S.F. Express CTO Tian Min: For Logistics Companies, Giving Up Computing Is Giving Up On Survival And The Future
摘要： S.F Express is in fact exploring the science of computing, so as to improve efficiency and lower the cost.
China’s logistics industry is developing at a rocket speed. The scale of the logistics industry in China and the industry’s fusing with other industries, as well as its relationship with the economy and society, have pushed this particular industry to a new height.
According to the 2016 China Express Industry Development Index Report from the State Post Bureau of People’s Republic of China, China’s express business volume maintained its top one position in the global market in 2016, accounting for 40% of the global market. China’s express market contributed 60% of the global express industry’s growth, the report also writes. In addition to that, China’s express industry had generated four trillion yuan for online retailing in 2016, which took up 12.5% of the total retail sales of consumer goods, supporting directly the outbound agricultural product sales of ￥100 billion while creating a value of ￥120 billion for the manufacturing industry. It’s apparent that the express industry has been an indispensable driving force that nurtures new economy and accelerate the economic development.
With a scale that’s even unprecedented worldwide, China’s express industry grows without a model to learn from. Besides, the very industry is also constantly challenged by new technologies such as cloud computing, big data, AI, and IoT etc., which disrupt the express industry’s operation model and management concept rapidly. It’s like having to change the plane’s engine when flying.
Tian Min, CTO at S.F. Express, believes it’s time to go back to the very nature of logistics and catch the true essence of this field. In his talk at AI & Smart Logistics Roundtable Forum, hosted by S.F. Express, TMTPost, and CARDOPT, he pointed out that logistics originally means the science of computing. He emphasized that if a logistics company gave up on computing, it would mean giving up the future and survival. “A good logistics company’s management and operation must be based on data and computing.”
What follows is Tian Min’s speech at the AI & Smart Logistics Roundtable Forum, hosted by S.F. Express, TMTPost, and CARDOPT:
To implement AI technology, it’s necessary to understand the working mechanism behind it. The very fundamental things you need to know about are the models and algorithms of statistics and operation research etc. No matter in which business scenarios you are using AI technology, you need to know about the basics of AI.
Logistics is the science of computing
Today I want to elaborate on the term, logistics, first. Is it logistics merely about moving shipments around?
The word logistics originated from the Greek word Logistikos, which means accountant or responsible for counting. The first theoretical analysis of logistics was by the Swiss writer, Antoine-Henri Jomini, who studied the Napoleonic wars. In 1838, he devised a theory of war on the trinity of strategy, ground tactics, and logistics. The earliest concepts and theories on logistics were formed in America in the 1930s. The word originally meant the distribution of goods. In 1963, the concept was brought to Japan, and Japanese people translated it into ‘the moving of goods’ (物的流通) and then started to use the abbreviated "goods moving"（物流）to represent logistics after the 70s.
The Chinese word logistics was imported from Japan. It’s vivid, but doesn’t tell the computing science nature. Today when we talk about logistics, we think about the moving of goods, but that shouldn’t be the case. Indeed, we should talk about the moving of goods, but also the science of computing.
We are hosting this forum with TMTPost and CARDOPT today because we need to explore and study the science of the computing in logistics. The improve of efficiency and reduction of cost in logistics can only be achieved through the science of computing. Additionally, we need to fully understand the origin of logistics so as to accurately analyze the current conditions of logistics and the issues it is facing. We need to do all these to build a future with smart logistics system.
Smart logistics, driven by data and computing
Smart logistics is achieved by technological means such as big data, cloud computing, smart hardware etc. that enhance logistics system mind, and the process of sensing learning, analysis, decision making and executing. The logistics system is therefore improved to be smarter and automated to drive industry development, lowering logistics cost and bringing up efficiency.
A good logistics company will plan and operate according to data and computing. If a logistics company gave up on computing, it would mean giving up the future and survival. The logistics in the future will exhibit many characteristics, such as connectivity, data-driven. That said, all elements in logistics would be connected and digitalized. Data will be used to drive insights, decisions and actions. Deep collaboration and highly-efficient execution would form among enterprise groups, enterprises, and organizations. The logistics system as a whole will be optimized based on a smart algorithm and all parties would have their part of the role, which they would fulfill efficiently.
S.F. Express has rich data, such as data on waybills. If we could turn data into information, an immense amount of value can be produced. The data-driven smart logistics will be a deep collaboration on a societal scale. It will not be just about one company. S.F. Express today is more open. Our Hive Box is accepted by the consumers in a short time. This shows the tendency of connectivity, data-driven, deep collaboration and efficient execution.
