Key points
- China’s recent development in AI thwarted US suppression policy
- Chinese startup DeepSeek released its latest open-source model
- Its performance is comparable to heavyweights like OpenAI
- Chinese companies have developed cheaper and more competitive models
ISLAMABAD: Industry observers have said that China’s recent rapid development of its artificial intelligence (AI) industry, particularly in large-scale language model technology, has significantly thwarted the US suppression policy.
According to Xinhua News, on the opening day of the World Economic Forum (WEF) 2025 Annual Meeting, Chinese AI startup DeepSeek released its latest open-source model DeepSeek-R1, which has achieved an important technological breakthrough, using pure deep learning methods to allow AI to emerge with reasoning capabilities spontaneously.
According to DeepSeek, this model’s performance in tasks such as mathematics, coding, and natural language reasoning is comparable to that of leading models from heavyweights like OpenAI.
China AI technology
China’s AI technology has become a hot topic in the relevant discussions of the WEF annual meeting in Davos, Switzerland.
Max Tegmark, a well-known AI expert at the Massachusetts Institute of Technology, told Xinhua that China has made a lot of progress in AI over the past year.
Tegmark said, “China lagged slightly behind in cutting-edge large language models a year ago, but it has now caught up.” He emphasised the “stupidity and mistake” of undermining scientific and technological cooperation owing to geopolitical concerns.
Just a few months after OpenAI released the reasoning model o1 in September 2024, the Tongyi Qianwen team of the Chinese tech giant Alibaba launched the experimental research model QwQ-32B-Preview at the end of November. In various tests, it showed reasoning capabilities comparable to or even exceeding OpenAI’s o1 model.
In late December, DeepSeek released the hybrid model DeepSeek-V3, whose evaluation scores in different tests have surpassed open-source models such as Llama-3.1-405B, and its performance is comparable to the world’s top closed-source models GPT-4o and Claude-3.5-Sonnet. In addition, DeepSeek-V3 is more cost-effective.
US-China competition
The Economist magazine wrote, “The United States tried to prevent China from catching up in the field of AI and that China’s recent progress is upending the industry and embarrassing American policymakers.” “The success of the Chinese model, coupled with changes in the entire industry, may reshape the economic pattern of the AI industry,” it said.
The New York Times noted that compared with US giants such as Google and OpenAI, Chinese companies have developed a cheaper and more competitive model.
Jeffrey Ding, an assistant professor at George Washington University who specializes in emerging technologies and international relations, said, “The US restrictions on Chinese chips forced Chinese engineers to train it (the model) more effectively so it could still be competitive.”
In addition to large language models, physical AI will also bring good opportunities to China. Li Yifan, co-founder of Hesai Technology, said at the WEF annual meeting that when digital AI is combined with physical products — whether in cars, robots, or other consumer electronics — Chinese companies have significant advantages in supply chain management, manufacturing capabilities, closed-loop processes, cost control, and large-scale production.
