"We searched 420,000 chemical compounds and found a hair-loss drug candidate in a single day. Within this year, we plan to launch an autonomous laboratory in which AI handles the entire process of material analysis and experimentation."
Lim Woo-hyung, co-director of the LG AI Research, described the plan in a keynote speech at the LG AI Talk Concert 2026 held at LG Science Park in Gangseo District, Seoul, on the 14th, unveiling an upgrade roadmap for EXAONE Discovery, an artificial intelligence model specialized in analyzing new materials.
EXAONE Discovery learns from molecular structure data to quickly identify candidate compounds suited to a given purpose. Developing a new material normally means running repeated experiments on hundreds of thousands of candidate compounds to find one that fits the intended use. EXAONE Discovery replaces that manual screening work with AI, cutting development time and costs.
Working with LG H&H (051900), LG AI Research used EXAONE Discovery to identify Ramcidil, a drug candidate for hair-loss treatment, from among 420,000 compounds in one day. "Ramcidil emerged as a candidate compound that can activate scalp cells and boost energy metabolism," Lim said. "We are now preparing to commercialize it."
LG AI Research will take EXAONE Discovery a step further into an AI autonomous laboratory. The autonomous laboratory is a platform that combines EXAONE Discovery with a specialized AI laboratory model that handles experiment design and result analysis, along with hardware such as robotic arms that handle actual samples. The goal is to have AI take charge of everything from candidate screening to experiment design, execution and analysis, sharply shortening development times for new materials.
For the project, LG AI Research is partnering with LG Chem (051910) and GS Caltex (078930). GS Caltex plans to apply the autonomous laboratory to its research on coolants for data centers, which it has been pursuing using EXAONE Discovery.
"Rather than running every experiment, the autonomous laboratory first proposes the experiments most worth checking," said Han Se-hee, head of the Materials Intelligence Lab at LG AI Research. "It allows fast exploration with a minimum of experiments, so researchers can move away from repetitive work and focus on solving new problems."
LG AI Research also unveiled new expert AI models tailored to manufacturing and finance the same day. EXAONE Omni Inspect, for manufacturing, is a quality-inspection model that uses cameras to detect defects in products and components. Conventional quality-inspection AI had to learn from tens of thousands of images and required retraining whenever a product's appearance or its components changed. EXAONE Omni Inspect, by contrast, needs to learn from at most a few thousand images and requires no retraining when the product changes. LG AI Research plans to apply the model to actual manufacturing lines within this year.
In finance, the company is using EXAONE Forecast, which analyzes data that changes over time to predict the future. LG AI Research recently ran a proof of concept with Saudi Aramco, the Saudi Arabian state oil company, using EXAONE Forecast to predict crude oil prices, recording accuracy of 98.55%.
EXAONE Forecast is also being applied to EXAONE Business Intelligence, a financial agent that analyzes some 8,000 listed stocks, to improve its performance.
In robotics, the institute is developing a world model that understands the laws of physics. The goal is to move beyond conventional action models, which learn and imitate human behavior, and have robots learn for themselves how the physical world works.
"Ultimately, we plan to develop this into robot intelligence that learns the laws of physics on its own, without a person teaching it step by step," Lim said.
Separately, LG AI Research announced the coming release of K-EXAONE 3.0, its next model, which will be used in the government's sovereign AI foundation model project to select a national champion in AI.
