Two years ago, when large Chinese technological companies such as Baidu and Alibaba were chasing the progress of Silicon Valley in artificial intelligence with Splashy and new chatbot ads, Deepseek has adopted a different approach. He got cleared on research.
The strategy gave its fruits.
The Chinese start-up has shaken the world of technology with its affirmation of having created a powerful artificial intelligence model that was significantly cheaper to be built compared to the offers of its best funded American rivals.
In the rivalry between China and the United States for the domain of artificial intelligence, Deepseek seemed to come out of nowhere. In fact, in recent years it has risen to the stars in the world of Chinese technology with a path that was far from conventional.
His mission of pursuing the research reflects that of companies like Openai, the Silicon Valley company that marked an American signature on the AI in the autumn of 2022. But the similarities end most.
The origins of Deepseek are in finance, not technological for the good of technology. His parent company, a Chinese Hedge Fund called High-Flyer, has not begun as a laboratory dedicated to the protection of humanity from the AI as to the Open, but as a company that uses the IA to make bets in the Chinese stock market.
The Alta Friggi had thrived by capitalizing on a market dominated by Chinese retail investors, who are known to have jumped into and outside the actions impulsively. In 2021, High-Flyer was under pressure from the regulatory repressions in China on a speculative trade, which the Beijing authorities believed were in contrast with their attempts to keep the markets calm.
So to Alto Frigio he pursued a new opportunity that has aligned himself better with the priorities of the Chinese government: to advance
“We want to do things with greater value and things that go beyond the investment sector, but it has been interpreted badly as speculation on the actions to the Ai,” said the CEO of High-Flyer, Lu Zhengzhe, Chinese State Media in 2023 “We have created a new team independent of the investment, which is equivalent to a second start-up.”
Deepseek was born. As with many other Chinese start-ups, Deepseek has arrived in a consolidated market with a different commercial approach.
The latest Deepseek artificial intelligence model is believed to be almost powerful as American rivals but much more efficient. His success suggests that the protagonist of the AI of Silicon Valley has reduced. The turning point of Deepseek, despite Washington’s efforts to limit Chinese access to the advanced chips for the IA, raises questions about how effective these long -term checks are, although the founder of Deepseek has recognized that the restrictions on the chip are A limitation.
Deepseek was not based on the creation of products to the revolts of consumption for revenue and only this month has released its first chatbot, which allows anyone to generate text and photos with simple commands. Instead, the company used the money that the high-flyers gained from the equity trade to the ambitious research of the Bankroll. The approach distinguishes it from the US rivals, which in the end are consumer technological companies.
This unconventional approach also allowed Deepsek to evade the rigid regulations that the Chinese government has placed on the use of artificial intelligence by the public. Because his goal was research and sale to companies that use his model – and, until the release of his chatbot this month, not to consumers’ applications – his first works have not triggered the same restrictions as the government.
Deepseek is managed by his CEO, Liang Wenfeng, a thin and besmbered engineer who studied at the University of Zhejiang in the eastern city of Hangzhou. He repeatedly said in the few interviews that he gave to Chinese media that to reach American innovation, Chinese companies must put research before profits. Deepseek and High-Flyer did not respond to requests for comment.
What Chinese technological companies “lack innovation are certainly not capital, but a lack of trust and knowledge on how to organize a high density of talents to obtain effective innovation”, said in an interview widely widespread with the Chinese technological outlet 36kr.
Those who worked with Mr. Liang describes him as a manager capable with a deep technical background, according to interviews and public finances.
“It is certainly an underp,” said Zihan Wang, a computer engineer who worked on a previous deep model, referring to a type of introspective personality of the Myers-Briggs test, a popular personality test among young people in China. “The stars are really good researchers and have the will to explore,” said Wang. “It’s not one of those people who want to control everything.”
Mr. Liang was not too annoyed by details such as the timing of the project and occasionally sent stimulating research questions to the entire team of researchers, Wang said. But above all, Mr. Liang seemed pushed to advance the technology and was not focused on profits.
Unlike many Chinese companies, which tend to focus on taking programmers, Liang has gained the reputation to employ people outside the calculation. Major poets and humanities of the best Chinese universities on the Deepseek staff train the model to write classic Chinese poems and ACE questions taken from the admission exam to the country’s difficult college.
“Most of the team graduated in the best universities in China,” said Yineng Zhang, a main basic software engineer of San Francisco who works on Sglang, a project that is not part of Deepseek that helps people to base themselves on Deepseek system. “They are very intelligent and very young.”
For years, Chinese technological companies have opened the way to artificial intelligence applications used in artificial vision, such as facial recognition. But the release of chatgpt by Openi has prompted a reckoning. When no Chinese society immediately issued something comparable, many concluded that American companies had an advantage in advance
In China, the IT were determined to demonstrate that they could compete. In 2023, many companies in China published their large language models, the technology that supports chatbots as a chatgpt.
But making advanced models would require the use of a large number of chips that would cost hundreds of millions of dollars.
Even the high flight was spending. By 2021, it was one of the handfuls of Chinese companies that had been able to accumulate more than 10,000 advanced NVIDIA chips.
Yet Deepseek’s search gave him a surprising advantage. Last year, he drastically reduced the prices that charged the developers who build applications using his model, causing price war with larger rivals.
Wang, the engineer who previously worked in Deepseek, said there were few discussions on commercial applications for the technology they were building. Instead, he said, the company focused on creating an artificial intelligence system that could be used by a series of people for many purposes.
“During my period there, we didn’t talk much about how we do money,” Wang said. “They only focused on creating an excellent foundation model.”
A crucial part of Deepseek’s popularity is that he made his developers public. This type of information sharing, called Open Source, was a milestone of the development of computer software, internet and now artificial intelligence.
In the United States, researchers and artificial intelligence entrepreneurs have long followed the progress of Deepseek technology. Last year, the company transformed its head when it released systems designed to generate its computer programs.
A new challenge for the company could arrive with its new high profile. On the same day he released R1, the model behind his new chatbot, last week, Mr. Liang appeared in a round discussion with Li Qiang, Chinese premier.
The sudden popularity of Deepseek pushed him to the center of the efforts of the Chinese Communist Party to stimulate innovation, and this could prove difficult to manage, said Jimmy Goodrich, senior consultant for technological analysis at the Rand Corporation, a Think Tank financed at the federal level. “It’s a great situation for Deepseek. I’m sure they weren’t on the five -year government of the government, “he said.
“Can they maintain this carefree vision chaotic when both the party and the world are looking at?”
Zixu Wang Research contributed by Hong Kong.