The second batch of new occupations was released, and artificial intelligence trainers were selected
It is not a simple requirement for AI to be a teacher
Recently, hundreds of "artificial intelligence trainers" were wearing masks and earphones in Kesi Network Technology Co., Ltd., Guangshan County, Xinyang City, Henan Province. Xinhua News Agency reporter
Li Anshe
Today, artificial intelligence is playing an important role in various fields. Especially in the e-commerce service, ticketing travel, health consultation, life shopping and other service experience ends, work efficiency has been greatly improved, and behind this is the support of artificial intelligence trainers. 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, including artificial intelligence trainers. It is understood that this is the second batch of new occupations released since the promulgation of the 2015 edition of the "Occupational Classification of the People's Republic of China". The latest analysis shows that it is estimated that by 2022, the number of domestic and foreign employees of artificial intelligence trainers is expected to reach 5 million.
The work is mainly oriented to the actual use of the product
According to the notice jointly issued by the Ministry of Human Resources and Social Security, the State Administration of Market Supervision, and the National Bureau of Statistics, the definition of artificial intelligence trainers is the use of intelligent training software, database management, algorithm parameter settings, and human-machine in the actual use of artificial intelligence products Interaction design, performance test tracking and other auxiliary operations personnel are mainly concentrated in the two fields of emerging industries and modern service industries.
Looking at the definition, it may be difficult for someone to distinguish the difference between artificial intelligence trainers and other artificial intelligence developers. In this regard, Duan Yucong, deputy director of the Special Committee of Hainan Provincial Blockchain Industry Research Institute and professor of Hainan University, analyzed that Intelligent trainers mainly use intelligent training software to work on the actual use of artificial intelligence products. Artificial intelligence developers mainly use development software, and their work is oriented to the development and design stage of artificial intelligence products; developers mainly complete functional and systematic work, and the goal of artificial intelligence trainers is to improve artificial intelligence. Product performance.
At present, the occupation of artificial intelligence trainers includes multiple types of work, the work content of different types of work is different, and the required skill focus is also different. Generally speaking, artificial intelligence trainers are mainly divided into data annotators and artificial intelligence algorithm testers. Data annotators have different requirements in different companies. Artificial intelligence algorithm testers have relatively high skills requirements, need to have the theoretical basis of relevant data processing, be familiar with related technical methods, be proficient in the use of development languages and simulation testing tools, and have experience in algorithm development or testing.
How do artificial intelligence trainers train artificial intelligence? Duan Yucong said that first, the artificial intelligence trainer needs to have a certain understanding of the data and knowledge involved, and then "cleanse" the data to obtain structured core knowledge and key data, specify data labeling rules, and "feed" the data to Artificial intelligence "tunes" it and constantly adjusts parameter optimization algorithms to allow robots to better serve humans. For example, by training AI to understand emotions, artificial intelligence can obtain sensitive information in human speech and text, and provide personalized and humanized services according to the user's status.
About 500,000 people are currently engaged in related positions in China
"The improvement of technology and productivity has greatly enriched people's daily lives, and the structure of social needs has also changed." Ge Heng, director of the Standards Division of the Vocational Skills Appraisal Center of the Ministry of Human Resources and Social Security, said that the artificial intelligence trainer is accompanied by New occupations arising from the application of emerging technologies meet the needs of economic and social development, meet the increasing needs of people, and conform to the development and changes of new industries, new formats, and new models.
New occupations and new positions created by the intersection of industry development have created a growing demand for the artificial intelligence trainer market. "Now the world is promoting the research and development of artificial intelligence technology. Artificial intelligence has penetrated into many industries, such as: data analysis, manufacturing, education, security, municipal administration, logistics, human-computer interaction, transportation, etc. And these new cross-industries, All need talent to participate in the realization, so the demand for artificial intelligence trainers is very high for the current society. "Said Deng Zhengjie, associate professor of Hainan Normal University.
In an interview with the media, Wang Zhiyu, an artificial intelligence training expert of the Alibaba Group customer experience business group, said that Alibaba has been incubating the first batch of artificial intelligence trainers in China since 2015. There are currently 200,000 artificial intelligence trainers in the entire Alibaba ecosystem Nearly 60,000 people have been trained and certified by the Ali artificial intelligence training system.
Wang Zhiyu mentioned that apart from e-commerce service groups like Alibaba, domestic and large AI companies basically have their own artificial intelligence trainers. "The entire industry involves AI companies, whether they are Party A, Party B, technology companies or service companies, they are actually starting to train their own artificial intelligence trainers. This is also a bit different from before." He revealed that there are about 50 in China. Tens of thousands of people go to such positions.
Data analysis shows that, according to the current industry development speed and the expansion of application fields, there will still be a large talent gap in the short term, such as data annotators; and with the development of technology, the requirements for the quality of practitioners are also constantly increasing. The gap in technical talents such as intelligent algorithm testers will be more prominent.
New landing scenarios require improving comprehensive quality
"We need to circle the objects in hundreds of thousands of photos every day, and then mark them out, such as flower pots, carpets, coffee tables, sofas ... These marked pictures will be sent to the database and become artificial intelligence learning materials. For example, when a computer looks at tens of thousands of objects marked with a sofa, it can recognize what a sofa is. "Xiaochen, a young man from Guizhou, described his work.
"For us, the biggest challenge for AI trainers is cross-border integration, which requires trainers to understand both legal knowledge and algorithm-related knowledge." Li Zhengcai, CEO of Beijing Wisdom Zhengan Company, said that our main requirements for candidates are Build a bridge between AI engineers and data annotators.
Although the requirements of different talents for artificial intelligence trainers are different, it is not difficult to see that the main tasks of artificial intelligence trainers at present include not only labeling and processing pictures, text, voice and other services, but also analyzing and refining the characteristics of professional fields. Train and evaluate related algorithms, functions and performance of artificial intelligence products, design interactive processes and application solutions of artificial intelligence products, monitor, analyze and manage application data of artificial intelligence products, adjust and optimize artificial intelligence product parameters and configurations, etc.
Duan Yucong believes that at present, people who are engaged in simple tasks such as data labeling are indeed relatively high among artificial intelligence trainers. However, with the rise of unsupervised self-learning algorithms such as artificial intelligence reinforcement learning, the dependence on data will be reduced, and the focus of artificial intelligence trainers will also change. In addition, in the future, if we want to develop data training artificial intelligence for application scenarios that have spiritual needs, such as painting, music, and literature, artificial intelligence trainers will need to continue to learn and improve their comprehensive quality.
The application prospect of artificial intelligence annotation for human emotions and other subjective feelings is huge. Duan Yucong suggested that, regarding the excavation of the human spiritual world, relevant departments should make legislative and standardization of artificial intelligence governance, and make advance preparations for the professional ethics of artificial intelligence trainers.
Xie Yun, Ph.D., Ph.D., Department of Electronic Engineering, Tsinghua University, put forward higher requirements on how to make a better career plan. He said that the most important thing for artificial intelligence trainers is to get the methods through. Different method principles are applicable to different types of problems. Continue to practice and ask yourself according to the standards of craftsmen. However, artificial intelligence trainers should not be limited to the artisan pattern, but should explore through experience to further summarize a deeper principle and create universal results.