DeepSeek was founded by Liang Wenfeng, a 39-year-old computer science expert from Zhanjiang, Guangdong. Liang’s journey into AI innovation stems from a strong educational foundation and a passion for technology. Born into a family of educators, Liang’s father, a primary school teacher, fostered his early curiosity for learning.
Beijing, China – In a dramatic shake-up of the artificial intelligence world, Chinese startup DeepSeek has emerged as a formidable contender against industry leaders like OpenAI. The company’s latest open-source AI model, DeepSeek-R1, has outperformed OpenAI’s GPT-4 in critical areas, including mathematical reasoning, cost-efficiency, and openness.
Despite operating with just $6 million in funding, DeepSeek has managed to challenge established AI players, underscoring a significant shift in the global AI landscape.
The Visionary Behind DeepSeek
DeepSeek was founded by Liang Wenfeng, a 39-year-old computer science expert from Zhanjiang, Guangdong. Liang’s journey into AI innovation stems from a strong educational foundation and a passion for technology. Born into a family of educators, Liang’s father, a primary school teacher, fostered his early curiosity for learning.
Liang attended Zhejiang University, where he earned a Bachelor of Engineering in electronic information engineering in 2007, followed by a Master of Engineering in information and communication engineering in 2010. These years cemented Liang’s interest in technology and laid the groundwork for his future endeavors.
Before founding DeepSeek, Liang played a pivotal role at High-Flyer, one of China’s top quantitative hedge funds. He led Fire-Flyer, the firm’s deep-learning research branch, which utilized advanced supercomputers to analyze financial data.
In 2023, Liang made a bold decision to pivot Fire-Flyer’s resources toward creating DeepSeek with the ambitious goal of advancing artificial general intelligence (AGI). Speaking to Chinese tech publication 36Kr, Liang explained, “Basic science research has a very low return on investment. But we’re doing this because we believe in the mission.”
DeepSeek’s research team consists largely of recent graduates from elite Chinese universities like Peking University and Tsinghua University. These young researchers have already gained recognition for their academic contributions, including publishing in top journals and winning international awards.
DeepSeek’s flagship model, DeepSeek-R1, has made waves by outperforming GPT-4 in several benchmarks, particularly in reasoning and cost-efficiency. This achievement is especially notable given the U.S. export restrictions on advanced GPUs, such as Nvidia’s H100, which many believed would hinder Chinese AI progress.
Rather than relying on expensive hardware, DeepSeek has focused on software optimization and open-source collaboration, employing innovative techniques like Multi-head Latent Attention (MLA) and Mixture-of-Experts (MoE). These advancements have reduced the model’s computing requirements to just one-tenth of what comparable models, such as Meta’s Llama 3.1, demand.
DeepSeek’s decision to release its models as open source has won it widespread support from the global AI community. By sharing its breakthroughs, the company is attracting contributors and solidifying its role as a leader in collaborative AI development.
Challenging the U.S. AI Dominance
DeepSeek’s success highlights a growing shift in the ongoing tech rivalry between the U.S. and China. While U.S. export controls aim to limit China’s access to advanced computing resources, companies like DeepSeek are demonstrating that innovation doesn’t solely depend on brute-force hardware.
Join our Channel...