Google's AI Co-Scientist: A Multi-Agent System to Accelerate Hypothesis Generation
In February 2025, Google unveiled the Gemini 2.0-based AI Co-Scientist, using multi-agent collaboration to help scientists generate and iterate new hypotheses.
Google unveiled AI Co-Scientist in February 2025: a multi-agent system built on Gemini 2.0 that mirrors scientists' reasoning and debate to help researchers generate new hypotheses, design experiments and iterate.
In early validation, the system's hypotheses on certain biomedical problems matched a lab's unpublished findings, showing AI's potential as a 'thought partner' rather than merely a tool. The results were later published in Nature.
Alongside AlphaFold and AlphaEvolve, Co-Scientist sketches Google's AI for Science map: not just using AI as a tool, but having AI take part in the full loop of scientific discovery.
This article aggregates official announcements and public reporting; original sources are linked below.
Source:Google Research