Didi Vice President Ye Jieping resigned. He was the head of AI at the travel giant and the key scientist who defeated Uber China

Qian Ming from Aofei Temple
Quantum Bit Report | Public Account QbitAI

Ding, ding, Ye Jieping resigned.

Quantum位 learned that the vice president of Didi Chuxing, deputy director of Didi Research Institute, head of Didi AI Labs, tenured professor at the University of Michigan, and IEEE Fellow has resigned.

There are rumors that his next stop is Alibaba. However, people close to Ye Jieping revealed that he will not join Alibaba, but will consider fields with good data foundation and huge potential for AI application.

Some people speculate that the departure of this AI technology expert is related to Didi’s recent adjustments to AI Labs - Didi has internally changed the name of AI Labs to AI Tech, from a "laboratory" to a "technology business department" to strengthen its implementation.

However, some people also said that Professor Ye had hoped to be closer to his hometown in Zhejiang for many years and spend more time with his parents and family.

Within Didi, after Ye Jieping's departure, CTO Zhang Bo will temporarily take over AI Labs, while daily affairs will be handled by Didi's chief intelligent control scientist Tang Jian, who will serve as deputy director of Didi AI Labs.

In any case, everyone inside and outside Didi was surprised by Ye Jieping's departure.

Unlike the "trajectory" of scientists leaving companies in the past two years, although Professor Ye joined as an academic giant, he has always been a practical person in the implementation of technology.

It was precisely because of his joining in 2015 that Didi was able to transform from grabbing orders to dispatching orders, making a key contribution to Didi's victory over Uber.

Later, because He Xiaofei left to start his own business, Ye Jieping took the helm of Didi AI Labs and Research Institute. In five years, he established and trained a high-end technical team. Not only did the business get off the ground, but the academic achievements based on the business also made Didi famous at top global conferences such as KDD and CVPR.

So why did such a meritorious general leave at this time?

Let’s start with Professor Ye Jieping.

Top student at Fudan University, tenured professor at the University of Michigan, Didi war hero

Ye Jieping is from Jinhua, Zhejiang Province. He attended Yongkang No. 1 High School, one of the earliest high schools established in Zhejiang Province with a history of more than 100 years. It is a first-level key high school in Zhejiang Province. Other famous alumni include Hu Weiwu, the father of Loongson.

In 1993, Ye Jieping was admitted to the Department of Mathematics at Fudan University. In the same year, the mathematics discipline of Fudan University was approved by the State Education Commission as a basic discipline talent training base.

After graduating from Fudan University, he went to the National University of Singapore for further studies and stayed abroad for 18 years.

In 1999, after obtaining a master's degree in computer science, Ye Jieping went to the United States to pursue a doctorate in computer science at the University of Minnesota, majoring in machine learning.

At that time, machine learning was still in the cold winter. He later recalled that when he was studying at the University of Minnesota, there was no teacher specializing in machine learning. Because of his interest, he persisted and obtained his doctorate in 2005.

"Machine learning" is the starting point of his scientific research career in academia and also the driving force that helped him become an international academic giant.

After graduating with a Ph.D., he joined Arizona State University and worked there for 9 years and 5 months, rising from assistant professor to associate professor. In January 2015, Ye Jieping was hired as a tenured professor at the University of Michigan.

In his 10 years of teaching and research, he has continuously produced research results in the fields of machine learning, data mining and big data analysis, and has won the best paper awards at KDD and ICML. According to Google Scholar, he has published a total of 448 papers, which have been cited more than 20,000 times.

Along with this, there are also various academic part-time jobs and honors - member of the editorial board of several top international academic journals such as IEEE TPAMI, DMKD and IEEE TKDE, and program committee chair and area chair of several international conferences such as NeurIPs and KDD.

At the same time, Ye Jieping has also trained a large number of talents in the field of artificial intelligence. For example, Liu Ji, the head of Kuaishou's AI platform, was a student he taught when he was teaching at Arizona State University.

The summer vacation of 2015 was a turning point for Ye Jieping from academia to industry. He returned to China to visit his family during the summer vacation and was invited by He Xiaofei, then SVP of Didi, to visit Didi and exchange ideas, where he met Cheng Wei, Liu Qing, Zhang Bo and others.

At that time, Didi was in the process of transitioning its dispatching system from grabbing orders to dispatching orders. Didi invited Ye Jieping to come over not just for communication, but also to prepare some difficult work problems for him.

Although Ye Jieping's research direction is "machine learning and medicine" and he has not been exposed to research on traffic, he still relied on his technical accumulation and delivered a solution in more than 10 hours.

This short trip back to China completely changed Ye Jieping's planned development path. Before this, he had never thought about entering the industry.

The feedback from using technology to solve practical problems had a strong appeal to Ye Jieping. When he returned to the United States, he told his students that he was leaving.

He was inspired by Didi and accepted the olive branch, ending his overseas study and research career, which has been 5 years. During this period, he contributed technical strength to Didi's victory in the war with Uber, promoted and witnessed the implementation and application of Didi's AI technology, and was selected as an IEEE Fellow in 2020.

Applying "AlphaGo" to Didi's dispatching system

When Ye Jieping joined, Didi had just ended its "money-burning war" with Kuaidi and began to build an artificial intelligence team to apply machine learning on a large scale in the travel field, especially in data mining and real-time computing, to compete with Uber China.

Ye Jieping spent his early years at Didi on the "battlefield", using algorithms to make up for Didi's technological shortcomings and promoting a smooth transition from grabbing orders to dispatching orders.

This transformation played a key role in the technical confrontation between Didi and Uber China, and remains the basis for Didi's efficient vehicle dispatch.

