Nvidia and the Korea Advanced Institute of Science and Technology plan to release at least 1 million Korean-language AI training samples in 2027, the head of their joint research lab told reporters in Seoul on Tuesday, The Korea Herald reported. Seoul Economic Daily also covered the briefing.
Why it matters: Kim Hyun-woo, an assistant professor at KAIST’s Kim Jaechul Graduate School of AI and head of the lab, said the aim is better Korean training data and training methods, not a new foundation model. He did not give a specific release date in 2027.
“Our goal is to create at least 1 million Korean-language data samples and make them publicly available for training,” Kim said, according to The Korea Herald. He argued that Korea has invested heavily in models and computing but less in high-quality Korean text, and that translated or loosely filtered synthetic data can hurt performance. “Generating the data itself isn’t the problem. Filtering it for quality and curating what matters takes much more effort,” he said.
The lab, based at the Seoul graduate school, is the first concrete milestone since Nvidia and KAIST announced plans in July for what they called Asia’s first joint AI research lab between Nvidia and a university, backed by a five-year, $300 million collaboration, the Herald reported. Kim said Nvidia’s open Nemotron models will support the work, but that the datasets and methods are meant for broader use by Korean developers. Nvidia has also begun recruiting full-time researchers in Korea, with initial hires expected by year-end, the Herald said.