Technology & Humanity
Researchers Develop Conversational AI For Self-driving Labs
From Phys.org Technology · October 21, 2026
A research team led by Dr. Gwang-Noh Ahn at the Korea Research Institute of Chemical Technology (KRICT) has developed a human-in-the-loop optimization (HILO) framework. This technology uses large language models (LLMs) to translate a researcher's natural-language expertise into structured information that an AI can use for experimental planning in self-driving laboratories (SDLs).
To address the risk of incorrect human guidance, the team implemented an adaptive weight credibility detection (AWCD) mechanism. According to the researchers, this system monitors whether expert guidance aligns with experimental data and can automatically disable misleading information when contradictory evidence appears. In virtual tests, the framework significantly reduced the number of searches required to reach target performance in specific chemical reactions.
The team also released an interactive "Co-Scientist" interface to allow for real-time dialogue during the optimization process. The researchers intend to validate this technology using physical self-driving laboratory equipment.
A factual summary prepared by The Quantum Disciple. The original report belongs to its publisher.