KT's Supergiant AI 'Mideum' Expands Overseas... Collaborates with Thailand's Jasmine Group to Develop Large Language Model

After Developing a Thai-Only LLM, Collaborative Commercialization Efforts Across Southeast Asia

KT announced on the 23rd that it will begin building a Thai-LLM (Thai Large Language Model) using KT's super-large AI 'Mideum' in collaboration with Jasmine Group, a leading information and communication company in Thailand. After developing a Thai-specific LLM, the plan is to jointly commercialize it throughout Southeast Asia.

Through this collaboration, the two companies plan to carry out ▲ development of a Thai-specific LLM and business model ▲ analysis of the Southeast Asian market and establishment of marketing strategies ▲ transfer of LLM construction technology and know-how ▲ joint establishment of AI regulatory response measures for the Southeast Asian market. In particular, KT will transfer the technology and know-how necessary for building the LLM, while Jasmine Group will focus on building the GPU Farm, which serves as the foundation for Southeast Asian market analysis and model development. To this end, the two companies held a meeting between their representatives at the KT headquarters in Gwanghwamun, Seoul on the 19th and completed discussions on the commercialization of 'Mideum.'

KT and Jasmine Group will build a GPU Farm at a new IDC (data center) promoted by Jastel Co. Ltd, a 100% subsidiary of Jasmine Group, in the first half of next year, and then begin building the Thai-specific LLM from the second half, engaging in phased collaboration. Following this, based on the successful case in Thailand, they plan to jointly discover a global LLM business model and cooperate in joint commercialization in Southeast Asian markets such as Laos and Cambodia.

Global research firm Statista projected that the generative AI market in Southeast Asia will reach $7.6 billion by 2030. Machine learning platform company Dataiku, in an IDC InfoBrief report, forecasted that AI solution spending in Southeast Asia will increase significantly from $174 million in 2022 to $646 million in 2026.

KT views this achievement as an opportunity to tap into the so-called 'Sovereign AI' movement, where countries and companies aim to build their own LLMs to maintain full data sovereignty without being dependent on big tech companies. For example, the representative super-large AI in the U.S., GPT-3, is known to have 92.6% of its training data in English, 0.016% in Korean, and 0.013% in Thai. Due to the limited training data in non-English languages, there has been a limitation in reflecting the political and cultural contexts of those countries. KT plans to support Jasmine Group in securing AI sovereignty through building a Thai-specific LLM and further expand 'Mideum' to most global countries with demand for super-large AI.

To commercialize super-large AI, KT has been pursuing an 'AI full-stack' strategy that covers AI infrastructure, models, and application service areas. By integrating the AI semiconductor capabilities of Rebellion, in which it has made strategic investments, and the AI semiconductor driving software of More, KT is strengthening its infrastructure competitiveness. Additionally, last month, KT expanded its AI ecosystem by investing approximately 20 billion KRW in domestic AI startups Upstage and the education specialist Qanda, covering model and application service areas.

Kim Young-seop, CEO of KT, said, "Through this collaboration, we expect to contribute to strengthening Thailand's AI industry competitiveness together with Jasmine Group and successfully enter the Southeast Asian AI market jointly."

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