Nvidia, KAIST team up to develop Korean-language AI

ملخص مفصل

Nvidia and the Korea Advanced Institute of Science and Technology (KAIST) will jointly develop sovereign artificial intelligence optimized for the Korean language, aiming to…

Nvidia and the Korea Advanced Institute of Science and Technology (KAIST) will jointly develop sovereign artificial intelligence optimized for the Korean language, aiming to overcome shortages of training data and establish South Korea as a research hub for AI technologies serving languages with relatively few speakers.

The partnership is expected to strengthen South Korea’s sovereign AI ecosystem while helping Nvidia develop techniques that could be applied to other languages with limited training resources. Researchers say the initiative marks a shift in South Korea’s role from a major market for Nvidia’s graphics processing units to a partner in advanced AI research.

At a news conference held at Nvidia Korea’s Seoul office on Tuesday, the two organizations outlined plans for their joint AI research laboratory, including developing Korean-language datasets and making the resulting technologies freely available.

The announcement provided the first details of the laboratory’s research agenda since Nvidia and KAIST signed a memorandum of understanding in October to establish the facility.

The laboratory will be headed by Kim Hyun-woo, a professor at KAIST’s Kim Jaechul Graduate School of AI and a research manager at Nvidia.

Kim, who joined KAIST in September, holds positions at both the university and the U.S. technology company. The laboratory is currently recruiting full-time researchers, marking what the organizers described as the first time a global technology giant has directly recruited researchers in South Korea for such a joint research operation.

The initial research will focus on building high-quality Korean-language training datasets and developing methods to train AI models effectively using limited amounts of language data.

“AI performance is determined by data, but South Korea has far less training data available than English- or Chinese-speaking markets,” Kim said. “We need to develop methodologies to overcome this data gap.”

Korean, spoken by a relatively…

مطلع المقال، مقتطف من Pulse (Maeil Business). لقراءة المقال كاملًا، افتحه على موقع الناشر.