SelectStar announced that it will begin the next phase of developing a national AI after the SK Telecom consortium, in which it participated, passed the first evaluation of the "Independent AI Foundation Model (Dokpamo)" hosted by the Ministry of Science and ICT. This project is a national project aimed at securing an independent AI foundation model that will represent South Korea, and the first evaluation was conducted based on the development achievements of five consortia over the past six months.
SelectStar, responsible for the data division within the SK Telecom consortium, built and supplied AI model training data based on its large-scale data construction and verification technology. This contributed to strengthening model security and mitigating bias risks. Furthermore, SelectStar leveraged its proprietary AI reliability verification solution, "Datumo Eval," and its experience with red teaming to support the assessment of potential risk factors in real-world service environments.
The upcoming Phase 2 evaluation will focus on improving the quality of post-training data, aiming to enhance the performance of the SK Telecom consortium's AX K1 model. SelectStar will be responsible for the overall design and validation of a dataset designed to enhance complex inference capabilities in STEM fields, a multimodal dataset for image and video understanding, and a red-teaming dataset to identify model vulnerabilities.
The SK Telecom Consortium is a collaborative effort led by SK Telecom, the host company, and includes industry partners such as Selectstar, Krafton, Forty2Dot, Rebellion, and Liner, as well as academic institutions such as Seoul National University and KAIST. The consortium aims to strengthen South Korea's AI competitiveness by expanding and enhancing models to the level of parameters in the trillions.
Kim Se-yeop, CEO of SelectStar, said, “Passing this first evaluation demonstrates that SelectStar’s data construction and reliability verification capabilities can be effectively utilized in key national AI projects,” and added, “We will continue to contribute to the implementation of a trustworthy national AI model.”
- See more related articles
You must be logged in to post a comment.