source: google deepmind: gemini for science: ai experiments and tools for a new era of discovery
level: technical
google deepmind and google research unveiled gemini for science, a set of experimental tools on google labs aimed at accelerating the scientific method. the tools target three core research tasks: hypothesis generation, computational discovery, and literature insights. hypothesis generation uses a multi-agent system to produce and debate research ideas with verified citations. computational discovery runs thousands of code variations in parallel to test modeling approaches. literature insights structures scientific papers into searchable tables and creates summaries, reports, and audio overviews using notebooklm.
the tools are built on existing ai systems like co-scientist, alphaevolve, and empirical research assistance. enterprise versions are already in private preview with partners such as basf, klarna, and daiichi sankyo. google also launched science skills, a bundle integrating over 30 life science databases into google antigravity, enabling complex bioinformatics workflows in minutes. early tests showed a speedup from hours to minutes for rare disease analysis.
google is collaborating with over 100 institutions, including stanford and imperial college london, to validate the tools. a trusted tester community of researchers and nobel laureates stress-tests the systems. the company also works with conferences like neurips on agentic peer review tools. this builds on prior ai work like alphafold, used by millions of researchers. the goal is to help scientists handle growing data volumes and focus on high-impact problems.
why it matters: these tools could reduce manual research bottlenecks, letting scientists test more ideas and find insights faster across fields like drug discovery and climate modeling.
source: google deepmind: gemini for science: ai experiments and tools for a new era of discovery