension of what we have known so far, is becoming reality at a dizzying pace.
While this tectonic shift has enormous potential for human society, it is also giving rise to vague anxieties about the future and concerns about unforeseen risks. We at RIKEN, standing at the forefront of science, take these societal concerns seriously and face this era with a keen sense of the gravity of this moment.
At the same time, we believe we must not be intimidated by these changes or shy away from them, as this would stall our exploration of the unknown. Taking on the unknown, acquiring new knowledge, and broadening our horizons are fundamental joys for humanity and a driving force that has sustained human growth and development. This essence remains unchanged even in the age of AI.
AI has now reached an advanced stage where it can integrate and understand not only linguistic data but also data on natural phenomena captured by experimental instruments—including non-linguistic data such as scientific measurement data—as well as complex mathematical models and simulations. The rapid advance of agentic AI—allowing the formation of teams within computers—has made it possible to autonomously coordinate processes ranging from the generation of hypotheses to complex calculations, experimental design, and verification. This has paved the way for humans and AI to collaborate to dramatically expand the space of scientific exploration. Precisely for this reason, we humans must also refine the wisdom needed to make sophisticated use of this power as AI continues to evolve.

At this turning point, RIKEN is launching the RIKEN AI for X-disciplinary Initiative in Science (RIKEN-AXIS) to advance co-creation in AI for Science.
RIKEN brings together researchers from a wide range of natural science fields, advanced measurement facilities such as SPring-8 and cryo-electron microscopes, the supercomputer Fugaku, and a wealth of scientific data. Using our combined strengths to make full use of these resources, we will advance the development of new scientific AI in close integration with the cutting-edge scientific research that utilizes it. We will build an open scientific research platform by feeding insights and findings from the frontlines of research back into the improvement of AI technology, continuously refining and expanding the platform in collaboration with researchers both in Japan and abroad, and ensuring its rapid and flexible evolution.
During my tenure as President of the University of Tokyo, I coined the term Kyuchi-kyosou (究知協創) through careful deliberation with my colleagues. It embodies the philosophy of diverse individuals joining forces to advance the never-ending pursuit of knowledge in the correct way. In this initiative, we add further meaning to Kyosou (co-creation): active collaboration between humans and AI, as well as the convergence of knowledge across national borders. The guiding phrase “Empowering the Joy and Art of Discovery,” which we will champion under this new initiative, embodies our desire to use the power of AI to further enrich both the fundamental joy of understanding and discovering the unknown, and the art of intellectual creation—the ability to pose essential questions, discern the principles hidden deep within nature, and link these insights to original discoveries.
This initiative will foster a synergistic and iterative cycle among the advancement of scientific foundations and technologies of AI for Science, the acceleration of research through co-creation between AI and humans, and the construction of a shared platform that can rapidly and flexibly evolve. Through international collaboration and the nurturing of talent, we will create a cross-cutting enabling foundation that supports all three activities. To this end, we will pursue the following five pillars as an integrated whole.

• 1.Advancing the scientific foundations and technologies of AI for Science
  We will research and develop AI designed specifically for science and foundational technologies for AI agents, integrating linguistic and non-linguistic data, mathematical models, simulations, and automated laboratories to seamlessly link hypothesis generation with verification.
• 2.Accelerating cutting-edge scientific research through human-AI co-creation
  We will tackle the challenge of making innovative scientific discoveries across RIKEN’s diverse research fields and present to the world a new research model for the AI era.
• 3.Building and providing an agile, evolving shared platform
  In coordination with Japan’s national ARiSE initiative, we will link the AI for Science development supercomputer RIKYU, the supercomputer Fugaku, quantum computational and measurement technologies, and cutting-edge experimental facilities to provide a foundation—both domestically and internationally—that can rapidly incorporate findings from the frontlines of research and continuously evolve.
• 4.Achieving both AI sovereignty and international collaboration
  We will ensure data security and autonomy and, under an “open-closed” strategy, pursue international cooperation, including the U.S. Genesis Mission.
• 5.Developing talent to support the “Joy and Art of Discovery”
  We will create an environment where researchers, particularly early-career scientists, can tackle unexplored questions across disciplinary boundaries with AI as a reliable partner.

It is our responsibility as people engaged in science to discern the essence of AI and guide technology toward the sound development of society. RIKEN will expand the space of scientific exploration through a new form of Kyuchi-kyosou between humans and AI, and, together with all of you in Japan and around the world, pioneer a better future for society.

Makoto Gonokami
President, RIKEN