In the past five years, nearly 60 billion yuan has been invested in R&D. Midea Jin Chengzhen: AI is very popular, but it has not yet reached the stage where quantitative changes lead to qualitative changes.

Author｜Huang Nan

Editor｜Yuan Silai

Entering 2024, the technological development represented by large-scale models has pushed social and economic development into the era of digital intelligence. How to traverse the technological revolution cycle and find the best path for AI technological innovation and industrial implementation?

A traditional manufacturing company that mainly sells hardware and has transformed into a technology enterprise that provides software and solutions - Midea, has given its answer: Embrace AI + industry, and drive the transformation and upgrading of traditional industries with digitalization.

Recently, the 29th Technology Month "Visionary Conference" of Midea Group was held in Foshan, Guangdong, and several products equipped with Midea's Meiyan large-scale model were exhibited on the spot.

Visionary Conference

In the AI industrial chain, the needs and feedback from the technological innovation and product end need to smoothly enter each scene link of the application, participate in the actual implementation and iterative optimization, and finally enter the product innovation and industrialization. With the "AI +" strategy as the core, Midea Central Research Institute has proposed the layout of four main scenarios: "AI + Industrial Robot", "AI + Intelligent Manufacturing", "AI + Smart Home", and "AI + Medical Imaging".

Take the robot field as an example. In the past, due to the immaturity of technology, robots lacked the cognitive ability of complex scenes, and the application space was limited. For a long time, for example, industrial robots have long existed in the form of traditional collaborative arms, and their functions and practicality have a large gap with people's expectations. The arrival of the AI large-scale model has changed this situation. Its general ability to solve multiple tasks has made people see the possibility of robots from single-task delivery to large-scale application.

Ding Han, an academician of the Chinese Academy of Sciences and a professor at Huazhong University of Science and Technology, pointed out at the Visionary Conference, "The concept of robotized intelligent manufacturing is to manufacture with robots as the core, and there are four reasons: First, it can be parallel; second, it is efficient; third, it is easy to reconfigure; fourth, the integration is very convenient. In the future for a long time, the development trend of robots lies in continuously improving their adaptability to the environment and autonomy of operations. It will become a disruptive technology and may become the entry point of the third industrial revolution."

Ding Han, Academician of the Chinese Academy of Sciences and Professor at Huazhong University of Science and Technology

In this context, Midea has started to make layouts in areas such as computing power, AI algorithm platform, multi-modal intelligent sensing and control technology, and embodied intelligent large-scale model. Zhang Xiaoyi, Vice President and CDO of Midea, disclosed that the first version of Midea Software Engineering Institute's computing power platform has been released. At the basic model layer, Midea's strategy is not to develop the underlying large-scale model. "Whether it is a good open-source model, a closed-source model, a purchased model, or a model we trained ourselves, we will use it well according to different scenarios. Therefore, the algorithm platform and access platform are very crucial. One is to solve its service problem, and the other is to solve the problem of information security."

Based on the new-generation infrastructure of the large-scale model, Midea has also launched native applications, which are the two strategic projects of the Home Intelligence Entity and the Factory Intelligence Entity. Zhang Xiaoyi stated that the Midea Factory Intelligence Entity strategic project will incorporate KUKA robots.

In addition, Midea also disclosed on the spot that in the humanoid robot track, which is currently the hottest, Midea has also made some explorations, mainly focusing on the core component field.

Roundtable Forum - "The Future of Artificial Intelligence and Energy Transformation"

Today, in different countries and regions around the world, a large part of smart products are produced by Chinese companies. The huge overseas digital intelligence demand market has become a must-win territory for domestic enterprises. Currently, the overseas market has become Midea's second home field. For example, the "North American Cold Climate Heat Pump Series" technology of Midea's Household Air Conditioning Division has accelerated the decarbonization transformation of houses in New York, USA, and has been expanded to several cities including Boston and Chicago under the promotion of the American Energy Efficiency Alliance.

According to official public data, in the past five years, Midea has invested nearly 60 billion yuan in the R & D side, and the R & D investment in 2023 exceeded 14.5 billion yuan. A total of 33 R & D centers have been established globally. At the same time, in terms of the construction of the talent echelon, the number of R & D personnel of Midea has exceeded 23,000, including more than 7,000 masters and doctors. Through the construction of a global R & D system, the establishment of a global top talent introduction mechanism, the introduction of dozens of strategic cooperation experts, and the establishment of multiple academician workstations, joint scientific and technological research has been carried out.

Midea robots exhibited at the Visionary Conference

In order to further understand Midea's technological layout and product outlook on the AI + side, Yingke recently also interviewed Jin Chengzhen, Technical Director of Midea Group Central Research Institute. The following is the conversation content, edited:

Yingke: Home appliances are an already very mature industry, especially white goods, which belong to the "supply - demand" driven industry. It can be seen that the development of AI technology in recent years has also injected a lot of new vitality into this field. From the perspective of the entire industry, what changes has it experienced in the product and technology ends?

Jin Chengzhen: AI is indeed very hot recently, but from the perspective of technology and product ends, it has not accumulated to a stage where quantitative change can cause qualitative change, and Midea has not yet come up with a complete solution. But in this regard, we already have a series of projects under study. By introducing AI technology and coordinating the multi-product matrix layout, we hope to have results to show everyone in the short term.

