LG AI Research has announced an expanded partnership with Google Cloud to advance its EXAONE 3.0 generative AI models and the enterprise AI agent ChatEXAONE. It will leverage Google Cloud’s AI-optimized infrastructure for training and development.
This collaboration builds on their previous work, which began in December 2021 with the development of EXAONE 1.0, South Korea’s first bilingual and multimodal generative AI model. It continued with EXAONE 2.0 in July 2023, aiming to drive innovation for real-world industrial applications.
As part of the latest announcement, LG’s EXAONE 3.0 models and the ChatEXAONE AI agent were trained using Google Cloud’s advanced Cloud TPUs and GPUs, integrated into its AI Hypercomputer.
The collaboration enables LG to advance in scalability and performance, addressing the computational needs of these sophisticated models. By leveraging Google’s top-tier AI accelerators, LG has reduced inference processing time by 56%, memory usage by 35%, and operating costs by 72% with the EXAONE 3.0 7.8B model compared to its predecessor.
The EXAONE 3.0 model, which supports both Korean and English, has performed well in Korean language tasks. Hwayoung (Edward) Lee, Lead of AI Biz Dev and DX Group at LG AI Research, noted that EXAONE 3.0 was trained on an extensive dataset comprising 60 million pieces of specialized data, 350 million images, patents, software codes, and other critical information. “Google Cloud’s AI Hypercomputer has been essential in achieving our goals for model efficiency and cost-effectiveness,” Lee stated.
Security considerations were crucial in LG’s decision to partner with Google Cloud. Given LG’s involvement in sectors with national technology interests, such as large batteries, displays, and 5G, ensuring data governance and privacy was a top priority. Google Cloud’s comprehensive data principles and MLOps pipeline offered LG the strong infrastructure required to maintain data integrity while developing ChatEXAONE.
Additionally, LG plans to list EXAONE 3.0 on the Google Cloud Marketplace and is exploring the development of APIs designed to meet the growing needs for enterprise AI solutions. With this, the company plans to broaden the accessibility and functionality of LG’s AI models, catering to a wider range of industrial applications.
Ki-sung Chi, Country Director of Google Cloud Korea, commented on the new partnership: “Our collaboration with LG AI Research demonstrates the power of joint efforts in advancing AI technology. By providing LG with our cutting-edge AI infrastructure, we are not only accelerating their groundbreaking work but also influencing the future of industries on a global scale.”
Launched in December 2020, LG AI Research, the AI research hub of South Korea’s LG Group, is dedicated to advancements in artificial intelligence. The research center’s flagship project, EXAONE, a 300-billion parameter multimodal AI model, was introduced in 2021.
EXAONE, which stands for “Expert AI for Everyone,” has the capability to process both textual and visual data, aiming to enhance decision-making through its vast data-learning capacity and cutting-edge AI technologies.
Google Cloud provides a comprehensive suite of AI, infrastructure, and development tools designed to support business solutions. With its advanced AI stack, including custom-built chips and generative models, Google Cloud serves as a crucial partner in LG’s AI initiatives.
Also Read:
- SK Hynix Unveils World’s First 16GB DDR5 Chip with 6th-Gen Technology
- SK Telecom, KT, and LG Uplus Unite to Standardize APIs, Streamline App Development
- Samsung Promises Seven Years of OS Updates for New AI Smart TVs
- Hanwha Aerospace Launches First Overseas Production Facility in Australia
- Kakao Announces Merger of DK Techin and Kakao Brain
