Shulou(Shulou.com)11/24 Report--

CTOnews.com, October 9 (Xinhua)-- AI models are a hot topic in the field of artificial intelligence in recent years. They can perform a variety of amazing functions, such as generating realistic text and images, or having smooth conversations with humans. But behind these large models, there is a group of unknown data annotators who work hard to tag the original data every day, providing huge amounts of data for AI technology training.

The work of Tu Yuan Pixabay data annotators is not easy, they have to face boring tasks, low income, long-term instability and the risk of being replaced at any time. They are the cornerstone of the development of AI technology, but they receive little attention and respect.

According to Tech Planet, data annotators use the most primitive piece rate to calculate their wages, and most practitioners earn only 5000 yuan a month. Some of them are college graduates, some are baby mothers, and some are career changers. They process pictures, text, voice and other data in cubicles in third-and fourth-tier cities to provide services for Internet companies and car companies.

CTOnews.com notes that the data tagging industry has also experienced ups and downs. In 2017, when AI technology expectations were booming, data annotators could earn a lot of money, with 50 cents for a 2D frame. However, with the intensification of competition in the industry and the unsmooth development of technology, the unit price marked by the data is getting lower and lower, and now the lowest price is only 4 cents.

Data tagging companies are also under tremendous pressure. They need to have a certain scale and capital reserves to obtain orders from the source, and to bear the long payback cycle, high employee mobility, quality and cycle instability and other problems. Haitian AAC is currently the first motherboard listed company in the data tagging industry, with a profit margin of just over 10% last year and lost money in the first half of this year.

What worries data annotators even more is that they may soon be replaced by the AI they helped create. Some companies at home and abroad are developing tools that can label data automatically, using the mainstream large models in the market to label data sets. These tools claim to improve labeling efficiency and reduce costs, and achieve accuracy similar to or even higher than manual labor.

Of course, not all data annotations can be replaced by AI. Some data annotations that require professional knowledge and logical analysis capabilities still require manual participation, such as medical care, finance, self-driving and other fields. But it also means that the threshold of the industry will continue to rise, and it may take more study and effort for data annotators to survive in the industry.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.