IDST: What is Alibaba’s most mysterious department doing?

[[130588]]

Key

[“IDST needs to do what other departments in Alibaba are unwilling to do and cannot do.” This is how Wang Jian defines the role of IDST.]

In the past two days, Alibaba invested 15 million yuan to launch a data competition code-named "Tianchi". The contestants competed on how to use existing big data to make algorithmic recommendations for Taobao women's clothing matching and predict the inflow and outflow of Yu'ebao funds. The driving force behind it is Tu Zipei, the head of IDST, which is known as Alibaba's most mysterious department. He is the author of "Big Data" and "Top of Data".

On March 25, a reporter from China Business News conducted an exclusive interview with Tu Zipei in the conference room of Alibaba's Xixi Park. As a military man, he has a special heroic spirit, which is rare in the IT circle.

Tu Zipei was transferred from Silicon Valley to Alibaba as vice president at the end of last year, responsible for the commercialization and innovation research of big data. The news attracted wide attention in the industry, but the new department, which is mainly composed of scientists, has not been unveiled to the outside world.

Within Alibaba, IDST, which does data research, has intersections with almost every BU (business unit) of the group. How can this department, which plays the role of a transit station, shoulder Jack Ma’s business dream of “letting data drive the future”?

Do things that other departments are unwilling or unable to do

IDST (Institute of Data Science & Technologies), the Data Science and Technology Research Institute, is obviously different from the operating style of the business units within Internet companies just from the name. If the various business units are likened to the martial arts factions fighting on the battlefield, then IDST is full of academic style, and most of its employees have a background as scientists.

As the person in charge, the English name printed on Tu Zipei's business card is Jack, the same as the familiar Jack Ma (Ma Yun). His position only says Vice President of Alibaba Group, and there is no other related title such as IDST.

Currently, the department has about 150 employees, distributed in Hangzhou, Beijing, Silicon Valley and Seattle, of which Hangzhou has about 50 people. In terms of organizational structure, IDST belongs to Alibaba Cloud; in terms of hierarchical relationship, Tu Zipei reports to Dr. Wang Jian, the chief technology officer of Alibaba Group, who is the original designer of the idea of ​​IDST.

In fact, the three leaders of IDST all have important backgrounds. In addition to Tu Zipei, Qi Yuan is a tenured professor in the Department of Computer Science and the Department of Statistics at Purdue University, specializing in machine learning and artificial intelligence; Jin Rong is a tenured professor at Michigan State University, and has won the National Science Foundation Award (NSF Career Award). Tu Zipei told reporters that in terms of division of labor, Qi Yuan leads the Chinese team, while Jin Rong leads the American team, and he is responsible for research in business and data economy.

Before going abroad, Tu Zipei worked in the Guangzhou Armed Police Border Defense Corps for eight years, and then worked in the Guangdong Provincial Economic and Trade Department for two years. His connections in the government are also a resource that Jack Ma values. You should know that since last year, Alibaba Cloud has reached cooperation agreements with more than a dozen provinces across the country. Connecting with government agency data is a treasure for any private data company.

"IDST must do what other departments in Alibaba are unwilling to do and cannot do." This is how Wang Jian defines the role of IDST. Why are they unwilling to do it? Compared with those business departments that can achieve immediate revenue growth, it is difficult to achieve results quickly in areas such as data research, machine learning, and natural language processing. Tu Zipei also said that since it is an innovation department, it does not have traditional performance evaluation targets. Why can't it be done? Just listening to these cutting-edge research fields will tell you how high the threshold is.

Similar research laboratories have taken root in powerful technology companies around the world. For example, Google Brain, also known as the Google X laboratory, conducts research on artificial intelligence and machine learning, which can be applied to fields such as speech recognition and driverless cars. Microsoft's similar deep learning system is called Adam. In May last year, Andrew Ng, the former "Google Brain" scientist who was the first to make a machine recognize "cat", returned to China from Silicon Valley and joined Baidu's Deep Learning Institute to promote artificial intelligence research.

From Silicon Valley to China, not only scientists, but also a large number of high-level talents in cloud computing are demonstrating this flow trend, such as from Oracle to Alibaba Cloud. Tu Zipei believes that this trend shows that in cutting-edge fields such as big data and cloud computing, the gap between China and the United States is minimal, and more and more people see the opportunity to return to China to make achievements.

Invest 15 million yuan every year in data competitions

How much change will the analysis and use of big data bring to Internet companies and users' lives?

Last year, Alibaba held a Tmall algorithm recommendation competition, in which a team of six college students took away the first prize of 1 million yuan. The algorithm they designed was 16.9% more efficient than that of Alibaba engineers. This algorithm was then applied to the "Double 11" shopping festival. In layman's terms, it makes product recommendations more accurate. It is estimated internally that this algorithm has brought tens of millions of yuan in economic value to the platform, which has also allowed Alibaba to see the power of new external forces.

Wang Yiting, the project leader and senior manager from the IDST department, told the reporter of China Business News that one of the six people will soon join Alibaba, one will start a business based on this instant success, and the others have not yet graduated. The Tianchi competition, which just started on the 26th, is similar to this. It will last for a year and has the meaning of a "league". Last year, a total of 7,276 teams from home and abroad participated in the competition.

The purpose of IDST to hold such competitions is to attract talents, innovate commercial applications of products, and promote the construction of big data ecology. This was also confirmed by Tu Zipei, who told our reporter that Alibaba invests about 15 million yuan in this competition every year, of which the largest expense is computing costs. Alibaba has provided 1,000 servers specifically for contestants, and the project will also connect with venture capital to incubate the contestants' entrepreneurial dreams. "If there is any commercial benefit here, it is only that the contestants' algorithms are applied to Alibaba's business."

In this Tianchi competition, contestants will compete on three subjects, one of which is the matching of women's clothing on Taobao. Alibaba will desensitize the big data on Taobao and give it to the contestants, who will then need to design an algorithm to use image recognition and other technologies to let machines match clothing, replacing the previous manual selection and matching. Of course, the "eyesight" of the machine will ultimately be evaluated by consumers.

For Yu'e Bao, which has a fund size of 578.9 billion yuan, the contestants need to accurately predict its future daily fund inflows and outflows to ensure high returns for users and avoid runs caused by large-scale redemptions. In addition, they need to predict the shopping preferences of Taobao mobile users.

Topics such as recommending music based on users' preferences on Xiami Music will be implemented in the next competition.

All business is digitized, all data is business-oriented. Jack Ma has been "preaching" his data views in many high-end occasions around the world in the past year. Tu Zipei said that China currently has 14% of the world's data, and it is expected that this proportion will reach 21% by 2020, but the data utilization rate is less than 0.4%. The space for data commercialization and the future value of the data economy have already emerged.

Data openness and data commercialization will be two intersecting lines that will continue to extend in the next few years. Tu Zipei said that in the past, writing books and giving speeches were for knowledge dissemination, but this time, when he moved his family from the United States to Hangzhou, his role requirements have changed to knowledge innovation, which will be more challenging.