"While creating a good AI (artificial intelligence) model is important, the mission of the LG AI Research Institute is to solve difficult problems that have remained unsolved for a long time in industrial settings."
Im Woo-hyung, Co-Yeon Gu (Director) of LG AI Research Institute, stated this on the 14th at 'LG AI Talk Concert 2026' held at LG Science Park in Magok, Gangseo-gu, Seoul, regarding the direction of LG's 'Expert AI'.
The LG AI Research Institute is accelerating AI transformation in science and manufacturing by leading with Expert AI that directly solves problems in industrial settings. The focus has been on solving problems using industry-specific expertise and on-site data, then reflecting the results back into AI research to advance technology. Since its establishment in 2020, it has successfully solved over 100 industrial challenges. Going forward, the institute plans to focus on evolving AI from merely analyzing problems to technologies that can independently judge and act.
A representative achievement is new material development. Using the 'ExaOne Discovery' platform for new material discovery developed with LG H&H, the LG AI Research Institute analyzed over 420,000 candidate substances in a single day and identified 'Ramscidil', a substance effective in improving hair loss. This drastically shortened the typical 22-month cosmetic material development period. LG H&H is currently preparing to commercialize Ramscidil.
The role of Expert AI has also expanded from candidate substance exploration to actual experiments. On this day, the LG AI Research Institute unveiled 'AI Autonomous Laboratory' for the first time, which combines materials-specialized AI with robots. When AI sets experimental conditions, robotic arms and automated equipment carry them out, and AI learns the results to design follow-up experiments.
Park Jae-seop, AX (AI Transformation)Chu Jin (Team Lead) of R&D (Research & Development) at LG Chem, emphasized, "Robots are similar to the hands of researchers, while materials-specialized AI and data systems correspond to their brains." He explained that through role division, repetitive tasks are entrusted to AI and robots, allowing research teams to focus on formulating new hypotheses and creative problem-solving.
The scope of AI judgment has also expanded into manufacturing processes. 'ExaOne Tabular' predicts quality and defect possibilities with only small amounts of process data. At LG Innotek, retraining existing models previously required a minimum of 14 days and over 300 hours of production process data. However, through ExaOne Tabular, changed conditions could be reflected within a maximum of 50 hours.
Yoo Jeong-seon, in charge at LG Innotek, introduced that the lead time (required time) for responding to models was reduced by approximately 85%. In the case of 'ExaOne Omni Inspect', inspections can continue without separate retraining even if products or processes change. This reduces the burden of relearning AI.
Representative cases of problem-solving by the LG AI Research Institute include battery life and capacity prediction, defect product selection, and discovery of new drug candidate substances. The LG AI Research Institute has published 368 papers at top global AI conferences and filed 1,080 patents domestically and internationally.
President Im stated, "For the past six years, LG has enabled AI to understand industrial problems, and now it is preparing for the stage where AI can independently judge and act." He added, "Based on the robot foundation model, we are moving toward a future of autonomous factories where entire factories move as one intelligence beyond individual robot automation."
in INDUSTRY
