The recent rise of Deepseek and Alibaba’s Qwen artificial intelligence platforms has been a hot topic among technology experts and investors in China and the US. However, specific information about China’s engineering teams remains lacking.
An Asia Times analysis using public information on the background of the two chatbot developers may help to draw more clearly how Deepseek and Qwen emerged.
Based on published research papers and media reports, the engineering teams of Deepseek and Qwen will not collaborate or overlap.
The only connection between the duo is that in a paper on January 22 this year, Deepseek researchers stated that Qwen2.5 and Meta llamas had “distilled” to develop Deepseek-R1. The launch of the Deepseek-R1 caused a slump in the US stock market in late January.
Some analysts believe that open source AI DeepSeek used “knowledge distillation” to extract data from Openai’s ChatGPT and train AI models. However, no conclusive evidence has been made public.
According to a paper from January 22, Deepseek-R1 has 16 core contributors, some of which are directly related to Microsoft Research Asia (MSRA) and China Computer Federation (CCF).
Core contributor Yu Wu was overseen by Ming Zhou of MSRA in Beihang University’s doctoral program from 2014 to 2019. He was a full-time intern at MSRA from 2013 to 2019, and in 2019 he participated as an associate researcher.
Core contributor Dayago was also overseen by Minzo in his PhD program at Sanyatsen University in Guangzhou from 2018 to 2023. He was tutored by Nandhan from 2020 to 2023 and by Deuutan from 2017 to 2020 of Musla’s natural language computing group. .
Core contributors Zibin Gou and Zhihong Shao, both of whom are co-authored by Zhihong Shao of Tsinghua University, with Nan Duan of MSRA.
Another Deepseek contributor, Zhenda Xie, was advised by a prominent MSRA scientist at Tsinghua University from 2018 to 2023 from saving Guo. He worked as a research intern at MSRA at the same time.
Zhou Ming and Nan Duan still work at MSRA. In 2016 and 2018, they jointly led the CCF Computer Terminology Approval Working Committee. Zhou is currently the vice president of CCF.
Of course, the connection between Deepseek’s team and MSRA researchers does not mean that MSRA has stakes in the Hangzhou-based company. However, if the US puts more pressure on China’s technology sector, MSRA may have to suspend its work and internship programs in China.
In early 2023, the MSRA reportedly suspended recruitment of interns from seven Chinese universities and Beijing Post and the University of Electro-Communications due to concerns about its ties with the People’s Liberation Army.
The US-approved Northwest Polytechnic University (NPU) and Harbin Institute of Technology (HIT) are one of these schools.
Damo Academy
Hangzhou-based Alibaba, founded by Chinese tycoon Jack MA, developed a chatbot differently.
Mobilizing cash flow from e-commerce companies, Alibaba founded Damo Academy in 2017 to begin researching AI. Damo stands for discovery, adventure, momentum, and outlook.
The Academy has established an advisory committee of 10 well-known educators and researchers. It established an advisory committee of six from the US and four from Chinese universities.
Alibaba has assigned Aliyun’s Chief Technology Officer, Zinglen Zhou, to run the Damo Academy. Zhou received his PhD in Computer Science from Columbia University in the United States. He was a Microsoft R&D partner for four years before joining Alibaba in July 2016.
Chang Zhou, an algorithm engineer who worked for Qwen’s data processing, joined Damo Academy in 2017. He graduated from Fudan University in 2012 and received his PhD from Peking University in 2017.
Before joining Alibaba, he had already worked with Alibaba software engineers on several projects and co-authored two papers in 2017.
GAO received his PhD from Peking University in 2003 and has published over 30 research papers. He has a project funded by China’s 863 program and the National Natural Science Foundation of China (NSFC).
In March 1986, 200 Chinese scientists proposed the 863 program to then-China leader Deng Xiaoping. The government established the program in November of the same year.
In 2022, Chang Zhou co-authored a research paper with a group of scholars, including Bin Cui from Peking University.
CUI is currently the Associate Dean of Peking University’s Computer Science School and the Associate Director of CCF’s Technical Committee on Databases.
Last July, Chang Zhou decided to take a team of around 10 engineers to join bytedance. Alibaba reportedly filed a lawsuit against Zhou, claiming he had no right to join his competitors.
Qwen vs Deepseek
The Alibaba research team appears to have a stronger advisory committee than Deepseek. It also has a more expanded history in data management research.
This may be why Alibaba can build the original AI model, QWEN2.5, but DeepSeek only collects distilled versions of other AI models. This is also the reason why Apple Inc recently partnered with ALIBABA to launch an AI-powered iPhone.
IT columnist Amanda Caswell wrote in a recent article that Qwen2.5 defeated Deep-R1 in all seven tests she set up. She says that QWEN2.5 provides a more structured and easier to read answer, and the DeepSeek-R1 response lacks depth and originality.
From a Beijing perspective, both Qwen and Deepseek are equally essential as China upgrades its industry and helps hedge against US decoupling and sanctions.
According to media reports, Chinese President Xi Jinping said he will soon chair a symposium to raise public sector sentiment. Alibaba co-founder Jack Ma and Deepseek founder Liang Wenfeng will be taking part.
Yong Ji-an is a contributor to the Asia Times. He is a Chinese journalist and specializes in Chinese technology, economy and politics.
Read: Apple Partners Selling AI-powered iPhones in Alibaba and China
