LG Uplus Advances On-Device AI with Lightweight Model Development

LG Uplus is making significant strides in the artificial intelligence landscape with its commitment to developing "human-centric AI," a strategy that places a premium on ethical usage and the stringent security of user data. In collaboration with LG AI Research, the company is actively refining its on-device AI capabilities, aiming to deliver enhanced privacy and efficiency to its customers.

The Rise of On-Device AI and the ixi-O Service

The focus on on-device AI is a strategic pivot away from traditional cloud-based AI services. Unlike cloud solutions that require data transmission to external servers, on-device AI operates directly on the user's smart device. This inherent characteristic significantly reduces the risk of personal data leaks, a growing concern in the digital age. LG Uplus has already demonstrated its pioneering spirit in this domain with the successful launch of ‘ixi-O’, recognized as South Korea’s first AI agent service built upon on-device AI technology. This initial success has paved the way for further innovation.

Developing a Lightweight Small Language Model (sLM)

Building on the foundation of ixi-O, LG Uplus and LG AI Research are now concentrating on the development of an advanced on-device small language model (sLM). This new model is being engineered to leverage a Neural Processing Unit (NPU), a specialized hardware component designed for efficient AI computations. The integration of an NPU is crucial for optimizing the performance and power consumption of AI models operating directly on devices.

Enhanced Efficiency and Reduced Power Consumption

According to Lee Hong-lak, Senior VP and chief scientist at LG AI Research, the development of this new sLM-based ixi-O is projected to result in a remarkable reduction in power consumption. The goal is to achieve an efficiency level that is one-fourth of its current operational cost. This substantial improvement in power efficiency is a critical advancement, particularly for mobile devices where battery life is a key user consideration. LG Uplus plans to finalize the development of this cutting-edge on-device AI model within the first half of the year. Subsequently, it will be progressively integrated into mobile devices slated for release in the latter half of the year, marking a significant upgrade in the AI capabilities offered to consumers.

A "One Team AI" Strategy for Enhanced Customer Convenience

The collaborative effort between LG Uplus and LG AI Research, branded as their ‘One Team AI’ strategy, was prominently showcased at the ‘Mobile World Congress (MWC) 2025’. This strategy underscores a unified approach to AI development, with the overarching aim of enhancing customer convenience through advanced AI solutions. By focusing on on-device AI, the companies are not only prioritizing security but also aiming to create AI services that are more responsive and seamlessly integrated into users' daily lives.

Expanding AI Applications: Security and Generative AI

Beyond the core sLM development, LG Uplus is actively exploring and integrating AI into various aspects of its services. The company is developing AI-driven anti-phishing technology, a crucial security feature designed to combat the rise of voice phishing and fraud facilitated by AI-generated deepfake technologies. This technology, commercialized globally as an on-device AI solution, can detect AI-manipulated voices during calls within seconds, alerting users to potential threats. This initiative highlights LG Uplus’s commitment to leveraging AI for enhanced user safety.

Furthermore, LG Uplus is embracing generative AI with its ixi-GEN model, which is optimized for telecommunications. This model, built upon LG AI Research’s Exaone, offers high security and efficiency, positioning LG Uplus as a contender in the business-to-business (B2B) generative AI market. ixi-GEN can be deployed on local servers, minimizing security risks, and is trained on clean data to reduce copyright infringement concerns. The company plans to integrate ixi-GEN into eight of its services, including chat agents and mobile shop advisor services, by the end of the year. To further democratize AI, LG Uplus has also introduced the ‘ixi Solution’ platform, an AI development tool designed to make AI accessible for practical, real-world applications by both professional developers and ordinary users.

Strategic Partnerships and Future Vision

LG Uplus is also forging strategic partnerships to broaden the reach and impact of its AI initiatives. A notable collaboration is with Samsung Electronics to jointly develop an AI smartphone. This project aims to embed LG Uplus’s AI capabilities, such as the ixi-O call assistant, into Samsung’s Galaxy series of smartphones, potentially creating a "real AI phone" that transcends current AI assistant functionalities. While the new AI smartphone is initially targeted at LG Uplus users, the intention is to eventually make these AI services available to users of other mobile carriers.

The company’s commitment to AI is further underscored by its significant investment plans, with an annual investment of 400 billion to 500 billion KRW in AI technology, projecting a cumulative investment of 2 to 3 trillion KRW by 2028. This substantial financial commitment reflects LG Uplus’s strategic vision to transform into an AX (AI Transformation) company. The core philosophy driving this transformation is a customer-centric approach, focusing on delivering tangible value and practical AI-powered services that enhance daily life, rather than just showcasing advanced technology.

LG Uplus’s multi-faceted approach to AI, encompassing on-device sLM development, generative AI integration, enhanced security features, and strategic partnerships, positions it as a forward-thinking player in the rapidly evolving AI ecosystem. The company’s dedication to a human-centric approach ensures that technological advancements are aligned with user privacy, security, and overall convenience.

Tag: lg uplus

LG Uplus Pioneers On-Device AI with Lightweight Model for Enhanced Privacy and Efficiency