Researchers from Sweden and the United Kingdom have developed an AI-driven laboratory system capable of autonomously executing all stages of biological research. The platform formulates scientific hypotheses, designs experimental protocols, issues commands to automated lab equipment, and analyzes the resulting data to plan subsequent steps. During validation tests on yeast cells, the system—powered by the GPT-4o model—analyzed a dataset of 60,000 data points to generate 1,933 hypotheses, successfully confirming key predictions such as the synergistic inhibitory effect of arginine and caffeine on cell growth.

Notably, the system demonstrates adaptive learning, refining its future experimental models even when predictions prove incorrect. While human intervention is still required for manual tasks like solution preparation and sample transport, hypothesis selection and protocol design are fully automated. Researchers emphasize that AI is not replacing scientists, but rather evolving into a full partner in scientific discovery by dynamically shaping the direction of future research.