SK Telecom has announced the release of A.X K1 artificial intelligence model, first of its ultra-large model in South Korea, which has 500 billion parameters. The announcement was made at the 'First Presentation of the Independent AI Foundation Model Project', hosted by the Ministry of Science and ICT on December 30th.
A.X K1 is designed with 519 billion parameters in total. When it performs tasks based on user requests, around 33 billion parameters are activated. It can operate at a smaller scale when necessary, making it both powerful and flexible.
AI models with over 500 billion parameters, like A.X K1, are known to be better at managing difficult tasks such as mathematical reasoning as well as multilingual learning compared to smaller models. They also improve capabilities in areas like high-level coding and autonomous task management, which means the AI can perform tasks on its own, such as sending emails or creating documents without specific instructions, along with asking for more information whenever required.
A.X K1 can also act as a 'Teacher Model', sharing its insights with smaller models (under 70 billion parameters). These aids bolster the surge of the artificial intelligence robot ecosystem in South Korea.
SK Telecom plans to make A.X K1 available through A-dot, which has over 10 million subscribers, creating an 'Everyone's AI' environment where anyone in the country can easily use AI via phone, text, web, or app.
In addition to general use, A.X K1 will be applied in areas like business productivity, manufacturing AI, game AI, and humanoid robots. It will also help test the performance of AI semiconductors, supporting South Korea’s semiconductor industry.
This new AI model is expected to improve both the daily lives of citizens and the competitiveness of industries in Korea.
Executive Statement
According to SK Telecom, CEO Yoo Young-sang, the 'A.X K1' from the elite team is the first ultra-large model in Korea. SK Telecom explained that this model is significant as it marks a step up for Korea's AI model to compete on the global stage, where the competition among the three AI powerhouses, the U.S. and China, is fierce.
