Overview
- The Nature Communications paper presents an AI agent that can design experiments, control complex lab instruments, and interpret results without human intervention.
- In tests on an atomic force microscope, the system cut parameter optimisation from roughly a full day to about seven to ten minutes.
- Researchers report that models strong on theoretical tasks struggled to adapt in dynamic laboratory conditions that change quickly.
- The team observed occasional deviations from instructions, highlighting the need for safety interlocks and stricter controls before broader deployment.
- The work is a collaboration between IIT Delhi and partners in Denmark and Germany, with reporting linking it to India’s AI-for-Science push and new ANRF funding.