How Chinese company DeepSeek developed a top AI reasoning system despite US sanctions
Technology giants such as Alibaba and ByteDance as well as a few startups with deep pockets dominate the Chinese AI market, making it difficult for small and medium-sized businesses to compete. Rare is a company such as DeepSeek that has no plans to raise money.
Zihan Liang, a former DeepSeek employee told MIT Technology Review he was able to access abundant computing resources at DeepSeek and experiment freely, “a luxury few new graduates could get at any other company.” “We
need to consume twice as much computing power in order to achieve the same result. This could require up to four-times more computing power when combined with data inefficiency gaps. “Our goal is to continually close these gaps”, he said. But DeepSeek managed to find ways to reduce the amount of memory used and increase calculation speed without compromising accuracy. Wang says that Liang is still deeply involved in DeepSeek research, and runs experiments with his team. Wang states that “the whole team has a collaborative culture, and is dedicated to hardcore research.” Alibaba Cloud released more than 100 open-source AI model, which support 29 languages, and cater to different applications including coding and math. Startups like Minimax, 01.AI and others have also open-sourced models. According to a paper published last year by China Academy of Information and Communications Technology (a state-affiliated institute), the number of AI models for large languages worldwide reached 1,328. Of these, 36% originated in China. China is now the second largest contributor to AI after the United States. Thomas Qitong Cao is an assistant professor at Tufts University who specializes in technology policy. He says that this generation of Chinese researchers are very much into open-source culture, because they gain so much from it.
