Who Will Take the First Leap
in the AI Business?

What was the motivation behind the establishment of LG AI Research, often referred to as the LG Group’s “AI think tank”?

The LG Group launched LG AI Research in December 2020, aiming to lead globally in both fundamental and applied AI research while also enhancing the LG Group’s overall AI capabilities. The core goal is to solve the challenges faced by LG’s affiliates using AI and to develop commercially valuable technologies across various industries. Since its inception, LG AI Research has focused on overcoming the technical limitations of deep learning and expanding enterprise AI research. In December 2021, we announced the development of the large-scale AI model EXAONE (EXpert AI for EveryONE).

Currently, LG is the only company in Korea that has commercialized both a bilingual Korean-English model and a bidirectional image-text multimodal model. With the launch of EXAONE 3.0, which is on par with global open-source AI models, we are committed to strengthening further our global competitiveness and advancing the AI research ecosystem.

This seems like an approach that is difficult to find in other companies.

The LG Group also faced many challenges at the beginning. During the early days of the company’s launch, there was lack of domestic AI talent and even an understanding of AI, giving rise to serious difficulties. With the belief in seizing future technologies, however, the group has spared no effort in supporting this initiative over the past four years. As a result of such investment and effort, LG AI Research has been at the center of accelerating AI transformation across LG’s business sectors including production lines, product development, and customer service within LG’s affiliates.

What sets LG’s AI platform apart?

From the beginning, LG AI Research focused on “expert AI” to enhance industrial competitiveness. For AI to be used in business, it needs to possess expertise, reliability, and cost-efficiency. That’s why we collaborated with global business firms to train the AI using specialized data, ensuring that the data was free from copyright issues. Based on this, we developed models and technologies tailored to key industries and verified our technological capabilities through collaborations with the Ministry of the Interior and Safety and the Korean Intellectual Property Office.

In 2023, we introduced three platforms based on EXAONE 2.0, with the goal of creating synergy between humans and AI. These are: EXAONE Universe, a conversational AI platform for professionals; EXAONE Discovery, a platform for accelerating the development of new materials, substances, and drugs in the chemical and bio fields; and EXAONE Atelier, a multimodal AI platform that aids human creativity. These platforms are not intended to replace humans but seek to complement professional expertise and enhance work efficiency, enabling individuals to focus more on creative aspects—this is a different direction compared to other generative AI models.

Starting August this year, we launched the open beta service for ChatEXAONE for LG employees. Built on EXAONE 3.0, this service is an “Enterprise AI Agent” that offers real-time web-based Q&A, document and image-based Q&A, coding, and database management, providing functions and specialized insights tailored to users’ job roles. With ChatEXAONE, LG employees can utilize AI for a wide range of tasks from searching, summarizing, translating, and data analysis to report writing and coding. ChatEXAONE will lay the foundation for all essential components of an enterprise AI system, secure application cases, and bring about changes in how we work.

LG has achieved management innovation through the adoption of AI in a relatively short period. Could you share some relevant examples?

Since the release of the first version of LG’s large-scale multimodal model EXAONE in December 2021, LG AI Research has continued to improve its performance through ongoing research and development. Currently, EXAONE is applied across various industries within the LG Group, and its effectiveness has been proven.

First, in collaboration with LG Innotek, we developed an AI-powered vision-inspection system that can accurately identify defective products with only a small amount of normal data. By utilizing this new system, the lead time for building a defect selection process is reduced by 90% compared to the traditional use of data, significantly improving the efficiency of production processes. Second, we provided core technology to LG Uplus’s generative AI ixi-GEN, successfully leading the development of an AI model specializing in the telecommunications field. ixi-GEN is a telecommunications-focused AI model trained on a vast amount of LG Uplus data and is used in various areas, including Q&A related to telecommunications services and customer support. Lastly, with LG Display, we developed a customized AI solution aimed at improving the quality of display products. In this process, EXAONE was utilized to learn and analyze knowledge specific to displays. For example, when asked, “How can we improve the quality of ○○?” EXAONE analyzes display-related data and suggests optimal solutions, thereby enhancing the work efficiency of employees.

