DeepSeek, the Rising Star in AI, Relies on Young Talent to Challenge US Tech Giants
DeepSeek, a Chinese AI start-up, has been making waves in the tech industry with its innovative approach. The company recently unveiled its DeepSeek V3 large language model, showcasing remarkable performance that rivals or even surpasses that of major US competitors like Meta Platforms and OpenAI. This achievement is particularly noteworthy as it demonstrates China's potential to excel in AI despite facing limitations in resources and funding.
The driving force behind DeepSeek's success lies in its team of "young geniuses," as revealed by insiders and Chinese media reports. The company's founder, Liang Wenfeng, a former AI student at Zhejiang University, leads a group of talented individuals who are either fresh graduates or early in their AI careers. This unconventional hiring strategy prioritises ability over experience, setting DeepSeek apart from other local AI firms.
Among the standout members of DeepSeek is Gao Huazuo, a physics graduate from Peking University, and Zeng Wangding, pursuing a master's degree at the AI Institute of Beijing University of Posts and Telecommunications. These individuals, along with other key team members, have been instrumental in driving key innovations in the research of the MLA architecture, showcasing the company's commitment to nurturing young talent.
DeepSeek's groundbreaking V3 model was developed using a fraction of the resources typically employed by its competitors. Trained in just two months using less powerful Nvidia H800 chips and a budget of only $6 million, the model incorporates cutting-edge training architectures and techniques, including Multi-head Latent Attention and DeepSeekMoE. This cost-effective approach has drawn praise from industry experts, highlighting DeepSeek's ability to achieve remarkable results on a limited budget.
Liang Wenfeng, the enigmatic founder of DeepSeek, is described as a reserved yet intuitive leader with a keen eye for technical detail. Former employees speak highly of Liang's mentorship style, where he guides his team members through suggestive phrases rather than direct commands.
DeepSeek unveils its DeepSeek V3 model, showcasing performance on par with US tech giants.
The company's emphasis on hiring young talent over experienced professionals sets it apart in the industry.
Key team members, including Gao Huazuo and Zeng Wangding, drive innovations in AI research.
Source: SCMP