본문 바로가기
bar_progress

Text Size

Close

LG Uplus-AI Institute Declares 'One Team AI': "Creating a National Representative LLM Will Be Competitive"

Developing On-Device sLM Based on Exaone
Achieving Both Security and Performance... Showing Confidence
"Providing AI That Customers Can Use With Peace of Mind Is a Telecom Operator's Duty"

LG Uplus-AI Institute Declares 'One Team AI': "Creating a National Representative LLM Will Be Competitive"

LG Uplus and LG AI Research Institute expressed confidence in AI technology development by advocating the 'One Team AI' strategy. Lee Sang-yeop, CTO of LG Uplus, responded to a question about participating in the government's selection of a national representative AI team by saying, "If we create a national representative LLM (Large Language Model) or sLM (small Language Model), One Team LG AI will be competitive."


On the 5th (local time), LG Uplus and LG AI Research Institute announced on the 9th that they unveiled the 'One Team AI' strategy in Barcelona, Spain, where the Mobile World Congress 2025 (MWC 2025) was held. The briefing was attended by Lee Sang-yeop, CTO of LG Uplus, and Lee Hong-rak, CSAI (Chief AI Scientist, Vice President) of LG AI Research Institute.


Last month, the government announced the 'National AI Capability Enhancement Plan,' aiming to develop world-class artificial intelligence (AI) models and stated its intention to select a 'national representative AI team.' The strategy is to consolidate dispersed corporate AI capabilities into a consortium and focus support on computing resources and research funding.


Currently, attention is focused on whether major companies with AI technology such as Naver, Kakao, LG, and SK will participate and how the consortium will be formed. In particular, the possibility of cooperation and the form of participation among companies that possess their own foundation models have emerged as key issues.


When asked about the intention to participate in the selection of the national representative AI team, the CTO said, "Rather than discussing decision-making from a technical perspective, the 'Exaone 3.5' 32B model of LG AI Research Institute, which was unveiled last December, was listed as a 'notable AI model' by the U.S. nonprofit research organization Epoch AI." He added, "LG Uplus is good at enhancing specialized performance with domain-specific data, so if we create a national representative LLM or sLM, One Team LG AI will be competitive." However, regarding the specific participation method, he said, "It is under internal discussion."


Both companies focused on 'on-device AI' as the core technology of the 'One Team AI' strategy. LG Uplus implemented this technology for the first time in the domestic telecommunications industry through the 'ixi-O' service. On-device AI technology processes user data directly on the device without transmitting it to external servers, greatly enhancing data privacy and security.


They are also developing an advanced version called 'on-device sLM' that uses an NPU (Neural Processing Unit) instead of a CPU. This upgrade is expected to reduce power consumption to one-quarter and CPU usage to one-tenth compared to the current implementation. This technology is planned to be based on LG AI Research Institute's 'EXAONE' model.


Another important pillar of the 'One Team AI' strategy is AI ethics and security. As part of efforts to enhance the ethicality of AI models, LG AI Research Institute has established an AI ethics governance system. They also developed a 'Data Compliance Agent' that can automatically analyze legal risks of training data. In addition, various ethical safeguards have been implemented in AI services, including personal information anonymization, automatic blocking of harmful content, and refusal to respond to violent or inappropriate questions.


Regarding concerns that focusing on safety and security might slow down AI technology development, the CTO responded, "I do not think that paying attention to assured intelligence (security-enhanced AI) slows down the pace of AI technological innovation." He emphasized, "It is important to train well with refined and high-quality data through distillation," adding, "Providing AI that customers can use with peace of mind is an obligation that a telecommunications operator must fulfill."


The CSAI said, "LG AI Research Institute has secured global-level technological capabilities from cutting-edge research to model development and will create human-centered agent AI through this. Through the trustworthy and reliable ixi-O, we will expand synergy as One Team with LG Uplus so that customers can use differentiated customized AI services."


© The Asia Business Daily(www.asiae.co.kr). All rights reserved.

Special Coverage


Join us on social!

Top