Didi: The problem we solved is actually more complex than AlphaGo

In the past, this company, which was driven by “burn money”, had the privilege of surviving the tickets for the first half of the Internet. The company’s goal now is to become a “world-class technology company.” This is a new vision that Didi CEO Cheng Wei described to the outside world.

The second half of the drop

Liu Qing once told Cheng Wei that Didi may have been the fastest-growing company in the history of China's Internet, not even one of them.

On April 10th this year, it broke through 10 million for the first time in a single day, exceeded 15 million in June, and exceeded 16 million in July. This is the result that Cheng Wei mentioned in the award of the first Di-Tech algorithm contest of Didi.

In the past four years, Didi has achieved rapid development relying on giants' support. According to Cheng Wei, Didi is one of the companies that got the last few tickets in the first half of the Internet. The first half of the Internet is the development of terminals and the interconnection of products. He made an image of the analogy. The connection between people is Tencent, the connection between people and goods is Taobao, the connection between people and information is Baidu, and the connection between people and cars is a drop.

Four years ago, Cheng Wei did not expect that the taxi software would one day rely on algorithm drivers. With the increasing volume of drops: the volume of orders, the number of users, and the big data generated each day, these are far beyond the scope of artificial rules. Cheng Wei said frankly: “We went to Google, Facebook, and Microsoft to understand the structure of all of their core engines, hoping to learn some of the best methodologies. Looking back, I discovered that our platform is more difficult.”

This is due to the fact that vehicle scheduling and user matching are more complex than search engines, and the dimensions, complexity, and real-time considerations go beyond other industries. For example, the question of how to allocate orders: how to do when there are more people and fewer people, how to do it during the peak period and the peak period, consideration of nearby passengers and drivers when there is demand, consideration of passengers and nearby passengers, preferences of passengers, drivers, etc. There are too many variables and factors.

For Dripping, in the first half lucky to get the ticket, how to maintain the advantage in the second half? Cheng Wei said that the second half of the drop is artificial intelligence.

Technical layout

As the chief person in charge of the technology line, Debt CTO Zhang Bo expressed that the development of artificial intelligence in the past three years mainly relied on three elements: big data, cloud computing, and artificial intelligence algorithms.

He told us that the layout of the entire technical aspects of the droplets revolves around the above three-point layout:

Big Data. Now we have some traffic data, and every day we have millions of drivers running trajectories. We will then strengthen our access to traffic data, cooperate with traffic control departments, obtain traffic light data, road camera data, and data on urban road planning, etc.

cloud computing. At the end of May this year, Alibaba Cloud Chief Scientist Zhang Wenxuan joined Didi, Zhang Bo said that Zhang Wenjian will strengthen the construction of this block as a whole. For Zhang Wenjun's joining, some people think that this is a landmark event for Didi to become a technology-based company, which will help DDT to attract more technical talents;

artificial intelligence. The artificial intelligence team including artificial intelligence scientists, He Xiaofei, Dean Research Institute, international leader in machine learning, and Ye Jieping, vice president of Dripping Research Institute, are looking for the best talents in the world. At present, Didi has established Silicon Valley R&D center in Silicon Valley. In the future, it may also consider building similar R&D centers in the UK and Russia.

"The problem solved by the drop is actually more complicated than AlphaGo." Ye Jieping, vice president of the Drip Research Institute, told us. He cited an example, the capacity to mobilize a similar AlphaGo problem, the next 10-15 minutes the relationship between supply and demand in each region, multi-vehicle area, we hope the car transferred to a small place; further than such as route planning, from Which route to go from A to B, and how to make decisions at each intersection, may make a partial judgment. This is similar to AlphaGo, hoping to take control of the whole situation from a long-term perspective.

Drip algorithm competition

The Di-Tech algorithm contest was also initiated in this context. Zhang Bo hopes that the platform set up by the contest will attract more talents with excellent algorithms and machine learning through the data and computing resources.

According to official statistics, the first algorithm competition held by Didi attracted a total of 11,440 technical personnel. The final team size was 7,664. The top three teams were selected through nearly three months of competition.

The contest titled “Solving Supply and Demand in the Travel Industry Problem”: Give participants 30 minutes of real regional desensitization order information, which uses the algorithm to predict the next 10 minutes, and then compares it with the real situation, and then obtains A better algorithm makes predictions possible in the future. According to Ye Jieping, the topic is mainly to examine the ability of the players to deal with data, business understanding, mining features, and establish a suitable model.

In the eyes of the outside world, the Drip Preparatory Algorithm Competition is more representative of Didi's transition from a "burnt money" driven company to a "technology" driven company. Cheng Wei said that the first algorithm competition is just a beginning. In the future, it will continue to expand the scale and influence and attract more talents to participate.

Stainless Steel Bright Rod

SUS304 Stainless Steel Bright Rod, TP304 Stainless Steel Bright Rod, 316L Stainless Steel Bright Rod

ShenZhen Haofa Metal Precision Parts Technology Co., Ltd. , https://www.haofametal.com