LG Corp. (003550) will begin full operation of an "autonomous laboratory" by year-end, where robots and artificial intelligence carry out the entire experimental process on their own, from discovering chemical compounds to analyzing them. The plan is to upgrade "expert AI" tailored to science, manufacturing and finance to solve difficult problems at industrial sites and gain an edge in the global race for technological supremacy.
LG AI Research held the "LG AI Talk Concert 2026" on the 14th at the Convergence Hall of LG Science Park in Magok, Gangseo-gu, Seoul, unveiling industry-specific AI models with expert-level problem-solving capabilities along with its blueprint.
In the science field, the autonomous experiment system "EXAONE Discovery" was introduced. The model goes beyond designing new materials and predicting synthesis results to conduct actual physical experiments. The institute will begin full operation by year-end of a fully autonomous laboratory that combines EXAONE Discovery, a model specialized in materials analysis, with experiment design and analysis AI and precision hardware such as robotic arms. It is already developing immersion cooling fluid materials for next-generation AI data centers with LG Chem (051910) and is expanding the scope of new materials discovery with GS Caltex (078930). In an earlier collaboration with LG H&H, it identified "Ramcidil," an ingredient that helps ease hair loss, from among 420,000 candidate compounds in a single day, and it is developing a next-generation oral peptide drug with D&D Pharmatech. Lim Woo-hyung, co-head of LG AI Research, said, "By having robotic arms and AI perform the prediction and analysis work that researchers must repeat countless times, experiment time can be dramatically shortened." By using AI to first conduct experiments with high verification value, researchers can reduce repetitive work with minimal attempts and focus on solving new problems.
AI specialized for manufacturing and finance was also introduced. "EXAONE Omni Inspect," a manufacturing quality inspection model, detects product defects after learning from only a few thousand images and requires no retraining even when appearances or components change, and it is scheduled to be applied to production processes within this year. "EXAONE BI," a forecasting solution for the financial sector, has multiple AI agents work together to generate everything from data analysis to inference, forecasting and explanation. Since its official launch early this year, it has been expanding its business to clients including the London Stock Exchange Group (LSEG) and Koscom.
Next-generation research to advance robot intelligence is also picking up pace. Moving a step beyond existing "action models" that simply imitate human movements, the institute has begun developing a "World Model" that understands and reasons about the laws of physics in the real world. Along with this, the institute officially announced the release of "K-EXAONE 3.0," the next version of its flagship proprietary foundation model.
Since its launch in December 2020, LG AI Research has had 368 papers accepted at top global AI conferences and filed 1,080 patents at home and abroad over the past six years. On that basis, it has resolved more than 100 difficult problems at industrial sites, including battery life prediction, defective product screening and drug candidate discovery, creating tangible business value.
