AI Times, 19 Jun 2025
The Korean AI landscape has seen a surge in interest surrounding on-device AI processing, driven by factors like data privacy concerns and the need for real-time responsiveness. This trend aligns with the global push towards edge computing, but with a uniquely Korean flavor influenced by the dominance of mobile technology and the rapid adoption of 5G. According to the article, Eastsoft, a prominent Korean software company specializing in AI, has entered into a supply agreement with EQBR, a company specializing in financial infrastructure system integration. This partnership aims to deploy Eastsoft’s ‘AI Live Chat,’ a conversational AI human solution, within kiosk environments.
Eastsoft will be providing EQBR with a lightweight version of their ‘Allen LLM’ and their AI Human technology packaged as a Full On-Device SDK. This approach is noteworthy because it emphasizes on-device processing, unlike cloud-based solutions that rely on server communication. This is particularly relevant in Korea, where data security and low latency are highly valued. The initial rollout targets kiosks, with a projected volume of 2 billion KRW. This focus on kiosks suggests potential applications in various sectors, including finance, retail, and public services, mirroring the trends we see with companies like Samsung SDS and LG CNS exploring similar AI-powered kiosk solutions.
Technically, optimizing a Large Language Model (LLM) for on-device deployment requires careful consideration of resource constraints. The article mentions the utilization of a smaller, “sLM” (small Language Model) variant of Allen LLM, likely achieved through techniques like knowledge distillation or pruning. Furthermore, the integration of Retrieval Augmented Generation (RAG) suggests the system can access and process external knowledge bases to enrich its responses. This is crucial for providing accurate and contextually relevant information in a kiosk setting, where users might inquire about specific products, services, or procedures. The inclusion of Text-to-Speech (TTS) capability further enhances the user experience by enabling natural language interaction.
This partnership reflects the growing trend in Korea of incorporating AI into everyday interactions. While similar deployments are occurring globally, the Korean market’s emphasis on speed and convenience, coupled with strong government support for digital transformation initiatives, creates a unique environment for rapid adoption. The success of this project will likely depend on factors such as the accuracy and responsiveness of the sLM in real-world scenarios and public acceptance of AI-driven interactions in kiosk settings. Will this mark a significant shift towards more personalized and efficient customer service experiences in Korea, or will there be challenges related to user privacy and data security that need to be addressed? The coming months will be crucial in answering these questions.