In the rapidly evolving landscape of generative AI, distinct regulatory and ethical approaches have emerged, reflecting the values, ambitions, and constraints of various global players. We previously delved into the contrasting strategies of the United States and the European Union, two titans in the realm of artificial intelligence. Today, we broaden our lens to encompass the unique approaches of Japan and China. These Asian powerhouses have carved out their own pathways in the generative AI space, informed by their cultural, economic, and geopolitical contexts.
Japan’s Approach to Generative AI
With a vision centered on human-centric AI, Japan places emphasis on respecting human dignity and values, promoting social welfare and diversity, and fostering trust and collaboration. Japan’s approach to generative AI is grounded in the Social Principles of Human-Centric AI, published by the government in 2019. These Social Principles enumerate 10 key aspects including transparency, accountability, fairness, security, privacy, education, research, governance, and international cooperation. The overall idea is to realize these goals not by reigning in AI use, but through AI.
To maximize the positive societal impact of generative AI while minimizing its negative ramifications, Japan employs a strategy of ongoing development and revision of AI-related regulations. This strategy follows a risk-based, agile, and multistakeholder process. Several initiatives have also been launched to encourage data sharing and utilization among public and private sectors as well as internationally, notably the Data Free Flow with Trust (DFFT) framework.
Similar to the EU, rather than adopting a one-size-fits-all regulatory approach, Japan’s policy evaluates generative AI applications based on their specific benefits and risks. This enables a more flexible and adaptive regulatory environment that can respond to changing technologies and circumstances. Various laws and guidelines have been developed or revised to guide this sector, such as the Act on the Protection of Personal Information (APPI), the draft AI Guidelines, the Copyright Act, and the soon-to-be proposed Generative AI rules.
With regard to AI regulation, then, Japan is leaving its mark globally. But when it comes to actually developing their own LLMs, Japan is perceived to be behind the US, EU, and China due to a lack of computing and human resources. But it has plans to catch up. Through various programs and projects like the super ambitious Moonshot Research and Development Program (MRDP) and the Artificial Intelligence Strategy Council (AITSC), significant investments have been made in research and development. The country has also capitalized on its supercomputing capabilities, notably with publicly controlled Fugaku, to develop large language models based on Japanese data. Private sector efforts are underway to add much needed infrastructure. Nevertheless, Japan is home to several notable startups in the generative AI space, including Kotoba Technology.
China’s Approach to Generative AI
China’s stance on AI is fundamentally geared toward becoming a global leader in artificial intelligence by 2030, as laid out in its Next Generation Artificial Intelligence Development Plan. This has led to fierce competition, particularly with the US, further fueled by concerns around AI’s potential to tip the scale in the geopolitical and military context. A distinguishing feature of China’s approach is the symbiotic relationship between the government and technology companies like Tencent, Alibaba, and Baidu. This unique public-private partnership has made China a powerhouse in AI research and applications, from natural language processing to autonomous vehicles.
Ethically, China’s AI development is guided by a concept known as “Ethical AI,” introduced in the Beijing AI Principles. These principles, which somewhat parallel Japan’s Social Principles of Human-Centric AI, cover aspects like shared benefit, sustainability, and security. However, unlike the democratic governance models of Japan, the US, and the EU, China’s approach to AI ethics and policy is more centralized and tightly controlled by the state. The government has the final say on what counts as ethical or what needs to be regulated, often tying these decisions closely to broader state objectives, such as social stability or national security.
Honing in on generative AI in particular, the recently enacted Interim Measures for the Management of Generative Artificial Intelligence Services serve as a case in point for China’s centralized, state-driven approach to AI governance while not losing sight of the concern for lawful rights of individuals. These Measures provide a comprehensive legal framework that addresses key ethical concerns like discrimination, intellectual property rights, and public morality, all while aligning closely with broader state objectives such as national sovereignty and social stability, making China a leading, proactive player in AI regulation. By setting forth requirements that explicitly mandate the adherence to “Core Socialist Values,” as well as explicit technical and ethical guidelines for AI development, these Measures exemplify China’s commitment to harmonizing rapid technological advancement with its unique socio-political landscape. This regulatory approach reflects China’s desire to maintain a delicate balance: fostering innovation and global leadership in AI, while ensuring that such advancements are in lockstep with its national ethos and global ambitions. Some voices say that, at least for now, innovation takes priority, while others emphasise the urgent interest of the Chinese government to control information as this is key for its capacity to shape public opinion and ensure the Chinese government’s legitimacy.
China’s innovation in generative AI is robust, led by both state-funded projects and commercial enterprises. Massive datasets available in China, coupled with less stringent data privacy laws compared to the EU, give an edge to Chinese companies developing AI technologies, at least with regard to computer vision due to more prevalent surveillance cameras, but less so with regard to written text; in this aspect the Chinese corpus cannot compete with what is available in English, making it fall behind the US in the development and implementation of LLMs.
Similar to leading US companies, open-sourcing AI models has also been the approach by some Chinese companies. Yet, this is not of great advantage for the Chinese as computing power is restricted, given the difficulty of procuring NVIDIA’s advanced GPU chips.
Conclusion
In conclusion, the world of generative AI is as diverse in its regulatory and ethical considerations as it is in its technological applications. Whether it’s the United States’ market-driven model, the European Union’s human-centric, risk-based framework, Japan’s agile and multistakeholder approach, or China’s state-centered strategy, each offers valuable lessons. As we forge ahead into an era where generative AI will increasingly become a staple of everyday life, understanding these different models helps us appreciate the complex tapestry of considerations that guide AI development and deployment. It can hopefully also aid in debunking the argument against regulation of AI as it would lead to a disadvantage in international competition. When we note how strictly AI is regulated in China to ensure stability and that they have to grapple with resource restrictions both in terms of computing power as well as talent that is driven to the US, we can hopefully focus more on the significant risks that AI poses for all of society, on a global level, and tackle the very hard but very necessary challenge of making sure we get the regulatory piece going. A comparative view of what other nations are doing may also prove useful to prevent inaction out of fear of not getting it right at first try. Quoting Tom Wheeler, former chairman of the Federal Communication Corporation, from a recent podcast appearance concerning AI regulation: “There are no paths. Paths are made by walking.”
