LG AI Research's 'K-EXAONE' passed the second-stage evaluation of the Ministry of Science and ICT-led 'Independent AI (Artificial Intelligence) Foundation Model Project' on the 18th.
LG AI Research stated, "We focused on laying the foundation for developing flagship-level ultra-large-scale foundation models capable of competing in the global market." To achieve this, they expanded the model size by more than three times compared to previous versions and tackled technical challenges across training and inference of ultra-large models.
The company explained that this second phase is significant as it secured a technological foundation for stably training and operating ultra-large models while improving response quality and usability performance that users can actually experience.
LG AI Research plans to develop 'K-EXAONE' into a frontier-level expert AI model representing Korea based on the technologies acquired in this phase.
Im Woo-hyung, Co-Fellow at LG AI Research, said, "Significantly expanding the model size is an inevitable strategic choice to compete equally with global frontier-level models." He added, "Since it is difficult to narrow the technological gap with global big tech companies using only small-scale models, we will continue our challenge with the goal of achieving equal or superior performance at the same scale tier."
He further stated, "While the performance improvement compared to the first phase may appear somewhat gradual, this is a necessary technical verification process for leaping toward frontier-level AI that LG aspires to achieve." He noted, "Especially, this second phase involved optimizing training of ultra-large models within a limited timeframe. Based on the technical assets and know-how accumulated in this phase, we will successfully achieve our final target performance."
LG AI Research plans to hold a talk concert in one month to provide detailed explanations of future technology development directions.
in INDUSTRY
