Liang Wenfeng went to the DeepSeek's explosive impact on the artificial intelligence industry overnight to the Global Tech from a boyish and eye -like Math Whiz.
Millennial mathematics nerd -according to the conveyed part I wore a sweater vest, a suit jacket, and no tie Last week, a meeting with the Chinese Prime Minister is behind Deepseek, a Chinese AI startup that shook a global sector that developed an advanced model in a half of a few months of US rivals.
Rian, called Sam Altoman in China, did not start as an entrepreneur of Silicon Valley, but was a mathematical geek who started his hedge fund a few years after graduating from college.
Born in 1985, Rian, who grew up in a Chinese country, was a straight student who studied microscopic and wrote AI algorithms during the free time. According to the Wall Street Journal.
A few years after graduating from the prestigious University of Zhijiang, he launched his own quantitative hedge fund, HIGHFLYER, with his two computer scientists.
The fund is one of the largest quantum funds in China, according to the journal, which uses artificial intelligence algorithms and today has managed about $ 8 billion, according to the journal. However, not everything is sailing smoothly, and in 2021 it issued an apology for investors for a low -performance benchmark index.
He was deeply inspired by Jim Simmons, a genius with a chain smoking amount behind Renaissance technology, about 8,000 miles away.
“Every time you encounter difficulties in your workplace, you remember Simmons's words:” There must be a way to model prices, “Liang wrote in the introduction of Simons's biography.
According to the financial data provider Simu Paipaiwang, at least five funds managed by HIGH-FLYER have produced an average over 20 % over 20 % compared to market benchmarks in the past five years.
In 2021, just before the Biden administration began to restrict the export of AI chips, Liang began purchasing thousands of NVIDIA graphic processors with the aim of stockpiling 10,000. According to Financial Times。 His colleagues did not think about side projects a lot.
“When we met him for the first time, he was a very geeky man with a terrible hairstyle talking about building a 10,000 chip cluster to train his model. Liang's business partner. One told Financial Times.
“He wants to build this, but it will change the game.”
By the latter half of 2022, when Openai launched Chatgpt, when Openai released ChatGpt, only a handful of Chinese companies had more than 10,000 NVIDIA chips at hand.
“It's like buying a piano,” Liang told the Chinese Tech Publication 36kr in 2023 and discussed the purchase of chips. “First of all, you can afford it, because you have a group of people who want to play music with it.”
In 2023, Liang established DeepSeek. It claimed on Monday that it spent about $ 6 million to train an advanced AI model.
“Liang has built an exceptional infrastructure team that really understands how the chips work,” said the founder of a rival LLM company. “He took the best people from hedge fund to deep seek with him.”
Unlike Chatgpt, Deepseek's AI model is open source. In other words, it means that anyone can access it.
“For engineers, following your work to others gives a great sense of accomplishment,” he said last year. “Open source is more like a culture than a commercial behavior, and contributing to it is paid to us.”
According to the journal, his employees may call him a practical boss and sleep with his colleagues overnight with his colleagues while working on special projects.
He told the journal that he was surprised to see Deepseek exploding overnight, maintaining an inconspicuous thing.
