The Ministry of Science and ICT announced that three elite teams have passed the second evaluation of the 'Independent AI Foundation Model' project and will focus on enhancing model performance and real-world testing in the next phase. The strategy aims to narrow the technology gap with global frontier-level models while increasing practical usability.
On August 18, the ministry revealed the results of the second evaluation. The teams from Upstage, SK Telecom (SKT), and LG AI Research Institute advanced to the next stage. Motif Technologies, which joined through an additional application, was eliminated in this evaluation.
The three advancing teams plan to concentrate on enhancing model performance and conducting real-world testing in industrial and service environments.
Upstage aims to refine its model into one that the public can genuinely experience, utilizing the portal Daum and consortium partners. The company also plans to expand language support to Asian languages, laying the groundwork for entering the global market.
Upstage stated, "We are establishing an ecosystem where all citizens can benefit from AI while advancing Korean AI to the global stage. In the final testing phase, we will do our utmost to enhance the model that the public can experience through Daum and consortium partners." They added, "We will also expand language support to Asian languages to facilitate global dissemination."
SKT is focusing on creating a model that can be practically utilized in the domestic AI ecosystem, based on the success of its own AI model in various industries and everyday life.
SKT has been conducting real-world testing in the manufacturing sector by developing specialized agents for KG Steel and Connec, and in the defense sector through a memorandum of understanding with the Ministry of National Defense to provide lightweight models. The company is also collaborating with SK Biopharm in the bio sector. Moving forward, SKT plans to focus on reducing memory and processing speed burdens that arise during the operation of its proprietary AI model, 'A.X K2,' by companies and institutions.
LG AI Research Institute emphasized that it has secured a technological foundation for the stable training and operation of large-scale models and improved response quality and usability for actual users during the second phase.
In the third phase, the institute aims to further enhance the scale and performance of its models to compete directly with global frontier-level models. They also plan to hold a talk concert to reveal specific directions for technological development.
Im Woo-hyung, co-research director at LG AI Research Institute, stated, "Expanding the model size significantly to compete on par with global frontier-level models was a strategic choice we had to make. It is challenging to narrow the technology gap with global big tech using only small models, so we will continue to strive for performance that meets or exceeds that of models at the same scale."
Meanwhile, Motif, which did not pass the evaluation, has decided not to contest the results.
Im Jeong-hwan, CEO of Motif, remarked, "Regardless of the outcome, this project has proven that it is possible to develop AI models at a global frontier level in Korea based solely on technology. If the government had not decided to challenge the development of its own AI foundation model, opportunities for small startups like ours would not have existed, and we would not have achieved these results." He added, "Motif will continue to dedicate itself to developing global frontier models in the future."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.