Toshiba Group Announces AI Governance Statement, to Promote Provision of Trustworthy AI

- AI engineer headcount of 2,100 surpasses FY2022 target
- Group to train more people in AI literacy, strengthen development of framework for maintaining quality of AI systems

August 30, 2022

TOKYO—Toshiba Corporation (TOKYO: 6502) today announced the Toshiba AI Governance Statement for promoting the development, provision and use of trustworthy AI. The statement establishes seven principles aligned with Toshiba’s management philosophy, and will guide Toshiba Group in training human resources who are able to research, develop, and use AI; strengthening development of its framework for maintaining the quality of AI systems; and promoting Group-wide AI governance.

Digital transformation (DX) is a global trend, making the development and application of AI technology more important than ever. Toshiba Group is advancing the realization of carbon neutrality and a circular economy through DX, and promoting the application of AI that will help to solve social issues in critical infrastructure and other areas.

Against this backdrop, Toshiba introduced an AI engineer training program in 2019. This has increased the AI engineer headcount from 750 to 2,100, surpassing this fiscal year’s target of 2,000. The company has also set up an in-house platform for proactive sharing of information on over 200 AI assets, 55 of which are now showcased at "Toshiba AI Technology Catalog."

While AI adds to convenience, problems from its misuse or unintended operation point to the need for ethical guidelines and governance covering its utilization. In Japan, the government has formulated multiple guidelines*, and legislatures overseas are increasingly taking action, exemplified by the EU’s AI Act. Addressing AI governance is now a must for companies that use AI.

In April this year, Toshiba established the AI Center of Excellence Project Team, which is charged with leading group-wide promotion of the governance required for AI’s implementation in society. The project team formulated the Toshiba AI Governance Statement, in which seven principles clarify Toshiba Group’s vision of AI technology grounded in respect for human dignity, development of AI and talent, and an emphasis on fairness. It provides Toshiba with foundations for establishing AI governance and implementing initiatives for the development, provision, and use of trustworthy AI systems (Figure 1). 

Figure 1: Overview of AI Governance in Toshiba Group

Toshiba Group will increase the number of employees who understand AI. Alongside the 2019 engineer training program, it recently introduced new training that will enhance the AI literacy of non-engineering managers, to improve their ability to provide trustworthy AI to customers. Toshiba will also develop training programs for a wider range of employees, not limited to managers or engineers.

Toshiba is also developing frameworks to maintain the quality of its AI systems, including AI Quality Assurance Guidelines and quality cards that visualize quality assurance from the customer perspective. The cards detail the performance of the AI systems, such as their recognition and accuracy rates, and also indicate their strengths and weaknesses, based on data trends found in learning and evaluation processes. This allows customers to understand the AI system’s characteristics and use it correctly.

Another measure is the implementation of MLOps, infrastructure for machine learning system development and operation, to maintain reliable machine learning models and prevent performance degradation due to changes in the operational environment. AI systems with MLOps are now in use at Toshiba manufacturing sites and in social infrastructure, and the company looks forward to expanding their applications.

Toshiba Group is aware of its responsibilities to people and the future, and is realizing AI governance based on the AI Governance Statement as a way to further contribute to society. Inspired by a venture spirit handed down from generation to generation, Toshiba Group is firmly committed to develop, provide and use trustworthy AI systems to resolve social issues.

 

(Note)
*AI guidelines formulated by Japanese government:
- Social Principles of Human-Centric AI(261KB)
- AI Utilization Guidelines(1.19MB)
- Governance Guidelines for Implementation of AI Principles(1.02MB)