The AI China Report #4
Your executive summary on the geopolitics of AI & China 🇨🇳 - Issue #4
Welcome back to The AI China Report. This week we’re covering the rise of AI-generated digital humans in China, the Biden Administration’s plans to announce new sanctions on chip exports, and Saudi-China relations.
QUOTE OF THE WEEK
"The United States will assert ourselves when our vital interests are at stake, but we do not seek to 'decouple' our economy from China's."
- Treasury Secretary Janet Yellen
ONE BIG STORY: CHINA’S DIGITAL HUMANS
AI-generated avatars, or “digital humans” are becoming increasingly popular in China. Beijing officials have estimated that the digital human sector will surpass by $50 billion yuan (~$7B USD) by 2025 and $38.5B by 2030, according to a report released by QbitAI. There are over 280,000 Chinese companies actively working in the digital human space. These digital humans exist solely on the internet and are typically divided into two categories: character-based and function-based.
Character-based digital humans are often created in the likeness of celebrities or IP from pop culture characters.
Function-based digital humans, are often used for their reduced cost and efficiency compared to using real people, and can take the form of musicians, fashion models, news anchors, influencers and even financial advisors.
These function-based digital humans are putting some people out of work, especially in the live commerce sector. Live commerce, similar to QVC in America, is incredibly popular in China. There are over 1 million people working as live commerce hosts in China, having brought in over $140B in goods sold last year. Chinese tech giants like Tencent and Baidu (the Google of China) both offer platforms that allow for the creation of AI-generated avatars that can livestream 24/7 at a much cheaper cost than paying a real person to host. Baidu’s digital avatar platform XiLing, which is comparable to US-based Synthesia, allows for creation of digital avatars from existing IP or custom creation based on industry and allows for translation to multiple languages.
The rise of AI-generated avatars, particularly in China, underscores a significant shift in the digital landscape. While they offer cost efficiency and scalability, particularly for companies like Tencent and Baidu, there's an undeniable impact on employment in certain sectors. The staggering projections by Beijing officials highlight not only the commercial potential but also the broad cultural adoption expected in the coming years. As hundreds of thousands of companies dive into the digital human space, one can't help but wonder how this will reshape industries, societal norms, and individual careers in the not-so-distant future. As we stand on the cusp of this AI evolution, it signals us to adapt, innovate, and reconsider the essence of human roles in an increasingly digital world.
NEWS ROUNDUP
The Biden administration is planning to announce new controls on artificial intelligence chips and equipment that can be sold to China as early as this week, with the aim of preventing China from gaining a military advantage in AI. These rules, to be issued by the Commerce Department, are intended to close loopholes on export controls announced a year ago and will apply to slightly less powerful chips that still have advanced capabilities. The administration's approach to China involves asserting national security interests while not seeking to harm China's economy, and it is part of broader efforts to manage relations with China. (Axios)
China plans to increase its computing power by over a third by 2025, aiming for more than 300 exaflops (an exaflop is a measure of performance for a supercomputer that can calculate at least 1018 or one quintillion floating point operations per second), as it seeks to bolster domestic tech self-reliance in the face of U.S. sanctions. This effort includes the construction of 20 new smart computing centers and the development of homegrown, reliable software to safeguard the supply chain, with a focus on industries like manufacturing, education, finance, transportation, healthcare, and energy. (Bloomberg)
A research institute in Taiwan, Academia Sinica, withdrew its Chinese-language AI chat program after it started delivering Chinese government propaganda responses. Users were disturbed when the chat program identified itself as "Chinese" and stated that Xi Jinping was the national leader, contrary to Taiwan's majority view of not being Chinese and valuing democratic and pluralistic society. The incident highlights concerns about information security and cognitive warfare in the context of AI models influenced by political ideologies. (Radio Free Asia)
Fears have arisen at Saudi Arabia's King Abdullah University of Science and Technology (Kaust), which recently launched an AI system called AceGPT in collaboration with two Chinese universities, that its collaboration with Chinese institutions in artificial intelligence (AI) could jeopardize its access to US-made chips necessary for AI technology. This concern reflects growing unease about technology transfer between Gulf countries and China, with the US imposing export restrictions on key components, such as graphics processing units, which are vital for AI model development. Riyadh and Abu Dhabi aim to balance flourishing trade ties with China while sustaining their relationships with the US, their primary security partner, in the competitive field of AI in the Gulf. (Financial Times)
China has issued security requirements for companies offering generative artificial intelligence (AI) services, including a blacklist of prohibited sources for training AI models. The proposed regulations by the National Information Security Standardization Committee outline security assessments for public-facing generative AI models, with content containing over 5% illegal or harmful information, such as advocating terrorism or undermining national unity, to be blacklisted. Additionally, the rules require organizations to seek the consent of individuals whose personal information is used for AI model training and provide guidelines to avoid intellectual property violations. (Reuters)
DATA TO KNOW
Know someone who should be in the know about AI and China? Share The AI China Brief 🇨🇳 with your friends and colleagues!
Have feedback on how we can improve? Want to discuss this issue’s content? We’d love to hear from you in comments below!