The rapid development of artificial intelligence has also greatly increased the demand for talents in related fields. According to the data of the recruitment platform, in February this year, the recruitment volume of algorithm engineers, machine learning, deep learning and other positions increased significantly year-on-year, of which the recruitment demand for data labeling positions increased by more than 50% year-on-year.

At present, AI trainers are mainly engaged in annotation work. With the surge in market demand, this profession has attracted a lot of cross-border job seekers to pour in, and it has also led to related vocational training.

At a training institution for AI trainers in Shenzhen, students from all walks of life come to take classes.

According to the person in charge of the training institution, at present, artificial intelligence trainers are mainly engaged in annotation work, which can be subdivided into text annotation, image annotation, voice annotation and other types, with salaries ranging from 19000 yuan to 0 yuan per month. In the past two years, their training business has expanded from Beijing to Shenzhen, Hangzhou and Chengdu, but this is far from meeting the needs of the market.

Liu Meifang, principal of an artificial intelligence trainer training institution in Shenzhen, Guangdong: For example, if the resume is opened at 10 o'clock, our students can receive more than 0 replies within an hour, and many HR (human resources departments) take the initiative to greet them without needing to deliver them.

According to data from a recruitment platform, since the Spring Festival this year, the number of positions recruited for data labeling positions has increased by 56% year-on-year. Industry insiders said that with the rapid development of large language models and the increasing number of fields involved, the demand for artificial intelligence trainers will continue to increase.

AI Trainers: Coaches in the Digital Age

How exactly does an AI trainer work? What are the new requirements for trainers in the development of large models serving the professional field?

In Nanshan, Shenzhen, Guangdong Province, an artificial intelligence trainer Huang Peihui and his colleagues are practicing the effects of large models.

Huang Peihui said that the reason why the large model can accurately answer customers' questions is that more than 1000 similar questions have been entered into the large model in the early stage. For them, the process of screening customers to ask questions to customer service staff is to do "annotation", and these "questions" are "data", also known as "corpus".

In addition to such large models for public services, more and more large artificial intelligence models serving the professional field are also developing rapidly. At the Sixth Affiliated Hospital of Sun Yat-sen University in Guangzhou, Huang Yan, director of the Department of Pathology, is instructing the R&D staff of a large medical model to label digital pathology slides and provide training data for the pathology model.

In order to ensure the safety and accuracy of the data, the training of such large models must be checked by professionals such as doctors.

Sun Qigong, vice president of a medical large-scale model R&D technology company: Generally, senior medical experts are needed to guide or personally label the data, and misdiagnosis or missed diagnosis may occur if the mark is wrong.

In addition, in order to reduce the occurrence of "AI hallucination", that is, the content generated by the large model does not match the real data and deviates from the user's instructions, all vertical large models are building a multi-level annotation verification mechanism with the participation of experts in related fields, and converting their knowledge and experience into key training parameters to improve the accuracy of the large model.

Source: CCTV Finance