The most important factor for generative AI to be utilized effectively in industrial settings is high performance in real-world cases. While many high-performance models are introduced in various benchmarks, it’s rare to see models that demonstrate superior performance in actual field use. The recently released EXAONE 3.0 from LG AI Research is a model focused on delivering high performance in real-world use cases. This means that EXA-ONE excels in key aspects of human preference such as multi-turn conversations, task execution, reasoning, and mathematics, and, when combined with domain-specific documents and data owned by experts and enterprises, it can generate even better results.

Moving forward, LG AI Research plans to accelerate innovation by continuing to collaborate with LG’s affiliates and various partners, all centered on EXAONE 3.0.

Last December, the company drew attention by signing an MOU with world-renowned nonprofit genomic research institution The Jackson Laboratory (JAX).

The Jackson Laboratory chose to collaborate with LG, which possesses specialized AI technology for industrial applications, instead of focusing on the general AI models offered by global big tech companies. Our goal is to develop AI models tailored specifically for the medical field. Through joint research, the two companies are combining The Jackson Laboratory’s extensive medical data and LG’s AI technology to explore new possibilities in identifying the root causes of diseases such as Alzheimer’s and cancer and developing effective treatments. Currently, we are focusing on creating an AI model that can analyze the causes and progression of Alzheimer’s, predict the effectiveness of treatments, and assist in the development of new drugs and therapies. Additionally, we are working on a multimodal generative AI model that can diagnose cancer quickly using only pathology images and predict treatment outcomes, as well as a generative AI-based conversational agent that suggests personalized approaches to physicians based on an individual’s genomic characteristics. The synergy between the two organizations is expected to open new doors in the medical AI field and lead the global medical AI market.

What are your thoughts on the current status of the generative AI ecosystem and its implementation in the private sector?

Generative AI has now moved beyond the hype cycle and entered a phase where it must prove its practical value. Companies need to demonstrate how they can solve problems and create new value by utilizing generative AI.

According to a McKinsey Global Survey released in May, the number of companies adopting AI has increased significantly from 2019 to 2023. Even in fields that make the most use of generative AI such as marketing, sales, and IT, however, only 4% are actually using AI in production processes. This suggests that there are still technological, economic, and organizational barriers to the adoption of generative AI.

Still, this phenomenon is part of a necessary stage for the widespread adoption and application of generative AI. It is important to recognize that there is a performance gap between research-level AI and AI technologies implemented in actual industrial settings. If industries prepare well by customizing models, data, and managing risks, generative AI has the potential to fundamentally change how we work and live, creating new value in the process.

How do you view the adoption of AI in Korea?

While there is active discussion about the adoption of AI technologies including generative AI, the reality is that the pace is slower compared to US companies. I believe we should start by actively using AI technologies in smaller areas and boldly adopting the necessary technologies. Korea is known for its high receptivity to new technologies and ability to apply them quickly in practice. In the field of generative AI, however, these strengths have not yet been fully realized. To strengthen global competitiveness, proactive efforts are needed to create and support an environment for AI technology adoption through collaboration between the government, academia, and private sector.

What are the reasons companies face difficulties in adopting AI?

For companies to achieve continuous growth through AI, a data-centered approach is essential. Companies must accurately identify the problems that AI should solve and secure and refine high-quality data to address those problems. Moreover, it is crucial to strengthen the security of AI systems to minimize risks such as data breaches. With these efforts, companies can leverage AI technology to create new business opportunities and secure a competitive edge.

Could you share a message for CEOs regarding AI initiatives in their companies?

AI is developing rapidly, and has the potential to be applied across all areas. CEOs must strive to ensure that AI naturally integrates into everyday life and, further, into each business sector. Achieving even a 10% improvement in productivity through AI can provide companies with significant competitive advantages. CEOs should take the time to understand AI themselves and lead their organizations’ digital transformation to prepare for the future.