The Korea Intelligence Information Society Promotion Agency (NIA, Director Hwang Jong -sung) and the Korea Information and Communication Technology Association (TTA, Chairman Son Seung -hyun) announced on the 14th that they’ve released ‘AI Hub’ for the AI ​​Hub for AI (AI).
The models and data released this time were developed to measure ‘harmfulness’ in the standard verification technique of LLM data as a part of the event of the Early AI expansion ecosystem. Because of this of self -performance verification, he explained that the accuracy of 80-90%was achieved and the effectiveness of harmful expression detection has been proved.
The info consists of 200,000 data for detecting harmful expressions and 210,000 data for distinguishing harmful expression categories based on Korean harmful expression open source data.
As well as, it was said that the National Human Rights Commission also reference guidelines akin to guidance on disgusting expressions, defining hazardous expressions and categories, and releasing them to secure the consistency and reliability of information.
Hwang Jong -sung said, “It is anticipated that the disclosure of this artificial intelligence model and data might be a chance to strengthen the reliability and safety of Korean LLM.” “NIA will proceed to expand its support for artificial intelligence ecosystems by establishing and utilizing the muse for prime -quality artificial intelligence data that considers ethical responsibilities.”
By Park Soo -bin, sbin08@aitimes.com
