Building a Specialized National AI Model with University Collaboration in Tokyo

Overview of Tokyo’s National AI Project

In May of this year, Tokyo, in collaboration with the General Incorporated Foundation GovTech Tokyo, announced an initiative to develop a domestic AI model tailored specifically for administrative tasks. This engineered AI model aims to support administrative work by learning laws, regulations, and related expertise. The initiative is set to utilize the capabilities of advanced large language models, which are trained on vast textual data to generate language and summarize information.

Purpose of the Initiative

Large language models have been widely utilized across various fields in Tokyo and beyond. However, administrative work encompasses specialized terminology, often leading to inaccuracies—commonly referred to as 'hallucinations' in generative AI. These inaccuracies can significantly impact citizens' rights and governance, necessitating an AI that demonstrates transparency and reliability in its responses.
For this reason, the goal of this initiative is to create a secure AI model that specializes in administrative knowledge, ensuring accuracy and transparency in the processing of information.

Call for Proposals and Selection Process

In May, a formal call for proposals was issued, inviting academic institutions with a strong research background in large language models to participate. The submission period closed on June 12, during which six universities submitted their proposals. An external panel evaluated the proposals through presentations, considering their potential impact.

After a thorough review process, Tohoku University was selected to partner with GovTech Tokyo in this pioneering project. A formal agreement is expected to be signed in early July, establishing a framework for the project's execution, which will last until March 31, 2025.

What to Expect Moving Forward

This collaborative effort is set to reinforce public trust in governance by providing accurate information sourced from state documentation and regulations. For citizens, this means a clearer understanding of their rights and responsibilities. Furthermore, the AI model will allow for the verification of responses, thereby bolstering the model’s reliability.

The telegraphed goals of this initiative complement the broader Tokyo AI Strategy, which emphasizes improving the quality of public services and enhancing productivity and efficiency in government operations.

Visit the following links for more information on the initiative and the broader Tokyo AI Strategy:

This project reflects Tokyo's commitment to leveraging AI technology for enhanced governance, ensuring that the solutions developed are both accurate and beneficial for its citizens.