Today, a collaborative effort between Google DeepMind, Google Research, Google Cloud, and Google Labs introduced Co-Scientist, a multi-agent AI system designed to assist researchers in generating, debating, and evolving scientific hypotheses. This innovative tool, built on the Gemini framework, is poised to address the challenges of information overload in scientific discovery by producing thousands of hypotheses and refining them through a “tournament of ideas” and scientific debates. Since its inception, Co-Scientist has been tested on complex topics, yielding insights into liver fibrosis, approaches to Amyotrophic lateral sclerosis (ALS), and novel genetic leads for reversing aging. The system will soon be made available to individual researchers through a new feature called Hypothesis Generation.

Gemini: Gemini is Google’s family of multimodal AI models designed for reasoning, long-context handling, and tool use. Co-Scientist leverages Gemini’s core capabilities to power its agent coalition for hypothesis work. The model underpins the new tools aimed at accelerating scientific discovery.
Google Labs: Google Labs explores experimental AI applications and early-stage tools for broader audiences. It helped develop the Hypothesis Generation feature for Co-Scientist availability. The lab’s role emphasizes testing new AI approaches in scientific contexts before wider deployment.
Co-Scientist: Co-Scientist is a multi-agent AI system built on Gemini to generate, debate, and refine scientific hypotheses. It mirrors the research cycle of ideation, critique, and verification using specialized agents. The system is now being offered to researchers via an experimental tool in Gemini for Science.
Google Cloud: Google Cloud provides enterprise cloud infrastructure and AI services for developers and organizations. It supported the Co-Scientist rollout through its platform capabilities for hosting and scaling AI tools. The involvement aligns with Google Cloud’s work integrating advanced AI systems for research users.
Google DeepMind: Google DeepMind is Google’s primary AI research organization dedicated to advancing artificial intelligence capabilities. It played a central role in creating the Co-Scientist multi-agent system as part of the Gemini for Science initiative. The project draws on DeepMind’s strengths in building reasoning and agent-based AI tools for scientific applications.
Google Research: Google Research conducts foundational work across AI, computing, and scientific domains within Google. It contributed to the development and testing of Co-Scientist alongside other Google teams. The effort reflects Google Research’s focus on applying AI to accelerate discovery in biology and related fields.

AI Research: Co-Scientist was developed jointly across Google DeepMind, Google Research, Google Cloud, and Google Labs.
Research Access: Hypothesis Generation is rolling out as an experimental feature for individual researchers to register interest.
Scientific Applications: The system has been tested on challenges including liver fibrosis targets, ALS approaches, and genetic leads for aging reversal.