Liang Wenfeng has gone from boyish, bespectacled math whiz to global tech figure overnight thanks to DeepSeek’s explosive impact on the artificial intelligence industry.
The millennial math nerd – who reportedly wore a sweater vest, suit jacket and no tie to a meeting with the Chinese premier last week – is behind DeepSeek, the Chinese AI startup that shook the global sector with claims it developed an advanced model in just a few months at a fraction of the cost of US rivals.
Liang, called China’s Sam Altman, didn’t start as a Silicon Valley entrepreneur, but rather a math geek who launched his own hedge fund just a few years after graduating from college.
Born in 1985 and raised in Zhanjiang, China, Liang was a straight-A student who studied calculus and wrote AI algorithms in his free time, according to The Wall Street Journal.
A few years after graduating from the prestigious Zhejiang University, he launched his own quantitative hedge fund, High-Flyer, with two computer scientist pals, the report said.
The fund used an artificial intelligence algorithm to pick stocks – and today manages some $8 billion, making it one of China’s largest quant funds, according to the Journal. It has not all been smooth sailing, though, issuing a public apology to its investors in 2021 for underperforming benchmark indexes.
He was deeply inspired by Jim Simons, the chain-smoking quant genius behind the Long Island firm Renaissance Technologies located some 8,000 miles away – even penning the introduction to a Chinese version of a Simons biography.
“Whenever I encounter difficulties at work, I recall Simons’s words: ‘There must be a way to model prices,’” Liang wrote in his introduction to Simons’ biography.
Over the past five years, at least five funds managed by High-Flyer have produced average excess returns of more than 20% compared to market benchmarks, according to financial data provider Simu Paipaiwang.
In 2021, just before the Biden administration started limiting exports of AI chips, Liang began buying thousands of Nvidia graphics processors with the goal of stockpiling 10,000, according to the Financial Times. His colleagues didn’t think much of the side project.
“When we first met him, he was this very nerdy guy with a terrible hairstyle talking about building a 10,000-chip cluster to train his own models. We didn’t take him seriously,” one of Liang’s business partners told the Financial Times.
“He couldn’t articulate his vision other than saying: ‘I want to build this, and it will be a game change.’”
By late 2022, when OpenAI launched ChatGPT, only a handful of Chinese companies had more than 10,000 Nvidia chips on hand — and High-Flyer was one of them, according to the Journal.
“It is like buying a piano,” Liang told Chinese tech publication 36Kr in 2023, discussing the chip purchases. “Firstly, it’s because you can afford it. And secondly, it’s because you have a group of people who are eager to play music on it.”
In 2023, Liang founded DeepSeek, which on Monday claimed it spent about $6 million to train its advanced AI model – a fraction of what OpenAI and Google spent to train comparable rivals.
“Liang built an exceptional infrastructure team that really understands how the chips worked,” a founder at a rival LLM company told the FT. “He took his best people with him from the hedge fund to DeepSeek.”
Unlike ChatGPT, DeepSeek’s AI model is open source, meaning anyone can access it, a decision Liang made in an effort to ding major tech firms’ monopoly.
“For technologists, having others follow your work gives a great sense of accomplishment,” he told 36Kr last year. “Open source is more of a culture rather than a commercial behavior, and contributing to it earns us respect.”
His employees have called him a hands-on boss, sometimes sleeping overnight in the offices alongside colleagues while working on special projects, and putting little effort into his hair and clothing, according to the Journal.
He has been keeping a low profile, and was reportedly surprised to see DeepSeek blow up overnight, sources told the Journal.
Credit: Source link
