SKT Advances Multi-LLM... Targeting Generative AI Market for Enterprise and Public Use

AIDOT, Entropic, Conan, and others based on multi-LLM
Utilizing large-scale AI infrastructure and LLM partners
Promoting discovery of diverse industry customers including government ministries

SK Telecom announced its ‘Multi LLM (Large Language Model)’ strategy on the 23rd to target the ‘generative AI market’ for enterprise and public sectors. SKT stated that it is preparing technologically and commercially to provide customized services by combining ▲ its own A.Dot LLM, developed mainly for telecom-based services such as conversations and customer centers ▲ Anthropic’s LLM, which excels in ethical responses and large-scale text input ▲ and Konan LLM, rich in Korean language data.

The LLM foundation model is AI software equipped with basic functions such as document summarization, document generation, and Q&A. SKT revealed plans to promote a ‘Multi LLM’ business lineup that includes not only its own LLM but also industry-recognized third-party LLMs. It is expected that actual service provision will be possible from the fourth quarter when the integration of functions is completed.

Providing Full Stack Services with Strong Infrastructure and Ecosystem

AI services including LLM require stable GPU server infrastructure along with data for pre-training to ensure reliable service delivery. SKT has experience operating nationwide IT infrastructure and possesses a large-scale AI developer pool, enabling a ‘Full Stack service (Approach)’ in the enterprise AI field, from hardware infrastructure to LLM software. Recently, SKT made a strategic investment to become the third largest shareholder in Persona AI and is promoting cooperation in building LLMs for the AI Contact Center field, thereby establishing and expanding an ecosystem for various LLM business initiatives.

Aiming to Provide Industry-Specific Specialized LLM Services

The enterprise LLM services SKT aims to provide focus on matching each LLM service to the fields where its unique strengths can be best utilized. For example, A.Dot LLM has advantages mainly in conversational or chatbot services due to its trained data and service characteristics. Anthropic LLM has the feature of providing responses aligned with human-defined ethical principles, as it has been pre-trained on AI ethics. Additionally, it can input a large volume of text equivalent to a novel (100,000 tokens, 75,000 words) at once.

Discovering Customer Demand Across Various Fields

SKT is preparing to participate in pilot and main projects based on LLMs promoted by government bodies such as the Ministry of the Interior and Safety and the Ministry of Science and ICT. It is also advancing LLM-based service construction projects with individual clients, mainly in the financial sector. Discussions are underway with large manufacturing affiliates within the SK Group to apply LLM-based AI technology in manufacturing domains. Kim Kyung-deok, CIC Head (Vice President) leading SKT’s enterprise business, told internal members attending the multi LLM training course, “The important thing is the customer’s needs; it is crucial to quickly provide services that customers actually require based on various multi LLM combinations.”

This content was produced with the assistance of AI translation services.