The state of science and technology in the 21st century is undergoing a major transformation due to the advancement of information technologies (IT), such as artificial intelligence (AI), big data and machine learning, Internet of Things (IoT) devices, and smart robotics. The transformation of research and development (R&D) processes through digitalization is not only improving the efficiency of R&D, but also promoting new discoveries in various scientific areas and creating new scientific methodologies.

In this proposal, we aim to automate scientific research, which is also a promising field of application for AI. We propose the following three R&D themes focusing on the handling of hypotheses.

• (1) Large-scale and comprehensive hypothesis generation and search
• (2) High-throughput hypothesis evaluation and testing
• (3) Human-centric scientific research cycle integration

The automated AI system will be implemented as a human-in-the-loop system that executes a closed cycle of hypothesis generation and verification. This AI system is expected to become a powerful tool for cutting-edge research. For example, in the field of life science and medicine, such an AI system will enable the rapid development of vaccines for new viruses and the rapid discovery of safe and effective drugs. The other AI system in the field of materials science will enable the automatic design and optimization of materials with desired functionalities as well as its synthesis methods. It will also be used by many scientists to test their own working hypotheses and search new hypotheses beyond their cognitive bias. Moreover, the systematization and dissemination of its usage of the AI system in many scientific fields will raise the level of Japan's R&D capabilities.

In addition to the promotion of R&D of platform technologies for the AI system, the government needs to develop human resources and organizations that can freely use AI, big data technologies, and robotics, create a framework for data sharing, and foster communities. In this proposal, we propose the following three aspects of promotion methods.

• (1) Joint-use facility
• (2) Grand Challenge
• (3) Education and dissemination across disciplines

The R&D and operation of AI systems will be conducted in each fields for the time being, since the tasks and devices to be automated differ from field to field. As well as the formation of centers of excellence in each field, the establishment of a joint-use facility with a "hub" functions is required. Since there are various possible approaches to the realization of the AI systems, it is desirable to implement a R&D program that sets grand challenges and solicits a wide range of ideas to achieve them.

Incentive design is also important for the sharing of negative data and failures because the quantity and quality of data are key for machine learning. Education and dissemination of mathematics, logic, and other reasoning tools and verification protocols are also becoming increasingly important. In addition to hands-on training, it is desirable to promote R&D with attention to the systematization of education.