Amid growing demand for acquiring and utilizing artificial intelligence (AI) learning data, broadcasting content is being developed as data for AI learning.
The Ministry of Science and ICT (Minister: Yoo Sang-im, hereinafter referred to as ‘MSIT’) and the Korea Radio Promotion Association (President: Hong Bum-sik) announced that they will invite public bids for the “Broadcasting Video AI Learning Data Construction Project” from June 5 (Thursday) to July 4 (Friday) in order to accelerate the transformation of broadcasting media into AI and support the development of Korean AI models.
A total of 20 billion KRW is being invested in this project. The MSIT plans to select four consortiums consisting of broadcasters, AI technology companies, data processing firms, and research institutions, providing each with 4.83 billion KRW in support.
The selected consortiums must secure more than 10,000 hours of video held by broadcasters. Among the secured content, scenes without copyright or privacy issues will be selected to construct over 5,000 hours of AI learning data. This data will include various information such as the tone, expression, and background of individuals, and will be refined and processed to enable AI learning.
The MSIT will conduct step-by-step verification through a specialized agency to ensure the quality of the data. The entire data construction process will be reviewed, and testing with AI models will also be conducted. Consortia must also develop AI technologies applicable to broadcasting content production and services using the constructed data.
The MSIT plans to provide the constructed data for the “World Best Large Language Model (LLM)” development project. Some data will also be opened for use in AI-related research and education.
The transaction of AI data based on broadcasting videos will be expanded. Although transactions of broadcasting video AI data have not been active due to insufficient trading systems, the current consortium plans to disclose the status of broadcasting video AI learning data and establish data trading standards to promote data transactions.
Kang Do-sung, Director of Broadcasting Advancement Policy at the MSIT, stated, “Broadcasting videos accumulated over 70 years by domestic broadcasters are considered optimal data for training Korean AI models due to their rich content in language and actions,” and added, “We will actively support domestic broadcasting content to be utilized as key data in AI development.”