DeepSeek a breakthrough but bottlenecks remain
DeepSeek released its general large model DeepSeek-R1 last month, which attracted global attention with its low cost and high performance. The model's training costs were 10 percent of the industry benchmark using far less computing power resources than those of its international peers. This offers a new solution for breaking through the Western-dominated AI development model of relying on high inputs to make breakthrough.
At the same time, DeepSeek has adopted a completely open source strategy, disclosing algorithms, model weights and training details, so that global developers can learn from, improve and deploy models. The open source ecosystem helps to attract more developers and users to participate, promotes technology iteration, and is expected to change the winner-takes-all competition landscape.
Despite these breakthroughs, it should also be noted that China's original AI innovation still has a long way to go.
China's data infrastructure system construction is still in its infancy, the data acquisition and exchange mechanism is not yet sound, industry data and public data are difficult to obtain and access, and the data available for large models is limited. At the same time, data annotation is the basis for the supply of high-quality data. Due to the shortage of professional annotation talents, the quality of data annotation in China still needs to be improved, especially in areas such as medical care and autonomous driving where development needs are urgent and the professional requirements are high.
From a global perspective, the influence of Chinese domestic large models such as DeepSeek in the global technology ecosystem is still in its infancy. From a domestic perspective, the entire industry chain of China's AI development from basic research to technological innovation to scenario application has not yet been fully opened. The flow of factors such as technology, capital, data and talents that support the iterative development of large models is still blocked.
To this end, AI basic research and technological innovation should be continuously strengthened. The country should accelerate the construction of national strategic scientific and technological forces in the field of AI, promote the cross-integration of AI with basic disciplines such as mathematics, physics and brain science, and improve basic AI research. It should encourage open source AI technology, focus on open source projects, and promote open source contributors, service providers, users, operators and other entities to jointly promote AI technology innovation.
The authorities should provide more support to help cultivate and strengthen AI start-ups and provide scientific references for governments and financial institutions to accurately identify potential and high-value AI start-ups.
The country needs to give full play to its advantages in massive data and rich application scenarios, organize the advantages of scientific research institutions, leading technology companies, etc, focus on key vertical segments such as intelligent manufacturing and autonomous driving, and coordinate the layout of the large model industry application innovation engineering centers.
ECONOMIC DAILY
