Excluding China from the global AI development is a wrong game to play

Apparently, the U.S. and its allies have clear incentives to isolate China from the global artificial intelligence (AI) industry. Accusations of China’s dual use in AI have become an ubiquitous tone in the United States’ discussion of China’s current AI development. Nevertheless, China’s comparative advantage in AI operational application, coupled with its significant contributions to global AI development, presents an opportunity to enhance and synergize with the United States’ AI industry.

In March 2024, Britain and South Korea co-sponsored the Seoul AI Summit 2024. The participants were mostly the same as in the previous summit in the UK in 2023, except for China’s absence in the ministerial statement of the Seoul AI security dialogue. Recently, G7 countries expanded AI’s Hiroshima Code of Conduct into Africa without China’s participation. Meanwhile, U.S. Congress lawmakers introduced the “Enhancing National Frameworks for Overseas Critical Exports Act,” aiming to advance export control on Artificial Intelligence to its defined enemies, such as China. On June 21, 2024, the U.S. Treasury Department advanced outbound investment regulation, which will restrict American investment in technologies important to “the next generation of military, intelligence, surveillance or cyber-enabled capabilities” that threaten the United States’ security. All these distinctive interactions among the United States and its allies mirror their apparent goal of disconnecting China from cutting-edge AI technology and AI governance.

A critical question follows: In what ways will AI development change if China is not actively participating at the multinational level? Despite China not being as competitive as the U.S. in AI research and development (R&D), it has played a critical role in exporting its AI applications and associated digital infrastructures to developing countries to promote public goods. The footprints of China’s AI program export cover most regions in the Global South. According to the RAND corporation’s “China’s AI Exports Database,” China exported 155 AI projects in 64 countries between 2000 and 2017, including AI applications and AI infrastructures. In Pakistan, China managed the Safe City Islamabad Project which utilized AI to reduce criminal rates, and the Gwadar Smart Port City Master Plan, which integrated AI into daily port custom procedures. In Ukraine, China helped the Ministry of Infrastructure to build a surveillance and inspection system. In 2015, China and Zambia signed a cooperation framework to develop the smart Zambia National Information and Communication Technology Program, which helps Zambia to build IT talents and digital infrastructures.

Moreover, China’s massive manufacturing industry provides high-quality and sufficient data for foreign AI firms to build a resilient digital supply chain management system. Numerous multinational enterprises such as Tesla, Apple, BMW and Bosch have adopted AI into factory’s production management systems, significantly boosting production efficiency. In China, the U.S. based multinationals employed 1.7 million foreign workers, with over 45% of those employees working in manufacturing sectors. For those companies, having transparent information about the productivity of factories worldwide is indispensable. Therefore, China is an essential hub for global AI integration into manufacturing, providing sufficient and transparent production intelligence.  

More importantly, China’s industrial competence in operational AI applications such as smart factories, intelligent housing, and AIoT (artificial internet of things) can bolster the United States’ AI market. In April 2024, when Elon Musk visited China, he reached a deal with Baidu on the “Full Self Driving” system. As one of the leading AI firms in China, Baidu’s cooperation with Tesla on Automated Driving has indicated China’s industrial focus in AI applications. According to the Center for Security and Emerging Technology (CSET)’s report China’s Artificial Intelligence Industry Alliance, China’s major AI industrial group, the Artificial Intelligence Industry Alliance (AIIA), specializes in AI applications for the commercial market. 63 percent of AIIA’s members focus on AI-driven business solutions, data storage and processing services. In the last five years, substantial venture capital poured into China’s AI industry in which around $25,000 millions were in robot sensors and IT hardware. The CEO of Baidu also just admitted that “[w]ithout applications, having only foundational models, whether open-source or closed-source, is worthless.” In contrast with China’s AI industry, the United States’ AI companies such as Microsoft’s OpenAI, Meta, Google, and Nvidia, are leading entities in AI algorithms and AI chips, but not in AI applications. As a result, China’s AI application companies and the United States’ AI firms can benefit from each other and advance the global AI industry through complementing each others’ strength in AI development.   

In addition, it is already all but pointless to try to keep AI models away from Chinese firms. China’s access to advanced AI models mainly relies on an open-source AI ecosystem in which model developers grant free permission to other users to utilize their published AI. The 01.AI company in China, established by Kai-Fu Lee, introduced its first open-source AI model Yi-34B. According to 01.AI’s public release on social media, the Yi-34B large language model (LLM) is indeed built on Meta’s Llama 2 model, which is a rare example of leading open AI foundation model from the American company. Unless Meta does not permit other firms to reuse its Llama 2, Chinese AI firms have full legal access to Llama 2 models. As a result, the United States’ imposing AI model export to certain countries is ineffective and might disrupt the openness of the open-source AI model ecosystem. 

Despite the United States’ deep concerns about China’s dual-diligence in AI development, the acknowledgment of human intervention in the AI military system is shared between China and the United States. Official documents from the U.S. departments of defense highlighted that AI’s military capability will be “consistent with their respective obligations under international law, in particular international humanitarian law,” and documents from the Chinese defense departments also underlined that relevant AI-integrated weapon systems must be “under the control of human agency.” Two countries’ consensus in strict control of AI military integration is an authentic promise of preventing AI dual-diligence in lethal autonomous weapon systems(LAWS).

Meanwhile, China’s current AI R&D capability can not enable it to pose an AI threat to the United States. According to Marco Polo, in 2022, about 42% of most top-tier AI researchers work in the United States. The Stanford AI Index Report 2024 shows that 58% of machine learning models produced by the industry and only 17% of models are from academic institutions. In other words, the United States has more talents to innovate future cutting-edge frontier models, and the U.S. AI industry is attractive enough to keep much of these talents in the U.S.. Therefore, in the current preliminary period of AI development, the benefit of having two countries’ cooperation on AI is overwhelming and substantial. 

Both the United States and China are important players in AI development. China’s scale of AI market and comparative advantage in operational AI application can enhance the United States’ cutting edge innovation in the machine learning models and AI chips. Vice versa, Chinese AI companies also require technological cooperation with the American AI community. Unlike most other technologies, the AI ecosystem—especially as it continues to expand into other fields—requires multinational cooperation. As Langdon Winner argued in his article “Do Artifacts Have Politics,” any technical system requires a certain social condition as its operating environment. AI is indeed a technology that relies on shared information between countries and an unbiased algorithm that includes the opinions of all the countries. The United States and China, as the leading countries in AI development, need to first work together and then move beyond the bilateral cooperation toward multilateral AI groups, showing the world the power of leadership in safeguarding AI development.