Boiko, D. A., MacKnight, R., & Gomes, G. (2023). Emergent autonomous scientific research capabilities of large language models (arXiv:2304.05332)

Uses GPT-4 in a simple agent arrangement to perform chemical synthesis in a cloud lab.

The agent was able to design and submit instructions for an experiment “to perform Suzuki and Sonogashira reactions.”

They were also able to get it to synthesize several dangerous/restricted substances; see AI misuse.