Traditional industry solutions

It has been three years since the establishment of Microsoft Research Asia Innovation Exchange. From the very beginning, the concept of DTaaS (digital transformation as a service) was put forward, and recently Microsoft AI quantitative investment platform "micro-mine Qlib" and Microsoft multi-agent resource optimization platform "group strategy MARO" were officially released. The platform road of DTaaS has achieved initial results.

The number of member companies of Innovation Exchange has now expanded to 27, including leading enterprises and innovative companies from finance, logistics, education, health care, manufacturing, retail and other industries. AI scientists from Microsoft Research Asia have worked closely with industry experts in various fields to stimulate their wisdom, promote enterprises to accelerate digital transformation, help enterprises' business models keep pace with the times, and have carried out many forward-looking AI cooperative research projects with * * *, which have landed in many industries.

Microsoft is a platform company. In the process of independent cooperation projects (which can be called "points"), our researchers constantly abstract the AI logic in the core business scenarios, dig out the internal essence of the problems, and gradually extend the innovative technological achievements to a wider industry field (which can be called "fields"), and build these technologies into a common platform to realize the closed loop of AI application in a certain industry field. Only by realizing the leap from "point" to "surface" can AI truly change all walks of life.

Huaxia Fund, a member of Innovation Exchange, and Microsoft Research Asia have been cooperating in the field of quantitative investment and multi-factor stock selection since 20 17. Based on the strategy of "AI+ index enhancement", the two sides excavated a portfolio with low correlation with traditional investment methods, and realized the differentiated competition of Huaxia Fund in the financial market.

In fact, in the whole process of stock investment, stock selection is only a small step. In order to ensure the success of investment, we need to understand the relationship between stocks, so as to control risks and avoid the problem of "putting eggs in one basket", just as we should buy stocks carefully and relevant enterprises should diversify their investments. At the same time, constraints such as transaction cost and turnover rate need to be considered; When forming the optimal portfolio, we should also consider the execution of orders and trading factors.

Based on this idea, Microsoft Research Asia has built a micro-mine Qlib 1 AI quantitative investment platform on the basis of previous research, hoping to realize the AI closed loop of quantitative investment process. As an open source toolkit, the platform can be used by financial institutions and individuals, thus enhancing investors' technical reserves and comprehensive level, improving the efficiency of the whole market, and thus forming a greater virtuous circle in the investment field.

In the future, we will also consider expanding the open source platform from a horizontal and vertical direction-asset allocation and financial supervision. Asset allocation is an extension of stock investment. In addition to the secondary and primary stock markets, it can also help fund holders to plan more investment portfolios from bonds, foreign exchange and even gold, further sharing investment risks and ensuring higher returns.

On the other hand, the business form of financial services industry is becoming more and more complex, involving more and more institutions and individuals, and various operations are dazzling. For regulators, management is increasingly difficult. Looking for patterns, finding anomalies, monitoring risks, and mining insider information in complex environments are exactly what AI technology is good at. Therefore, in the process of communicating with partners, we also realized that AI can be a right-hand man in the field of financial supervision.

In the cooperation with OOCL, a member of Innovation Exchange, we have covered two major business scenarios in the logistics industry, namely supply and demand forecasting and route optimization, and optimized the existing shipping network operation by using the latest artificial intelligence technologies such as deep learning and reinforcement learning. The cooperation with SF mainly focuses on intelligent claims early warning, link prediction and dynamic pricing, and explores the application value of AI in more links in the logistics field.

These two cases cover many basic scenarios of supply and demand matching of logistics chain, which are very representative, but they are still breakthroughs in "points". In fact, from the perspective of big logistics, in addition to container and truck scheduling, it also involves warehousing management, cargo scheduling in warehouses, robot automatic sorting, and the relationship between warehousing and terminals, suppliers and retail terminals. All these sub-problems are integrated to form a complete logistics supply chain management platform.

Among them, one of the most fundamental problems to be solved by the logistics industry is the matching of supply and demand. Therefore, aiming at the core engine of resource supply and demand matching, which can be applied to all walks of life, we developed and opened the multi-agent resource optimization group policy MARO platform 2. Perhaps some enterprises have developed various IT systems to solve the sub-problems related to the matching of supply and demand of resources in the logistics chain, but our platform is the first in the industry that can be closely integrated with AI technology. Many business scenarios involving the matching of supply and demand of resources, such as the matching of bicycles and users, and the matching of tasks that data centers need to run with actual physical machines, can be solved through the MARO platform.

It can be considered that MARO is a multi-industry and cross-domain full-chain resource optimization AI solution. Users only need to provide a simple interface or data, and the platform will automatically generate an emulator for intensive learning and training, and finally give an industry solution. The open source MARO platform will not only be limited to the logistics industry, but also help more traditional enterprises to renovate resource matching tools, realize resource optimization in a data-driven way, and greatly save costs.

Similar to the construction and development of the general AI platform in the financial field, we hope to continuously enrich the general AI platform in the logistics field. Especially for small and medium-sized logistics enterprises, they will be able to directly use the general AI platform in the logistics field, including MARO platform, which greatly shortens the process of building their AI intelligent business system and forms a late-comer advantage.

Whether it is a general AI platform in the field of finance or logistics, it is based on the application "point" that AI is best at. As an assistant of human intelligence, artificial intelligence can show the ability to surprise enterprise decision makers when analyzing and solving complex problems only through short-term learning and debugging. When we find one core application "point" after another in different industries, we can gradually "get through" every traditional industry from point to surface with AI.

At the same time, it also actively cooperates with Microsoft's product department to integrate more AI decisions into Microsoft's product system. In the future, AI will surely achieve closer integration with different industries and different scenarios, and comprehensively lead every enterprise and industry to the AI era.

1 Microsoft AI quantitative investment platform-micro-mine QLIB:/Microsoft/QLIB

2 Microsoft multi-agent resource optimization platform-Group Policy MARO:/Microsoft/MARO