The number of employees of artificial intelligence trainers at home and abroad will reach 5 million

Teacher of AI (Online China)

　　Autonomous driving technology, intelligent service platform, CT image analysis and diagnosis system ... In recent years, the application of artificial intelligence (AI) technology has greatly improved people's work efficiency, and the silent dedication of artificial intelligence trainers is inseparable behind this.

　　A few days ago, the Ministry of Human Resources and Social Security, the State Administration of Market Supervision, and the National Bureau of Statistics jointly released 16 new occupations to the society, and the artificial intelligence trainers were among them. According to Alibaba Group's research data, it is expected that by 2022, the number of domestic and foreign employees of artificial intelligence trainers is expected to reach 5 million, and a large number of talent gaps are urgently needed to be met.

Help AI grow intelligently

　　In the Fangshan District Internet Financial Security Demonstration Industrial Park in Beijing, a "Course" tailored for autonomous driving systems is underway. Through the data annotation tool, 24-year-old artificial intelligence trainer Liu Xue circled a bus in the picture with a rectangular frame, and labeled the box as "bus", and then continued to mark other vehicles, traffic signs and bicycles on the image And other objects.

　　"These labeled pictures, text, and sound data will become learning materials for autonomous driving systems, helping artificial intelligence algorithms learn to identify entities on the road that affect vehicle driving." As an artificial intelligence trainer, Liu Xue did not participate in automatic The development of the driving system, but a large amount of data annotation later laid a solid foundation for the operation of the system.

　　"Compared with AI developers, the work of AI trainers is more biased towards the actual use of products." Duan Yucong, deputy director of the Special Committee of Hainan Provincial Blockchain Industry Research Institute and professor of Hainan University, said that AI trainers use intelligence Training software for database management, algorithm parameter setting, human-computer interaction design, performance test tracking and other auxiliary operations for the AI ​​products in actual use.

　　In other words, the AI ​​product at the beginning of the design is just like a child who has just entered school. The artificial intelligence trainer needs to "teaching according to the aptitude" according to the different landing scenarios preset by the product. Enough wisdom to deal with various problems in practical application scenarios.

　　In addition to training and cultivating AI, artificial intelligence trainers also need to process AI learning materials in the early stage, that is, to obtain structured core knowledge and key data through data screening, and "feed" the data to AI with specific data labeling rules and conduct "Training", and constantly adjust the parameter optimization algorithm based on its feedback, so that AI can better serve humans.

Efficient service to fight against the epidemic

　　During the New Coronary Pneumonia epidemic, artificial intelligence trainers and AI technologies have demonstrated efficient service capabilities in epidemic prevention, anti-epidemic and life travel.

　　On the front line of anti-epidemic, AI technology is widely used in the analysis and diagnosis of CT images of patients. Based on the image feature data of a large number of previously confirmed cases and suspected cases, the artificial intelligence trainer helps the artificial intelligence system to learn to distinguish the difference between the CT image features of the new coronary pneumonia cases and other patients such as influenza, so as to achieve rapid triage and effectively block the source of infection and Reduce the risk of cross infection.

　　Ma Chun'e, CEO of Shukun Technology, said that the diagnostic system has been used in nearly 100 hospitals nationwide, with a sensitivity of over 95%. In the future, through interdisciplinary and multi-modal comprehensive learning of artificial intelligence, further research will be conducted on special populations such as cases of Fuyang, asymptomatic infections, new coronary pneumonia combined with other basic diseases, and provide scientific support for precise prevention and treatment.

　　In Shapingba District of Chongqing City, a “non-contact” epidemic prevention and drainage system was conducted through an intelligent outgoing call system, and residents were asked whether there were any discomfort symptoms such as fever and whether they returned to work from the outside. Based on epidemic situation information and pre-learning, the system can automatically identify suspected patients and persons who need to be quarantined in the shortest time. "If 168,000 pieces of data require half a month by manual analysis, AI only takes about 4 hours." Liu Yi, director of the Shapingba District Political and Law Commission Office, introduced that the intelligent outgoing system can make 3500 calls per minute, which greatly improves the efficiency of the survey. .

　　With the help of artificial intelligence trainers, the intelligent outgoing call system has also been smoothly connected to major railway air ticketing systems, allowing the travel issues of 2.3 million passengers affected by the epidemic nationwide to be resolved in a timely manner. In addition, the artificial intelligence trainers of Alibaba's customer experience business group only need 22 hours to implement the health intelligent service workbench, which meets the growing demand for online health consultations and greatly improves service efficiency.

Talent gap needs to be met

　　"I was assigned to the '3D Semantic Segmentation' group to annotate the picture information." Last year, Xiao Chen, who was working abroad, returned to his hometown in Guizhou to receive corporate training and began his career as an artificial intelligence trainer. Xiao Chen said that although he has no concept of "artificial intelligence", fortunately, the computer knowledge he has learned comes in handy, and I feel that this is not difficult.

　　With the development of the AI ​​industry, more and more business personnel have become "cross-border" as artificial intelligence trainers. Alibaba's artificial intelligence training expert Wang Zhiyu said that the core competence of business staff transformation artificial intelligence trainers is their ability to understand the business. "We can teach him the idea of ​​intelligent training and data analysis, but we cannot teach him the ability to understand the business itself."

　　In response to the growing talent gap, the entire AI industry has begun to train its own artificial intelligence trainers. However, among the nearly 500,000 artificial intelligence trainer practitioners in China, the majority of people are engaged in simple tasks such as data annotation. However, those who have relevant knowledge and experience in analyzing and refining the characteristics of the professional field, training and evaluating related algorithms, functions and performance of artificial intelligence products, as well as designing interactive processes and application solutions of artificial intelligence products are "uncommon".

　　"Artificial intelligence trainers need to continue to learn to improve their overall quality." Duan Yucong analysis said that the AI ​​industry still has a lot of demand for data annotators in the short term, but with the rise of unsupervised self-learning algorithms, AI depends on data Sexuality will be alleviated, and the work focus of artificial intelligence trainers will also change, and the gap in technical talents such as AI algorithm testing will be more prominent.

　　Experts said that if artificial intelligence trainers can systematically accumulate experience, sum up perceptual experience into rational knowledge, through continuous practice, sum up deeper principles, further enhance creativity, and become more competitive in the workplace.

　　Jin Chen