On September 23, the Mifeng Pai global launch event titled “DATA FOR AI, Everyone Can Be a Teacher” was held in Shanghai. At the event, Mifeng Technology officially launched Mifeng Pai, the world’s first high-quality physical AI data crowdsourcing platform for all categories of products. It also introduced the robot trainer career program and a billion-yuan subsidy program, and initiated the compilation of the “Embodied Intelligence Crowdsourcing Data Development Report” as well as the Scene Data Alliance, providing a large-scale approach to addressing the lack of real data in the embodied intelligence industry.
Yao Maoqing, Chairman and CEO of MiFeng Technology, said that what robots really need to enter factories, stores, and homes is not algorithms or hardware, but massive and real human action data. MiFeng Pai transforms ordinary actions such as叠 clothing, making tea, and organizing goods into high-value data for robot training, turning the slogan “everyone can be a teacher” into reality.
Leifeng Net (public account: Leifeng Net) has learned that the MiFeng Party platform consists of two layers: at the bottom is the infrastructure including the MEgo series collection devices, MiFeng Party APP, and the MEgo Engine data post-processing platform; at the top are the scene network and collector network — wherever there is real work, that is the origin of data, and any ordinary person can accept orders, collect data, and settle payments at any time. Among them, MEgo View is the industry’s first full-sensory multi-view acquisition device, with multiple cameras covering a field of view of over 300 degrees; MEgo Gripper has a reconstruction accuracy of millimetre level and can be equipped with three-dimensional tactile sensing. After users download the Meifengpi APP, register under their real names, and obtain the device, they can earn income through the four steps of “order acceptance — acquisition — inspection — payment”. It supports multiple payment methods such as WeChat, Alipay, and Yinlian, with funds available within two minutes, and taxes are automatically paid by the platform. The platform currently covers 22 major categories, over 5,000 task types, and more than 50,000 real scenarios. During the internal testing from August 15 to September 15, nearly 20,000 users registered, with the top user achieving a monthly income of 5,111 yuan.
Regarding the establishment of new occupations, the press conference awarded the first batch of global “Robot Trainer” certifications to 16 beta-tester users, and officially launched the Robot Trainer career program for the entire society. Previously, departments such as the Ministry of Human Resources and Social Security had included “Embodied Intelligent Robot Application Technicians” in the category of new occupations. This action by the press conference aligns with the national industrial direction. Zhang Zhifu, Vice President of the MiFeng Party’s Business, announced a billion-yuan subsidy plan: 50 million yuan will be used for task subsidies, 20 million yuan for equipment subsidies, 30 million yuan will be invested in building a offline service network covering 40 cities, and 2 million yuan will be spent on dual insurance protections for people and devices; at the same time, a five-level growth system from M1 to M5 will be launched, with the goal of achieving 1 million daily active users and 1 million robot trainers within two years.
The industry ecosystem and governance are also advancing in tandem. FeiBeng Technology and the China Academy of Information and Communications Research jointly launched the preparation of the “Body-Like Intelligence Crowd Sourcing Data Report”. It is jointly built by the National Robot Testing and Evaluation Center, Guizhou Big Data Group, and China Telecom Data Development Center. A scenario data alliance was established at the press conference, with participants including Longqi Technology, Huayi Hotel, and Deji Hospital among 12 entities. The first batch of units have moved in; Liu Li, the technical leader of Miefeng Technology, read out the “Action Initiative on Safeguarding and Protection of Embedded Intelligence Data”, establishing guidelines for the crowdsourcing data industry from three aspects: collection standards, full-chain security, and industry collaboration. According to the press conference, currently, global high-quality real physical interaction data is only in the range of about 100,000 to one million hours, while the commercialization of robots requires tens of millions or even hundreds of millions of hours of data. The gap exceeds 99%, and crowdsourcing collection is becoming a new way to expand data supply.
Yao Maoqing said that the experience of hundreds of millions of hands will eventually merge into the capabilities of robots. Currently, the MiFeBeng faction has started operations in 40 cities across the country. In the future, it will work with thousands of industries to complete the data landscape for physical AI scenarios, and accelerate the arrival of the AGI era.
Leifeng Net original article, unauthorized reproduction is prohibited. For details, see Reproduction Guidelines.