DeepSeek, an AI startup notable for its advanced technology, is leveraging the creativity of young talents to compete with major corporations in the US.
On December 26, the Chinese AI startup DeepSeek introduced its large language model, DeepSeek V3. Despite having limited training resources, the model demonstrated performance comparable to, and even surpassing, some major US competitors like Meta and OpenAI.
This is considered a significant step forward, opening opportunities for China to develop its AI capabilities despite difficulties in accessing advanced chips and financial resources. According to technical reports, recruitment pages, interviews with former employees, and local articles, this breakthrough is the result of a humble founder working alongside a youthful research team.
In 2023, DeepSeek spun off from the investment fund High Flyer-Quant. The founder of High-Flyer Quant, Liang Wenfeng, is also the person behind DeepSeek, and he studied AI at Zhejiang University.
In an interview with 36Kr in May 2023, Mr. Liang revealed that DeepSeek prioritizes hiring young employees in the AI field, from fresh graduates to those just starting their careers. This policy clearly demonstrates the company’s stance of valuing ability over experience.
Mr. Liang stated: “Our key technical positions are primarily held by recent graduates or those with only one or two years of experience.”
Among the talents at DeepSeek, Gao Huazuo and Zeng Wangding are particularly recognized for their significant contributions to the research and development of the MLA architecture. Gao graduated from Peking University (PKU) in 2017 with a degree in physics, while Zeng began her master’s studies at the AI Institute of Beijing University of Posts and Telecommunications in 2021.
The profiles of both reflect DeepSeek’s different approach compared to other AI startups, most of which prioritize hiring more experienced researchers or PhDs trained abroad specializing in computer science.
Additionally, DeepSeek’s core team includes Guo Daya, a PhD who recently graduated from Sun Yat-sen University in 2023, along with two other young PhDs, Zhu Qihao and Dai Damai, both from Peking University (PKU).
Furthermore, there is former employee Luo Fuli, one of DeepSeek’s most prominent talents. With a master’s degree from Peking University (PKU), she has been hailed by the Chinese media as an “AI prodigy.”
She drew public attention when news broke that Xiaomi founder Lei Jun had offered a salary of $1.4 million per year to recruit her. DeepSeek’s V3 model was trained in two months, using approximately 2,000 Nvidia H800 chips with a total cost of only $6 million. Andrej Karpathy, one of the founders of OpenAI, remarked that this was a “joke budget,” implying that the cost is extremely low relative to the scale of the project.
Two weeks ago, many users reported that DeepSeek V3 had mistaken itself for . However, experts suggest that an AI model claiming to be something else is not unusual in the field of artificial intelligence, as most current chatbot tools are trained from general data sources.
According to ZNews
