How Deepseek went from shares to AI Star

Two years ago, when renowned Chinese technology companies such as Baidu and Alibaba were chasing Silicon Valley's advances in artificial intelligence with splashed ads and new chatbots, Depseek adopted a different approach. He focused on the investigation.

The strategy was worth it.

The new Chinese company has shaken the technological world with its claim that it created a powerful model of ai that was significantly cheaper to build than the offers of its best funded US rivals.

In the rivalry between China and the United States on the dominance of artificial intelligence, Depseek seemed to get out of nowhere. In fact, he has shot through the technological world of China in recent years with a path that was anything but conventional.

His mission of pursuing the investigation reflects that of companies such as Openai, the firm of Silicon Valley that marked an American firm on ai in the autumn of 2022. But the similarities end mainly there.

The Origins of Deepseek are in finance, not in technology for the good of technology. His parent company, a Chinese coverage fund called High-Flyer, did not start as a laboratory dedicated to safeguarding humanity of ai as an open ai, but as a business that uses ai to make bets in the Chinese stock market.

High-Flyer had prospered by capitalizing a market dominated by China's retail investors, known for entering and leaving the shares impulsively. In 2021, High-Flyer was pressed by regulatory repressions in China in speculative trade, that Beijing authorities considered that he disagreed with his attempts to keep the market calm.

So High-Flyer looked for a new opportunity that said he lined up better with the priorities of the Chinese government: ai advanced

“We want to do things with greater value and things that go beyond the investment industry, but it has misunderstood as speculation of ai actions,” said High-Flyer executive director Lu Zhengzhe, to the Chinese state media in 2023. “We have created a new independent investment team, which is equivalent to a second company.”

Deepseek was born. As with many other new Chinese companies, Depseek arrived at an established market with a different commercial approach.

It is believed that Deepseek's latest model for artificial intelligence is almost as powerful as US rivals but much more efficient. His success suggests that Silicon Valley's lead has been reduced. Deepseek's advance, despite Washington's efforts to limit Chinese access to the advanced chips necessary for ai, ask questions about how effective these long -term controls can be, although the Deepseek founder has recognized that the restrictions of Chips are a limitation.

Deepseek did not trust to make consumer -oriented products for income, and only this month launched their first chatbot, allowing anyone to generate text and photos with simple commands. Instead, the company used the money that the Alto Flyer won from the trade trade to the ambitious research of the investigation. The approach distinguishes it from US rivals, all of which are, ultimately, consumer technology companies.

This unconventional approach also allowed Deepseek to put aside the strict regulations that the Chinese government has put in public use. Because his approach was to investigate and sell to companies that use their model, and, until the launch of their chatbot this month, not the applications of consumers, their early work did not trigger the same government restrictions.

Deepseek is led by its executive director, Liang Wenfeng, a thin engineer and with glasses he studied at the University of Zhejiang in the eastern city of Hangzhou. He said repeatedly in the few interviews he has granted to Chinese media that to catch up with US innovation, Chinese companies must investigate profits. Deepseek and High-Flyer did not respond to comments requests.

What Chinese technology companies “lack innovation is certainly not a capital, but a lack of confidence and knowledge about how to organize a high talent density to achieve effective innovation,” he said in a widely circulated interview With Chinese technology 36kr.

According to interviews and public accounts.

“It is definitely an INTP,” said Zihan Wang, a computer engineer who worked in a previous Deepseek model, referring to an introspective personality type of the Myers-Briggs test, a popular personality test among young people in China. “INTP are really good researchers and have the will to explore,” said Wang. “He is not one of those people who wants to control everything.”

Mr. Liang was not too upset with details such as the project's deadlines, and occasionally sent stimulating research questions to the entire team of researchers, Wang said. But above all, Mr. Liang seemed driven to advance technology and not He focused on profits.

Unlike many Chinese companies, which tend to focus on the hiring of programmers, Mr. Liang has earned the reputation of using people from outside the computer science. Poets and specialties of humanities from the main universities of China in Depseek staff train the model for classical Chinese poetry and questions taken from the difficult entrance exam to the University of the country.

“The majority of the team graduated from the best universities in China,” said Yineng Zhang, a main software engineer based in San Francisco who works at the SGLANG, a project that is not part of Deepseek that helps people to build About the Deepseek system. “They are very intelligent and very young.”

For years, Chinese technology companies were pioneers in artificial intelligence applications used in computer vision, such as facial recognition. But the launch of Openai chatgpt caused a calculation. When no Chinese company immediately launched somewhat comparable, many concluded that US companies had an advanced advantage

In China, computer scientists were determined to demonstrate that they could compete. In 2023, many companies in China published their own large language models, the technology that supports chatbots such as Chatgpt.

But making advanced models would require using a large number of chips that cost hundreds of millions of dollars.

High-Flyer was also spending. By 2021, it was one of the few Chinese companies that had been able to store more than 10,000 advanced Nvidia A100 chips.

However, Depseek's investigation gave him a surprising advantage. Last year, he drastically reduced the prices that developers charged that they create applications using their model, which caused a price war with larger rivals.

Mr. Wang, the engineer who previously worked at Deepseek, said there was little discussion about commercial applications for the technology they were building. Instead, he said, the company focused on making an ai system that could be used by a variety of people for many purposes.

“During my time there, we don't talk much about how we earn money,” Wang said. “They simply focused on making a great base model.”

A crucial part of Depseek's popularity is that it has made public the work of its developers. This type of information exchange, called open source, has been an cornerstone of the development of computer software, internet and now artificial intelligence.

In the United States, researchers and businessmen of ai have long followed the progress of Depseek technology. Last year, the company caught attention when it launched systems designed to generate its own computer programs.

A new challenge for the company can come with its new high profile. The same day that R1 launched, the model behind its new chatbot, last week, Mr. Liang appeared in a round table with Li Qiang, China's prime minister.

Deepseek's sudden popularity has led him to the center of the efforts of the Chinese Communist Party to stimulate innovation, and that could be difficult to administer, said Jimmy Goodrich, principal technology analysis advisor for Rand Corporation, a group of experts financed by The federal government. “It is a great situation for Deepseek. I am sure they were not in the five -year plan of the government, “he said.

“Can you maintain this expensive chaotic vision when the party and the world are looking?”

Zixu Wang Research contributed from Hong Kong.