Tech giants like Alibaba and Bytedance, as well as a handful of startups with deep investors, dominate China’s AI space, making it difficult for small and medium-sized companies to compete. It’s rare to see a company like Deepseek without plans to raise funds.
Former Deepseek employee Zihan Wang said MIT Technology Review He had access to abundant computing resources and was given freedom to experiment when working at Deepseek.
In a July 2024 interview with Chinese media outlet 36KR, Liang said an additional challenge Chinese companies face under Chip sanctions is that AI engineering technology is less efficient. “We (most Chinese companies) have to spend twice as much computing power to achieve the same results. Combined with the data efficiency gap, this could mean up to four times more computing power is needed. Our goal is to continue to close these gaps,” he said.
However, DeepSeek found a way to reduce memory usage and speed up calculations without sacrificing much accuracy. “The team loves turning hardware problems into innovation opportunities,” Wang said.
Liang himself is deeply involved in Deepseek’s research process and runs experiments with the team. “The entire team shares a collaborative culture and commitment to hard-core research,” Wang said.
In addition to prioritizing efficiency, Chinese companies are increasingly embracing open source principles. Alibaba Cloud has launched over 100 new open source AI models, supporting 29 languages and embracing a variety of applications including coding and mathematics. Likewise, startups like Minimax and 01.ai have opened up their models.
According to a white paper published last year by China Information and Communications Technology, a state-run research institute, the number of AI large language models worldwide has reached 1,328, with 36% originating in China. This positions China as the second largest contributor to AI behind the United States.
“This generation of young Chinese researchers identifies strongly with the open source culture because they see many benefits,” said Thomas Qitong Cao, assistant professor of technology policy at Tufts University..
