Fan Wang and João da Silva
BBC News
Getty images
When Chatgpt took over the world of artificial intelligence (AI), an inevitable question followed: did she avoided the trouble for China, the biggest American technological rival?
Two years later, a new model of AI in China overthrew this question: can the United States stop Chinese innovation?
For a while, Beijing seemed to grip with his response to Chatgpt, which is not available in China.
Non -impressed users made fun of Ernie, the chatbot of the baidu search engine giant. Then, the versions of Tencent and Bytedance technological companies, which were rejected as fans of Chatgpt – but not as good.
Washington was convinced that he was ahead and wanted to keep it like this. Thus, the Biden administration has increased restrictions prohibiting the export of chips and advanced technology to China.
This is why the launch of Deepseek surprised the Silicon Valley and Le Monde. The company claims that its powerful model is much cheaper than the billions that US companies have spent on AI.
So how is a little -known business – whose founder is praised on Chinese social networks as “AI hero” – realize this?
The challenge
When the United States has prohibited the main manufacturers of chips in the world such as NVIDIA from selling advanced technologies in China, it was certainly a blow.
These chips are essential to build powerful AI models that can perform a range of human tasks, from response to basic requests to solving complex mathematical problems.
The founder of Deepseek, Liang Wenfeng, described the ban on chip as their “main challenge” in interviews with local media.
Long before the ban, Deepseek acquired a “substantial stock” from Nvidia A100 Chips – estimates vary from 10,000 to 50,000 – according to MIT Technology Review.
The main models of AI in the West use around 16,000 specialized chips. But Deepseek says he formed his AI model using 2,000 chips of this type and thousands of lower quality chips – which makes his product cheaper.
Some, including US Tech billionaire, Elon Musk, have questioned this assertion, arguing that the company cannot reveal how many advanced chips he really used in view of the restrictions.
But experts say that Washington’s ban has brought both challenges and opportunities to the Chinese AI industry.
He “forced Chinese companies like Deepseek to innovate” so that they can do more with less, explains Marina Zhang, associate professor at the Sydney University of Technology.
Video surveillance
The founder of Deepseek, Liang Wenfung (R), during a recent government meeting
“Although these restrictions pose challenges, they also stimulated creativity and resilience, aligning on the broader political objectives of China to achieve technological independence.”
The second world economy has invested massively in large technologies – batteries that supply electric vehicles and AI solar panels.
Transforming China into technological superpower has long been the ambition of President Xi Jinping, so Washington’s restrictions have also been a challenge that Beijing has met.
The publication of the new Deepseek model on January 20, when Donald Trump was sworn in as an American president, was deliberated, according to Gregory C Allen, an IA expert at the Center for Strategic and International Studies.
“Timing and the way in which a message is sent – that is exactly what the Chinese government wants everyone to think – that export controls do not work and that America is not the world leader in AI “, explains Mr. Allen, former director of strategy and policy at the US Department of Defense Joint Center for Artificial Intelligence.
In recent years, the Chinese government has fueled AI talents, offering scholarships and research grants and encouraging partnerships between universities and industry.
The National Engineering Laboratory for Deep Learning and other initiatives supported by the State have helped to train thousands of IA specialists, according to MS Zhang.
And China had a lot of brilliant engineers to recruit.
Talent
Take the Deepseek team for example – the Chinese media indicate that it includes less than 140 people, most of whom are what the Internet has proudly declared as a “local talent” of Chinese elite universities.
Western observers have missed the emergence of “a new generation of entrepreneurs who favor fundamental research and long -term technological progress on rapid benefits,” said Zhang.
The best Chinese universities create a “fast -growing AI talent basin” where even managers are often under 35 years old.
“Having grown up during the rapid technological ascent of China, they are deeply motivated by a motivation of autonomy in innovation,” she adds.
Getty images
Hangzhou, where the seat of Deepseek is, also houses other technology giants such as Alibaba
The founder of Deepseek, Liang Wenfeng, is an example – the 40 -year -old woman studied AI at the prestigious University of Zhejiang. In an article on The Tech Outlet 36kr, the people who know him say he looks “more like a geek rather than a boss”.
And the Chinese media describe him as a “technical idealist”-he insists on keeping Deepseek as an open-source platform. In fact, experts also believe that a flourishing open source culture has enabled young start-ups to pool resources and progress more quickly.
Unlike the largest Chinese technology companies, Deepseek prioritized research, which has made it possible to experiment more, according to experts and people who worked in the company.
“The 50 best talents in this area may not be in China, but we can build people like that here,” said Liang in an interview with 36kr.
But experts wonder how Deepseek can go further. Zhang says that “new American restrictions can limit access to American user data, which has a potentially impact on how Chinese and Deepseek models can become global”.
And others say that the United States always has a huge advantage, as, in the words of Mr. Allen, “their enormous quantity of computer resources” – and we do not know how Deepseek will continue to use advanced fleas To continue to improve the model.
But for the moment, Deepseek appreciates his moment in the sun, since most people in China had never heard of it only this weekend.
The new heroes AI
His sudden fame has seen Mr. Liang become a feeling on Chinese social media, where he is applauded as one of the “three heroes of AI” in the southern province of Guangdong, which borders Hong Kong.
The other two are Zhilin Yang, a leading expert at Tsinghua University, and Kaiming, who teaches MIT in the United States.
Deepseek delighted the Chinese Internet before the Lunar New Year, the biggest holidays in the country. This is good news for a besieged economy and a technological industry that is preparing for other prices and the possible sale of the American company in Tiktok.
“Deepseek shows us that it is only if you have the real deal that you will summarize the test of time”, reads a comment from Weibo at best appreciated.
“It’s the best new year gift. I wish our prosperous and strong homeland,” reads another.
A “mixture of shock and excitement, in particular within the open source community”, is the way in which Wei Sun, main analyst of the AI at Counterpoint Research, described the reaction in China.
Getty images
Deepseek’s success was applauded in China during his biggest holidays
Fiona Zhou, a technology worker in the southern city of Shenzhen, says that her social media flow “was suddenly flooded with deep posts yesterday”.
“People call it” the glory of Made-in-China “and say that it is shocked by Silicon Valley, so I downloaded it to see how good it is.”
She asked for “four pillars of (sound) destiny”, or Ba -Zi – like a personalized horoscope which is based on the date and time of birth.
But in his disappointment, Deepseek was wrong. Although she had an in-depth explanation on her “reflection process”, it was not the “four pillars” of his real Ba-Zi.
She says she will always do another blow at work, because she will probably be more useful for such tasks.