Yingke: In 2020, since Midea added "Technology Leadership" to its strategy, what has Midea done in recent years? Are there any satisfactory results that can be shared with us?

Jin Chengzhen: Taking the dryer as an example, after several years of continuous investment, the performance of the dryer has been significantly improved, reducing the drying time while improving energy efficiency. This is the result of our efforts on this single product of the dryer, and its market performance can also confirm this. The second is the U-shaped window machine launched for the US market, which has now become a benchmark product in Midea's product line. The third is the Dual Wash Station, which combines the washing machine and the sweeping robot, integrating these two completely different product forms into a new category.

These several products can all show Midea's tolerance for technological or product innovation, which also affects the development of the entire home appliance industry.

Yingke: How do you view the impact of AI on home appliance product innovation? Which scenarios will be its key application areas?

Jin Chengzhen: At present, the application of AI in home appliance hardware is still limited to the voice interaction link. Through cooperation and the invocation of large-scale model manufacturers, it can help Midea achieve natural communication in products. On the other hand, what users perceive less is that the addition of AI large-scale models can effectively improve the R & D efficiency; for example, for simulation data training, it usually takes 1 week under normal circumstances, but after introducing AI big data tools, it can be completed in one day.

Yingke: Different from other categories, how to achieve technological differentiation in home appliance products? What are the differences in the needs of users in different regions? What are the requirements for product R & D?

Jin Chengzhen: We divide the global market into several regions such as China, the United States, Europe, and Japan. Users in the Chinese and European markets will pay more attention to the energy efficiency level, followed by whether the product size matches the personal home use scenario; on the contrary, American users do not care much about this. In contrast, the Japanese market is a unique existence in terms of its product combination form or size in all aspects.

Based on these regional differences, it is determined that our R & D cannot rely only on Chinese engineers, but requires the participation of global excellent technical R & D talents. Therefore, Midea has established R & D centers in the United States, Germany, Italy, Japan and other countries, precisely considering the localized product design and the understanding of user needs.

At present, we are evaluating the expansion of the US R & D center by about three times on the basis of the original land scale and building scale, and the specific implementation plan is still under discussion.

Yingke: In recent years, what innovations have Midea's home appliance products made in technology compared to before?

Jin Chengzhen: Taking air conditioners as an example, the common solution in the past was one outdoor unit plus one indoor unit, but currently, the solution that Midea mainly promotes and has a large share is to adopt the multi-split air conditioning route, which can effectively save space. But a major challenge of the multi-split air conditioning is how to meet the energy efficiency requirements of different regions and the personalized connection needs of different users in a new form. At the same time, we have also introduced AI capabilities, which can adjust the comfortable air flow according to the space temperature by identifying and locating the specific location of the user.

Midea home appliances display at the Visionary Conference

Yingke: The following are several questions about the overseas R & D talent system. We know that Midea has gone global very early. Compared with the domestic layout, what is the main source of pressure on overseas talents for us?

Jin Chengzhen: First is the issue of popularity. Midea has a certain industry popularity in China, but globally, there are still many regions that do not know much about Midea, and the brand influence cannot be reversed quickly in a short period of time. More importantly, in terms of product power, we hope to be able to launch excellent localized products and impress customers through technical capabilities, such as the U-shaped window machine launched in the US region.

Overseas R & D centers have played an important role in Midea's global path, and more and more overseas talents are willing to join our R & D centers. Of course, it is inevitable to encounter difficulties in the process of overseas exploration and expansion, but this is also the focus of our attention this year. We believe that Midea can gain something after several years of efforts.

Yingke: How is the progress of Midea's global R & D talent system construction? What has been specifically done?

Jin Chengzhen: At this stage, it is still in the initial stage of construction and has not yet formed a very mature or standardized state. Next, we will focus on providing more growth mechanisms for learning and training. In the past, Midea Group's talent training courses were limited to the domestic market, but starting this year, we have also officially launched recruitment to allow overseas talents to participate in Midea's talent growth system.

Yingke: How does Midea introduce outstanding overseas talents? What are the deficiencies?

Jin Chengzhen: Currently, most of them are young PHDs, and some are field experts with rich overseas work experience, but mainly Chinese. At the same time, we also hope that more outstanding foreign field experts can join in the future.

Yingke: From the industrial end, which dimension has a greater impact on the development of the home appliance industry, hardware or software?

Jin Chengzhen: In the past, the competition in the home appliance industry relied on hardware strength, but with the changes of the times and the rise of the application ecosystem, Midea has become a company that also needs to layout and deeply cultivate in the software layer. Hardware manufacturing is Midea's gene. Based on the traditional advantages, we will also continue to integrate with AI and make continuous accumulation in the software layer.

Yingke: How do you view the technological development trend in the home appliance field in the next step? What are the challenges and opportunities?

Jin Chengzhen: The external perception of home appliances may think that this industry has reached its peak and there are no new tricks to play, but in fact, there is a certain threshold to enter this track, and we believe there are still many possibilities worth exploring. In the future, Midea will continue to exert efforts in the three directions of high energy efficiency, appearance and intelligence, continuously strengthen R & D investment, and optimize the innovation mechanism.