1. Overall Approach

• In line with its policy proposal entitled AI Utilization Strategy: For an AI-Ready Society (February 2019),^#1 Keidanren has been encouraging Japanese companies to take steps to become AI-ready, based on its fundamental recognition that artificial intelligence (AI) is one of the core technologies towards the creation of Society 5.0 for SDGs.

• Against this background, generative AI, and in particular the interactive ChatGPT (Generative Pre-trained Transformer) developed by the US company OpenAI, has the potential to fundamentally change the very nature of human society.

• While public interest in the use of generative AI is increasing rapidly in Japan and elsewhere,^#2 for Japan to continue targeting the establishment of Society 5.0 for SDGs, we must not lose sight of our ultimate purpose for making use of AI according to a clear and agile blueprint. Grounding our efforts in the accumulation and linkage of large-volume data,^#3 we must continuously pursue (1) further use of AI, and (2) measures to tackle the accompanying risks.

• At the same time, assuming we pursue (1) and (2) above, it will be of vital importance to (3) establish and reinforce Japan's own ability to develop AI. Japan differs from other countries in terms of language and cultural background; it can get only so far by taking advantage of AI developed in other countries. Unless we have both the will and the ability to create our own innovative AI models, our capacity to explore how it is being utilized and take a leading role in the creation of international rules is in doubt.

• The concepts expressed in the Principles for Utilization of AI that Keidanren formulated in 2019 are still universally applicable even today when generative AI is flourishing. Based on current circumstances and in light of these principles, we hereby propose measures to pursue initiatives (1) to (3) above.

2. Strategies to Promote Further Use of AI

• As we make increasing use of AI to bring about Society 5.0 for SDGs, we should avoid making utilization of AI itself into our objective, and instead integrate AI into the blueprint for achieving Society 5.0 for SDGs. Going forward, it will be essential to establish overall "command center" capability within the government and formulate a comprehensive national strategy setting out how Japan will make combined use of AI and other digital technologies with various types of data.

• For companies and other entities of all kinds to substantially enhance productivity by making maximum use of AI, they need to move beyond AI-readiness to rapidly become AI-powered (i.e., utilization of AI is a given throughout entire organizations). Keidanren has already proposed guidelines (see figure below) that indicate steps for achieving AI-powered status; from now on, we will revise these guidelines as and when appropriate in light of changing circumstances, including ongoing progress in the use of generative AI.

  

  ( View larger version )

• Use of AI to improve productivity will rely on simultaneous reskilling of existing human resources. In addition to the obvious need for reskilling to enable migration of labor into growth industries, it would also be advisable for human resources of all kinds in every sector to acquire the ability to utilize AI.

3. Measures to Tackle the Risks Accompanying Use of AI

• As we follow our strategies and principles for promoting use of AI, we need to set up "guardrails" to lead us in the appropriate direction, having recognized the merits and risks of using generative AI in particular. Examples of key considerations are set out below.

• Note that, to ensure appropriate use of AI, we should not require developers and service providers alone to make all the effort, but should address this issue throughout the entire AI ecosystem, making it clear where certain responsibilities lie. It is also necessary to encourage end users to take appropriate actions, including enhancing their own AI literacy. With regard to international rulemaking in relation to generative AI, Japan should take the initiative not only in the Hiroshima AI Process, but also in the Organisation for Economic Co-operation and Development (OECD) and other frameworks involving multiple stakeholders.

(1) Education

• In educational settings, while it is important to cultivate schoolchildren's ability to make full use of AI, there are concerns that its use may cause their thinking capacity to decline due to their inability to distinguish erroneous answers or content generated by AI.^#4 It is vital, therefore, to forge a broad consensus regarding the situations in which use of AI should be either encouraged or restricted according to age, taking into account ability to utilize information obtained using AI, among other relevant factors.

• We also need to consider wide-reaching revisions to the way future generations will be educated premised on the use of AI. Now that generative AI can easily create text based on information posted on the internet, we urgently need to consider the best comprehensive form of education that would teach children not to "avoid relying on AI," but to "use AI correctly" by making their own judgments regarding the accuracy and authenticity of information it generates.

(2) Intellectual property rights (copyright infringement, etc.)