Four core elements and five transformation strategies of AI
There are four core elements of AI: data, use scene, technology and algorithm. Firstly, you must have data, a very fundamental element. Then you must have a use scene. Without it, you have issues. Thirdly, you need to have the technology to pull it off, which includes hardware, software etc. In the future, the hardware might be designed by algorithm models instead of having only some options of hardware, because every model deals with different issues and data. There isn’t a single hardware system can handle these many complex issues and different data. Fourthly, the algorithm should be constantly studied and improved.
We work with prestigious universities both home and abroad and tech companies as well. S.F. Express wants to gather and connect the most talented minds and companies in the world to study these challenges and bring about solutions. We want to help the Chinese logistics company get to the next level.
There are five strategies for transforming AI, including successful case, data ecology, technical tools, seamless work process accessibility, and an open culture and organization. First you need to have a successful case. You can’t start something that’s unrealistic in the very beginning. Starting from scratch is hard. Besides that, you would need data and seamlessly integrate it into the operation and achieve a terminal-to-terminal data ecology. In addition, you need an open culture to enable the communication between experts, scholars and enterprises from all fields. Otherwise you can’t achieve the transformation of AI and smart logistics.
S.F. Express chose to adopt a diverse strategy because our vision is to provide our clients with more services based on our comprehensive logistics service ability. The services I am talking about here include business service, financial service, data and technology service.
S.F. Express is rich in data, including data on logistics operations like waybill data, logistics center data, IoT data, client sensing data, business data, financial data and data from external operations.
Six business scenes for S.F. Express’s AI
S.F. Express’s AI application scenes include smart logistics, smart service, smart decision-making, smart management, smart map, and smart packaging etc.
S.F. Express has tens of freight planes, tens of thousands of trucks, several thousands of logistics facilities, over 200,000 delivery staff. This is our company’s body. This body requires high collaboration and intelligence. In the upcoming years, we will apply AI and promote AI in every field. I have personally experience what it is like to be in the very forefront of the logistics industry with our R&D staff. We wanted to observe what parts of the job are highly repetitive that can be replaced by technological means so as to free human workers to engage in works that generate more value.
Our company has many excellent employees, but many daily works of theirs are very similar and repetitive. We can use machine learning to train an artificial brain to help make decisions. Eventually a smart brain that evolves itself can distribute tasks and manage the operation centrally, making sure every decision and execution serves to be optimal and that our clients can enjoy the best service.
The greatest challenge on service is maintaining consistency and stability. Our clients felt that logistics companies’ service quality is not stable. Sometimes the service is good, sometimes it’s bad. Sometimes it’s fast, and sometimes it’s slow. It’s just not stable. In the future, smart logistics can insure the consistency and stability of the service.
Let’s talk a little about the forecast on business volume. At present, business peak time is driven by sales promotions and shopping festivals, which is a great waste of social resources. To plan and allocate resources, we need to make forecast on the business volume from different perspective. We would even predict the business volume in the future five years of ten years. We also make forecast on the volume in a few days. We try to use technologies and methods like machine learning and time series analysis to make all kinds of forecast and study them. We look for elements that affect the business volume, such as the weather, the season, industrial structure, government policies, and GDP etc.
Cases of AI application
Route planning is one example of AI application. Traditional rout planning algorithms and tools can no longer keep up with the logistics issues that are presented by today’s complex and dynamic rapid changes. New mindset, new algorithm and new technology should be applied. Furthermore, mainstream map services today are controlled by Internet giants as part of their ecologies. That said, they are not neutral, and they are more of a consumer product instead of industrial. The logistics industry needs map service that provides high accuracy, stability and real-time map service. So we have been investing a lot of efforts in studying smart logistics map. By combining technologies like GIS and reinforcement learning etc., we explore and study route planning tools that are more suitable and can help optimize the cost and effectiveness.
Another example of AI application is digital smart management. Logistics companies have many spaces, facilities, and operators. In general, the operation is managed and operated by humans, and through human observation. We are studying to utilize technologies like machine learning etc. to automatically identify the objects in a space and enable the system to learn from extraordinary managers and operators. Gradually we can achieve assistive decision-making and decision-making automation.
The last application I want to talk about is the waybill recognition. Little might people know that it’s hard for machines to identify Chinese characters while English and numbers are easier. We use image recognition, address base, and convolutional neural network to increase the accuracy of the recognition, drastically lowering the work load and mistake rate.
There are many other applications. AI has a vast application scene in the logistics industry. We will constantly explore, study and build a smart logistics brain at Shunfeng.
The article is published with authorization from the author @Wu Ningchuan, please note source and hyperlink when reproduce.]