Koo Kwang-mo's Bold Bet... LG Invests $100 Million in Developing 'Mega AI'

AI Researchers Hold Non-Face-to-Face 'AI Talk Concert'
Mastering Large-Scale Data for Comprehensive Thinking and Judgment

Koo Kwang-mo, Chairman of LG

[Asia Economy Reporter Jeong Hyunjin] LG Group has decided to invest over $100 million (approximately 113 billion KRW) in the development of "super-large AI (artificial intelligence)." The goal is to develop AI at the level of the "top 1% human experts" that comprehensively thinks and judges like the human brain by learning large-scale data, and they are accelerating research and development.

LG AI Research Institute, LG's dedicated AI organization actively supported by Chairman Koo Kwang-mo, plans to directly enter fields mainly focused on by IT companies within six months of its launch, applying AI to global manufacturing company tasks such as software development, data analysis, and customer consultation. Through this, they intend to grow the AI sector as the main business to replace the smartphone business.

On the morning of the 17th, at the "AI Talk Concert" held online, LG AI Research Institute announced that it will invest more than $100 million over the next three years to secure and develop large-scale computing infrastructure capable of processing vast amounts of data quickly. Super-large AI is AI that, based on computing infrastructure, learns large-scale data volumes and independently thinks, learns, judges, and acts without being limited to specific purposes. In Korea, companies like Naver and Kakao, and overseas, Google, Amazon, and Microsoft are investing in super-large AI development.

Baek Gwang-hoon, head of LG AI Research Institute, said, "LG will actively build an open ecosystem for advanced super-large AI research, large-scale computing infrastructure, data acquisition, and commercialization together with global partners to secure world-class research capabilities."

Super-large AI, from Image Understanding to Data Inference

LG AI Research Institute plans to build and expand to the world's top level a global top 3-level AI computing infrastructure capable of performing 9.57 quadrillion operations per second for super-large AI development.

The super-large AI developed by LG AI Research Institute is characterized by its ability to understand not only language but also images and videos, and to perform data inference. LG has set a goal to unveil a super-large AI with 600 billion parameters, three times the 175 billion parameters of GPT-3, the most advanced super-large AI language model developed by the US AI research institute OpenAI, in the second half of this year.

Parameters play a role similar to synapses in the human brain that connect neurons to learn and remember information; the larger the parameter scale, the higher the AI intelligence. GPT-3 can have natural conversations like humans and create essays or novels, while the super-large AI developed by LG AI Research Institute can understand language, images, and videos, and also perform data inference.

LG AI Research Institute plans to develop super-large AI with trillion-scale parameters by the first half of next year. An LG official said, "This is expected to be the first case of such scale super-large AI development among global manufacturing companies."

Baek Gyeong-hoon, head of LG AI Research Institute, is presenting at the 'AI Talk Concert' held online on the 17th. (Photo by LG)

Utilization in LG Manufacturing Processes such as Product Process Shortening and Design

LG AI Research Institute plans to drastically change the way of working through super-large AI development. A representative example is its use in shortening the product development process. Previously, when developing new products, if modifications were needed from the initial product planning stage through design, engineering, and production, professional software developers had to redesign the process. By applying super-large AI, if the user interface is described in human language, AI can perform software coding, significantly shortening the product development process.

Additionally, super-large AI will be used as a virtual advisor summarizing customer consultation histories at customer centers, and AI capable of directly conducting sales related to contract signing with B2B customers will also be developed. Furthermore, super-large AI will automatically analyze 250 years of chemical research papers and patents to build a database, recognize molecular structure images within papers, extract physical property information from tables, and extract various experimental conditions from the text to build a comprehensive material database. It will also be applied to the development of anticancer vaccines, a new concept cancer treatment utilizing the human immune system.

Moreover, a "creative super-large AI" capable of collaborating with designers on product design and internal product engineering will be developed. For example, when inputting "robot vacuum cleaner design resembling a supercar," AI will automatically generate hundreds of design drafts, enabling designers to create detailed designs and engineering based on them.

This content was produced with the assistance of AI translation services.