Liang Wenfeng has gone from boyish, bespectacled math whiz to international tech determine in a single day due to DeepSeek’s explosive influence on the synthetic intelligence business.
The millennial math nerd – who reportedly wore a sweater vest, swimsuit jacket and no tie to a gathering with the Chinese language premier final week – is behind DeepSeek, the Chinese language AI startup that shook the worldwide sector with claims it developed a sophisticated mannequin in only a few months at a fraction of the price of US rivals.
Liang, known as China’s Sam Altman, didn’t begin as a Silicon Valley entrepreneur, however reasonably a math geek who launched his personal hedge fund only a few years after graduating from school.
Born in 1985 and raised in Zhanjiang, China, Liang was a straight-A scholar who studied calculus and wrote AI algorithms in his free time, in line with The Wall Road Journal.
A couple of years after graduating from the celebrated Zhejiang College, he launched his personal quantitative hedge fund, Excessive-Flyer, with two pc scientist buddies, the report stated.
The fund used a man-made intelligence algorithm to select shares – and at present manages some $8 billion, making it certainly one of China’s largest quant funds, in line with the Journal. It has not all been easy crusing, although, issuing a public apology to its traders in 2021 for underperforming benchmark indexes.
He was deeply impressed by Jim Simons, the chain-smoking quant genius behind the Lengthy Island agency Renaissance Applied sciences positioned some 8,000 miles away – even penning the introduction to a Chinese language model of a Simons biography.
“At any time when I encounter difficulties at work, I recall Simons’s phrases: ‘There should be a approach to mannequin costs,’” Liang wrote in his introduction to Simons’ biography.
Over the previous 5 years, not less than 5 funds managed by Excessive-Flyer have produced common extra returns of greater than 20% in comparison with market benchmarks, in line with monetary information supplier Simu Paipaiwang.
In 2021, simply earlier than the Biden administration began limiting exports of AI chips, Liang started shopping for hundreds of Nvidia graphics processors with the purpose of stockpiling 10,000, in line with the Monetary Occasions. His colleagues didn’t assume a lot of the facet undertaking.
“Once we first met him, he was this very nerdy man with a horrible coiffure speaking about constructing a ten,000-chip cluster to coach his personal fashions. We didn’t take him severely,” certainly one of Liang’s enterprise companions advised the Monetary Occasions.
“He couldn’t articulate his imaginative and prescient aside from saying: ‘I need to construct this, and it will likely be a sport change.’”
By late 2022, when OpenAI launched ChatGPT, solely a handful of Chinese language firms had greater than 10,000 Nvidia chips available — and Excessive-Flyer was certainly one of them, in line with the Journal.
“It’s like shopping for a piano,” Liang advised Chinese language tech publication 36Kr in 2023, discussing the chip purchases. “Firstly, it’s as a result of you’ll be able to afford it. And secondly, it’s as a result of you may have a bunch of people who find themselves desirous to play music on it.”
In 2023, Liang based DeepSeek, which on Monday claimed it spent about $6 million to coach its superior AI mannequin – a fraction of what OpenAI and Google spent to coach comparable rivals.
“Liang constructed an distinctive infrastructure group that basically understands how the chips labored,” a founder at a rival LLM firm advised the FT. “He took his greatest folks with him from the hedge fund to DeepSeek.”
Not like ChatGPT, DeepSeek’s AI mannequin is open supply, that means anybody can entry it, a call Liang made in an effort to ding main tech companies’ monopoly.
“For technologists, having others comply with your work provides an important sense of accomplishment,” he advised 36Kr final yr. “Open supply is extra of a tradition reasonably than a industrial conduct, and contributing to it earns us respect.”
His staff have known as him a hands-on boss, generally sleeping in a single day within the workplaces alongside colleagues whereas engaged on particular tasks, and placing little effort into his hair and clothes, in line with the Journal.
He has been preserving a low profile, and was reportedly shocked to see DeepSeek blow up in a single day, sources advised the Journal.