A system capable of generating ideas, running experiments and writing up results, including a negative finding that passed an initial round of peer review for a workshop.
The development is significant less because of one result than because it raises questions about authorship, reproducibility and how scientists should evaluate machine-generated research.