DeepSeek: Pioneering Innovation Amidst Challenges

Many Chinese companies are now following in the footsteps of DeepSeek, working towards innovation and collaboration to achieve significant global advancements, despite ongoing sanctions from the United States, experts remarked on Sunday.

These insights came after reports indicated that the US National Security Council is examining potential national security concerns related to DeepSeek, a private Chinese startup that has recently made headlines with its highly cost-effective AI large language model.

DeepSeek has successfully established a local supply chain alliance, comprising 278 companies. This initiative has significantly boosted the domestic production of key intellectual properties, increasing the rate from just 19 percent in 2022 to an impressive 64 percent today.

Pan Helin, a member of the Ministry of Industry and Information Technology's Expert Committee for Information and Communication Economy, highlighted that "in the context of competition in AI between the US and China, DeepSeek has developed a new technical approach for large language models that does not depend on high-end chips and effectively reduces the required computing power. This innovation allows DeepSeek to navigate past the computing power constraints imposed by the US on Chinese enterprises. Importantly, it reinforces the notion that a country’s innovation capacity, rather than political barriers, will define its success in the global AI arena," he explained.

Despite US restrictions on chip exports to China, DeepSeek remarkably created its models. Analysts from Jefferies, a market consulting firm, suggested that the latest iteration of DeepSeek's models incurred a training cost of merely $5.6 million, less than 10 percent of what it costs to train Meta's Llama model.

Pan further predicted that DeepSeek will not stand alone as a pioneer; numerous other Chinese firms are expected to achieve similar breakthroughs.

Since the launch of DeepSeek's new AI model R1 in January, US technological enterprises have been eager to adopt it, despite governmental attempts to hinder DeepSeek’s progress. Recently, Microsoft, an investor in OpenAI, announced support for DeepSeek’s R1 on its Azure cloud computing platform.

Additionally, leading chip designer Nvidia made the R1 model accessible to its NIM microservice users, stating that the model offers exceptional reasoning capabilities, high inference efficiency, and unmatched accuracy for various tasks.

Wang Peng, a researcher at the Beijing Academy of Social Sciences, observed, "While the US views AI primarily as a tool for maintaining its dominance, DeepSeek is oriented towards enhancing humanity, utilizing open-source technology and promoting global accessibility. China's vision for AI development appears increasingly unyielding, driven not only by financial investments but also by a blossoming culture of collaboration and innovation."

On Saturday, Huawei's cloud division joined forces with Chinese AI firm SiliconFlow to launch and activate DeepSeek's R1/V3 inference service based on Huawei's Ascend cloud platform. They remarked, "This partnership will ensure the model operates reliably in large-scale environments, satisfying commercial deployment requirements."

Liu Zhiyuan, an associate professor in computer science and technology at Tsinghua University, noted that DeepSeek is demonstrating that the technological gap between the US and China in AI development has substantially narrowed. However, Liu cautioned that DeepSeek still confronts significant challenges as it intends to expand its model and service capabilities, noting that AI technologies are rapidly advancing.

DeepSeek, AI, Innovation