BEIJING: Liang Wenfeng, the 39-year-old founder of Chinese AI startup DeepSeek, has in the matter of weeks become the face of China’s tech industry and its hope of overcoming an ever-tightening noose of export controls imposed by the United States.
Liang had kept an extremely low profile until Jan. 20, when he was one of nine individuals asked to give a speech at a closed-door symposium hosted by China’s Premier Li Qiang.
He gave two rare media interviews to Chinese media outlet Waves last year and in 2023, but apart from that has stayed mostly out of the public eye. DeepSeek did not respond to a request for an interview.
At the symposium, the millennial’s youthful appearance contrasted with the grey-haired academics, officials and state-owned conglomerate heads sat around him, pictures and video published by Chinese broadcaster CCTV showed.
But the fact Liang was invited to share his opinions on Chinese government policy highlights Beijing’s recognition of DeepSeek’s role in potentially upending the global AI order, in China’s favour.
DeepSeek launched a free AI assistant last week that the firm says uses less data at a fraction of the cost of current services, triggering a global selloff in tech stocks.
Last year, Baidu CEO Robin Li spoke at a similar symposium chaired by the Chinese premier.
Li, who announced China’s first ChatGPT rival in March 2023, said in an interview that same year China would never recreate Microsoft-backed OpenAI’s success and that Chinese firms should focus on applying existing AI models for commercial purposes.
Under Liang’s leadership, DeepSeek deliberately avoided app-building.
Instead, it concentrated research talent and resources on creating a model that could match, or better OpenAI, and it hopes in the future to continue focusing on cutting-edge models that will be used by other companies to build consumer and enterprise-facing AI products.
Liang’s approach stood out in a Chinese tech industry that was used to taking innovations from abroad, from smartphone apps to electric vehicles, and quickly scaling them up, often much faster than the countries where the inventions were first made.
“China’s AI can’t be in the position of following forever. We often say that there is a gap of one or two years between China’s AI and the United States, but the real gap is the difference between originality and imitation,” Liang said in an interview with Waves in July last year.
Liang’s interviews reveal a belief that China’s tech industry had reached a crossroads where it lacked the confidence but not the capital needed to engage in fundamental R&D breakthroughs.
‘Curiosity and desire to create’
“In the past thirty years, (China’s tech industry) has only emphasized making money, and ignored innovation. Innovation is not solely driven by business, it also needs curiosity and a desire to create,” he said in July.
DeepSeek has taken the decision to make all its models open-source, unlike its U.S. rival OpenAI. In open-source models, the base code is publicly available for any developer to use and modify at will.
Liang’s interviews reveal he has bought into the open-source culture that U.S. tech insiders previously argued was one reason the U.S.’s Silicon Valley held an edge over its Chinese counterparts.
“Even if OpenAI is closed-source, it cannot stop others from catching up…Open-source is like a cultural practice, rather than a business practice…a company that does this will have soft power.”
Liang grew up in the southern province of Guangdong, which during the eighties and nineties led the country in adopting market capitalism. Liang said back then he was surrounded by people who valued starting a business over studying, but he was more academically inclined.
He would go on to enrol in the elite Zhejiang University at age 17, majoring in Electronics and Communication Engineering before pursuing a master’s degree in Information and Communication Engineering, which he completed in 2010.
Liang then co-founded a quantitative hedge fund in 2015, which uses complex mathematical algorithms for trading as opposed to human analysis.
Alibaba releases AI model it claims surpasses DeepSeek-V3
The fund’s portfolio totalled over 100 billion yuan ($13.79 billion) by the end of 2021 but in April 2023, it announced on its WeChat account that it would expand its remit beyond the investment industry and concentrate resources to “explore the essence of AGI”.
DeepSeek was created a month later.
OpenAI defines AGI (Artificial general intelligence) as autonomous systems that surpass humans in most economically valuable tasks.
DeepSeek’s employees are mainly graduates and PhD students from China’s top universities, whom Liang believes prefer to work for DeepSeek because it is tackling the biggest challenges in AI.
“What attracts the best talent is obviously the solving of the world’s hardest problems,” he said in July.
“Our goal is still to go for AGI.”
Comments
Comments are closed for this article.