China Closes AI Gap with US Amid Chip Restrictions
China is closing the AI gap with the United States by rapidly deploying AI applications. Chinese IT giants are designing their own LLMs despite chip constraints. In LLM training, Huawei's Ascend 910B AI chip performs competitively.
Industry experts and observers point to China's quick breakthroughs in AI application deployment, as well as the government's support for the technology, as important drivers of success.
Chinese IT companies are competing to develop their own large language models (LLMs), which are critical for generative AI technologies such as ChatGPT, with some claiming to rival or outperform their American equivalents. This comes as the US tightens limits on advanced semiconductors used to train AI systems.
Winston Ma, author of "Digital War - How China's Tech Power Shapes the Future of AI, Blockchain, and Cyberspace," observes that China's scarcity of powerful graphic processing units due to US export restrictions is creating a push for AI efficiency in the country.
For example, Beijing-based start-up Shengshu AI recently introduced Vidu, a text-to-video converter, following in the footsteps of local companies Kuaishou and Zhipu AI. These Chinese businesses are democratising AI video tools around the world, in contrast to San Francisco's OpenAI, which has yet to make its tools widely available.
Furthermore, Chinese companies are promoting global AI development by releasing open-source LLMs, which allow anybody to build AI systems. Alibaba Group Holding's Qwen2 open-source LLM family, which debuted in June, won the ranks on Hugging Face, an open-source AI model developer community.
Analysts attribute China's rapid AI advances to its ability to overcome chip limitations and build the computational power required for training local LLMs. Despite US export limitations on Nvidia's high-end chips, Beijing and Chinese tech behemoths have effectively built significant intelligent computing capabilities using locally designed solutions.
Huawei Technologies, for example, has emerged as a prominent participant in this space, with their Ascend 910B AI CPU outperforming Nvidia's A100 when training LLMs. This accomplishment demonstrates China's determination to build a self-sufficient AI infrastructure.
The state's investment in computing resource infrastructure has also played an important role in easing concerns about chip shortages. China's enormous market size and strong public sector demand for AI and its applications create a favourable environment for future AI developments in the country.
China narrowing AI gap with US through rapid AI application deployment
Chinese tech firms developing own LLMs despite chip restrictions
Huawei's Ascend 910B AI chip shows competitive performance in training LLMs
Source: SCMP