In April 2016, on the eve of Didi ending its "war" with Uber China, Didi Research Institute was officially established, with He Xiaofei as the founding dean, Ye Jieping and Gong Fengmin, known as the "godfather of Silicon Valley security entrepreneurship", as vice deans, and AI technology research entered an in-depth stage.

Ye Jieping's research focuses on optimizing Didi's order dispatching and vehicle scheduling systems, while also applying cutting-edge academic research such as reinforcement learning used in AlphaGo to Didi.

Ye Jieping, who had already moved his family back to China, led a team to establish AI Labs in January 2018. Through machine learning theories and methods, he maximized the use of transportation capacity. It is located in Xi'erqi.

Afterwards, he also began to lead a team to study the application of technologies such as image, speech semantics, etc. in Didi's business.

In recent years, Ye Jieping has also become Didi's representative in the application of AI technology, delivering Didi's technological background at various academic and industry conferences.

He also led the researchers of Didi AI Labs to deliver results in the academic field, not only papers but also algorithm competitions.

In April 2019, the face detection DFS algorithm proposed by the Didi AI team and the Beijing University of Posts and Telecommunications team surpassed domestic and foreign technology companies and universities and research institutes in six evaluation results of the Easy, Medium and Hard evaluation subsets of the world's face detection public evaluation set WIDER FACE, and won five first places and one second place.

Now when you call Didi with your mobile phone, you may see a cylindrical camera above the windshield of many cars. This is the "Ji Shi" system launched by Didi, and behind it is the research results of Ye Jieping.

After Ye Jieping left, the next person in charge of Didi's artificial intelligence research department also came from within.

Judging from Didi’s technical executive lineup, it was initially rumored that the person who would take over would be: Zhang Wensong - Didi Chuxing’s senior vice president, head of the cloud platform business group and director of the open source committee.

Before joining Didi, Zhang Wensong worked at Alibaba for nearly seven years, serving as senior technical director of Taobao, chairman of Taobao Technology Committee, vice president of Alibaba, chairman of Alibaba Open Source Committee, and CTO of Alibaba Cloud.

In May 2016, Zhang Wensong joined Didi, which caused a sensation in the industry. Zhengming (Zhang Wensong's nickname) left Alibaba and joined Didi, becoming a hot topic in the technology circle.

Zhang Wensong currently oversees the Infrastructure Department, Operation and Maintenance Department, System Department, Big Data Architecture Department, IT & Machine Learning Department, Data Platform and Data Governance Department, Didi Cloud, etc., which is the foundation of the technology engine.

But now, Quantum位 has learned that AI Labs will be taken over by Didi Chuxing CTO Zhang Bo, and another IEEE Fellow Tang Jian will directly succeed Ye Jieping in terms of specific business matters.

Who is Tang Jian?

Tang Jian was previously the chief scientist of Didi Intelligent Control, an IEEE Fellow, and an ACM Distinguished Scientist, leading Didi's research and development in embedded AI and computer vision.

He graduated from Beijing University of Posts and Telecommunications with a bachelor's degree and received his Ph.D. in computer science from Arizona State University in 2006. He then taught at Syracuse University.

He has published more than 140 academic papers in top international journals and conferences, made pioneering contributions in the fields of deep learning model compression, AI-driven system control, and crowd intelligence perception, and has won the Best Paper Award many times.

Tang Jian currently also serves as a member of the expert committee of the New Generation Artificial Intelligence Industry Technology Innovation Strategic Alliance and the leader of the Intelligent Mobility Promotion Group, as well as an outstanding speaker of the IEEE Vehicular Technology Society.

AI scientists in the industry begin a new round of career choices

It's not just Ye Jieping. In the past two years, many AI scientists in the industry have changed their jobs.

At the end of July, Ma Weiying, vice president of ByteDance and head of AI, also resigned and joined Zhang Yaqin's team at Tsinghua University.

Prior to this, Wei Xiucan, head of Megvii Technology in Nanjing, returned to Nanjing University of Science and Technology; Zhang Tong, director of Tencent AI Labs, joined HKUST...

Their various movements continue to trigger discussions on the difficulties AI experts face in adapting from academia to industry.

Around 2016, as AlphaGo set off a wave of artificial intelligence, many AI experts in academia, either inspired by the times or invited by companies, "went into the sea" and entered the industry.

In the past few years, they have established artificial intelligence R&D systems for enterprises, cultivated artificial intelligence talent teams, and promoted enterprises to keep up with the pace of AI development.

Now, the AI ​​industry has entered a new stage of development. Compared with the scientific research breakthroughs in earlier years , the market and enterprises are more pursuing the implementation of algorithms.

This is more of an engineering consideration rather than the development of cutting-edge technology. In the view of industry insiders, this has led to a decline in the status of scientists within the company, weakened voice, and reduced resources, making it difficult for them to develop within the company.

For AI scientists who cannot adapt to the pace of corporate development, it is a natural choice to start a new round of career choices.

At present, Ye Jieping has bid farewell to Didi, and his next step is still unclear. Inside Didi, he explained his departure as follows:

The exchange with the big data team when I returned to China five years ago made me see the infinite potential and imagination of machine learning in the field of transportation. I am so lucky to have worked with you, who are warm, reliable, and empathetic, to improve travel efficiency and protect every trip with AI in the past 1,975 days and nights. This also made me realize that the best research comes from implementation. Joining Didi was a very correct decision, and this was also the five years with the most gains in my career.

I am very sad to say goodbye to you. I have been thinking for a long time and trying to explore the possibilities of AI in other fields. I am grateful to Didi and everyone for their trust, support and help. AI Labs has Teacher Tang, Teacher Zhu and Kevin. I believe they will continue to lead everyone to make more updated and powerful explorations.

We will not go far, and those who meet will meet again. Let technology play a greater role, and we are still on the same road.