• As use of generative AI increases, attention must be paid to how intellectual property rights such as copyrights and trademark rights are handled. If the copyrighted material of another individual or entity is used as data for machine training purposes, it can be used without the copyright owner's permission, provided it is used only as required for data analysis. However, Article 30-4 of Japan's Copyright Act provides that, if such use "would unreasonably prejudice the interests of the copyright owner" in light of the nature or purpose of the copyrighted material that is used, or the circumstances of its exploitation, the permission of the copyright owner is required.

• This being the case, we need to consider what kind of rules are appropriate and investigate the best form of legal system for Japan, including soft laws, while also paying attention to consistency among countries, in cases where: (1) data that could give rise to intellectual property rights is used for AI training purposes, (2) rights in relation to AI-generated content could arise, or (3) AI-generated content could infringe other rights including intellectual property rights.

(3) Governance and ethical considerations (dealing with bias, etc.)

• Considering the potential for biases in training data, or arbitrary hyperparameter tuning, a major challenge from now on will be ensuring AI-related governance, particularly in terms of ethical considerations. It will be essential for countries to build on the outcomes of the G7 Digital and Tech Ministers' Meeting in Takasaki, Gunma^#5 and cooperate to facilitate the use of trustworthy AI in a secure digital environment and to determine the best form of governance for both AI and the data sets it uses, ensuring fairness from an objective point of view as they seek to foster coordination among systems.

• In addition, further deliberation is needed to address governance issues such as the handling of personal information included in AI training data and the disclosure of personal information in generative AI outputs.

4. Establishing and Reinforcing Japan's Own Ability to Develop AI

• Most current generative AI models including ChatGPT have been developed using training data in English, making it difficult to realize these models' full potential in Japan, which has a different language, culture, and so on. Moreover, other considerations such as economic security also make it vital for Japan to possess the ability to develop cutting-edge AI independently.

• At least as far as development of foundational AI models is concerned, Japan currently lags far behind internationally; however, in areas of research and development other than foundational models, such as image recognition technology, there are a number of fields in which Japan has achieved comparative advantage.

• Establishing and reinforcing Japan's ability to develop AI so that it reaches a level comparable with other countries will require us to accumulate and link large volumes of data, provide profound support for research and development, and train human resources, in addition to drawing on our existing capabilities including basic research. In order to leverage a combination of data and use of AI as a means of bolstering Japan's competitiveness, individual ministries and agencies must collaborate in line with a clear and comprehensive policy to take action with an unprecedented sense of urgency.

• When doing so, it will be strategically important to involve industry, academia, and government in the development and responsible deployment of AI. To ensure that Japan's AI can use modes of expression particular to the Japanese language and culture, such AI development should be based on trustworthy, large-volume data, including data resources made available by national research and development agencies.

• Meanwhile, in addition to pursuing our own independent research and development—taking care to ensure that Japan's efforts are not evolving in a different direction from the rest of the world—we should lead international collaboration, including with countries in the Global South.

1. Keidanren, AI Utilization Strategy (February 19, 2019)
2. Against a backdrop of debate in individual countries over the use of generative AI, paragraph 38 of the G7 Hiroshima Leaders' Communiqué (May 20, 2023) included an agreement to "task relevant ministers to establish the Hiroshima AI process, through a G7 working group, in an inclusive manner and in cooperation with the OECD and GPAI, for discussions on generative AI by the end of this year." Within Japan itself, meanwhile, the Liberal Democratic Party's Project Team on the Evolution and Implementation of AIs submitted an AI White Paper entitled Japan's National Strategy in the New Era of AI to Prime Minister Kishida (May 9), and in the same month the government set up the AI Strategic Council, which summarized AI-related issues (May 26).
3. On May 16, 2023, Keidanren published a proposal entitled Creating New Value through Utilization and Linkage of Data (in Japanese only). The proposal emphasized the importance of initiatives to encourage utilization and linkage of data in order to generate value for consumers, with the ultimate goal of realizing Society 5.0 for SDGs through a specifically Japanese form of co-creation-oriented DX (digital transformation).
4. AI Strategic Council document (May 26, 2023) entitled Tentative Summary of AI Issues
   https://www8.cao.go.jp/cstp/ai/ronten_youshi_yaku.pdf
5. On April 30, 2023, the ministers agreed on the G7 Action Plan for promoting global interoperability between tools for trustworthy AI.
   https://www.meti.go.jp/press/2023/04/20230430001/20230430001-ANNEX5.pdf