Techonology

How deep went from stock trader to AI star

Two years ago, when large-nam Chinese technology companies like Baidu and Alibaba were chasing the progress of Silicon Valley in artificial intelligence with clear announcements and new chatbots, Deepsek took a different view. It made zero in research.

The strategy paid.

The Chinese Start-up has shocked the technical world with its claim that it created a powerful AI model that was much cheaper than the offerings of its better funded American rivals.

In rivalry between China and the United States on the domination of Artificial Intelligence, Deepsek was coming out from anywhere. In fact, in recent years it has touched the sky with a path through the world of China’s technology that was anything traditional.

Its mission to carry forward research mirrors of companies such as Openai, Silicon Valley firm, which marks an American signature on AI in the fall of 2022. But similarities mostly end there.

Deepsek originates in finance, not technology for technology. Its native company, a Chinese hedge fund, called a high-flag, devoted to use AI to use AI, not as a laboratory dedicated to the protection of humanity from AIs like Open AI. , Which uses AI to bet in the Chinese stock market.

The High-Flire flourished by capitalizing on the market dominated by China’s retail investors, which are known for jumping and out of shares. In 2021, the high-player pressured itself by regulatory rifts in China on speculative trade, which the authorities in Beijing felt that their efforts were on obstacles with their efforts to keep the markets calm.

So the high-flag gave a new opportunity, stating that it formed better with the priorities of the Chinese government: Advanced AI

“We want to do things with more prices and things that are beyond the investment industry,” Lu Zhenggi, Chief Executive Officer of High-Fliers,, in 2023, but it is misunderstood as AI stock speculation. . ” A new team independent of investment, which is equal to another start-up. ,

Deepsek was born. With several other Chinese start-ups, Deepsek came to an established market with a separate business approach.

The latest model of Deepsac for Artificial Intelligence is considered almost powerful as American rivals, but far more efficient. Its success shows that the AI ​​lead of Silicon Valley has shrunk. Despite efforts to limit the Chinese access to the advanced chips required by Washington, Washington’s success, Washington raises the question of how those controls can be effective for a long time – although the founder of the Deepsek has admitted that the chip has admitted that the chip The ban is a limit.

Deepsek did not rely on creating consumer-supporting AI products for revenue, and only this month released its first chatbot, which allows anyone to generate lessons and photos with a simple command. Instead, the company used the money that high-flagged made from stock trading for Bancolol ambitious research. The approach separated it from American rivals, which are all eventually consumer technology companies.

This unconventional approach also allowed Deepsek to sideline stringent rules, which the Chinese government has put on the use of AI by the public. Because its focus was research and sales on businesses that use its model – and, until the release of its chatbot of this month, not consumer applications – its initial work did not trigger the same government restrictions.

The Deepsek is run by its CEO, Leiang Wenfeng, a thin, Bangar Engineer, who studied at Jhejiang University in the eastern city of Eastern city of Hangzo. He repeatedly said in some interviews that he has given to Chinese media that Chinese companies will have to do research before profits to catch American innovation. Deepsek and Hi-Flor did not respond to the remarks requests.

Chinese technology companies “innovation reduction is definitely not capital, but there is a lack of confidence and knowledge about how to organize a high density of talent to achieve effective innovation,” he said in A Widely broadcast interview 36kr with Chinese take outlets.

Those who have worked with Mr. Liang describe them as a competent manager with a deep technical background, according to interviews and public accounts.

“He is definitely an INTP,” a computer engineer, Zeehan Wang, said, who worked on a deep model, which mentioned an introspect personality type from the Myers-Bigs Test, which was mentioned There is a popular personality test among people. “INTP is actually good researchers and desire to find out,” said Mr. Wang. “He is not among those who want to control everything.”

Mr. Liang was not very disturbed by details like Project Timeline, and sometimes the entire team of researchers was sent ideal research questions, Mr. Wang said. But mostly, Mr. Leiang seemed motivated to pursue technology and was not focusing on profits.

Unlike many Chinese companies, which focus on hiring programmers, Mr. Leiang has obtained a reputation to provide employment to people outside computing. Poets of top universities of Chinese and big companies of humanities trained the models to write classical Chinese poetry and ace questions taken from the country’s tough college entrance examination.

Yineng Zhang, a prominent software engineer at Besten, San Francisco, said, “Most of the teams graduated from top universities in China, which works on slipung, which is not part of Deepsek, which people have the head of the system of lampsakes. It helps in making. “They are very smart and very small.”

For years, Chinese tech companies pioneered artificial intelligence applications used in computer vision, like facial identity. But Openai’s Chatgate release inspired a recording. When any Chinese company did not release anything immediately, many people concluded that American companies had an edge in advanced AI.

In China, computer scientists were firm to prove that they could compete. In 2023, many Chinese companies released their big language models, the technology that reduces chatbots such as chatbots.

But a large number of chips will need to be used to create advanced models, which will cost hundreds of million dollars.

The high-flag was also spending. By 2021, it was just a handful of Chinese companies that were able to stock more than 10,000 advanced NVidia A100 chips.

Nevertheless, Deepsek’s research gave it a surprising benefit. Last year, it dramatically cuts prices that charged the developers that manufacture applications using their models, indicating value war with large rivals.

Mr. Wang, who previously worked in Deepsak, said that there was little discussion of commercial applications for the technique he was producing. Instead, he said, the company focused on creating an AI system that can be used by a series of people for many purposes.

“During my time, we did not talk much about how we earn money,” said Mr. Wang. “They focus on creating just a great foundation model.”

An important part of Deepsek’s popularity is that it has made public the work of its developers public. Sharing such information, called Open Source, is the cornerstone of the development of computer software, internet and now artificial intelligence.

In the United States, AI researchers and entrepreneurs have long followed the progress of Deepsek’s technology. Last year, the company replaced the heads on releasing the system designed to generate its computer programs.

A new challenge for the company can come up with its new high profile. The same day it released the R1, the model behind its new chatbot last week, Mr. Liang appeared in a round table with China’s premiere Lee Kiang.

Deepsek’s sudden popularity has insisted on the Chinese Communist Party’s efforts to inspire the innovation, and it may prove to be difficult to manage, Jimmy Gudrik said, a senior advisor for technology analysis for Rand Corporation, Funded think tank of a union. “This is a great prediction for Deepsek. I am sure he was not on the five -year plan of the government, he said.

“Can they maintain this chaotic careless vision when both the party and the world are watching?”

Zixu wang Contributed to research from Hong Kong.

T) Hongjo High-Flor AI Fundamental Research Company Limited (T) Lu Zhengze
#deep #stock #trader #star

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *