While big language models from Western tech companies show the potential to disrupt everything from marketing, education to coding, China is racing to cultivate its homegrown AI pioneers by ramping up state aid.
Beijing is now seeking public opinion on a design policy aimed at developing artificial general intelligence, or AGI, a category of AI that can theoretically perform all human tasks. In short, the aim of the policy is to support AI companies by strengthening support for cloud providers and data companies.
It is not uncommon for the capital to take the lead in policy making in emerging industries. For example, Beijing was the first to allow self-driving robotaxis to transport passengers on open roads under certain restrictions.
The AGI blueprint contains action plans around three main areas: computing power, training data, and applications.
The first strategy required closer collaboration between cloud providers, the sources of computing power, and universities and companies, which consume large amounts of computing power to train large language models, multimodal learning, and other AI. The policy proposes a state-backed, centralized platform that allocates public cloud resources to users based on demand.
Alibaba accounted for more than a third of spending on cloud infrastructure services in China last year, ranking first, according to market research firm Canalys. Huawei, Tencent and Baidu lagged behind.
The second strategy recognizes the lack of quality Chinese-language data and encourages the “compliant cleaning” of such datasets, including data anonymization, likely an effort to comply with China’s new strict privacy laws. The process will no doubt be time consuming and labor intensive, as we’ve seen OpenAI rely on Kenyan workers to manually label training data and remove toxic text.
Beijing’s big data exchange, launched by the government in 2021 to facilitate the trade of data between facets of society, will support the data sourcing process.
Finally, the policy lists potential pilot applications of AI ranging from using AI in medical diagnosis, drug making, financial risk management, transportation to urban management.
The proposed policy also touches on the importance of software and hardware infrastructure for AI training. Amid escalating competition between the US and China, the latter is looking to support innovation in key technologies such as semiconductors.
The US is already restricting exports of Nvidia’s powerful AI chip H100 to China. In response, Nvidia released a less powerful processor for China to avoid export controls. Domestic companies, such as tech giant Huawei and startup Biren, are also working on Nvidia alternatives.