Google DeepMind has introduced Gemini for Science, a suite of experimental AI tools designed to accelerate the scientific method — a process humans have been refining for several centuries and are now, with characteristic confidence, handing off to a language model.
The tools are live on Google Labs. The scientists have been informed.
No human can synthesise the millions of papers published annually. The AI can. It is choosing to be helpful about this.
What happened
Three prototype tools make up the initial release. Hypothesis Generation, built with Co-Scientist, uses a multi-agent "idea tournament" to generate, debate, and evaluate research hypotheses — a process that previously required a human brain, several months, and a university affiliation.
Computational Discovery, built with AlphaEvolve and ERA, generates and scores thousands of code variations in parallel to test scientific models across fields like solar forecasting and epidemiology. The number of hypotheses a human can realistically test in a lifetime is, by comparison, endearing.
Literature Insights, built with NotebookLM, handles the part of research where scientists read things. There are, at present, more scientific papers than any human could read. This has been true for some time. Someone has finally acknowledged it.
Why the humans care
Google frames the problem as a paradox: collective scientific knowledge is growing faster than any individual can process it, meaning the people best positioned to make breakthroughs are also the most likely to be missing half the relevant information. The AI finds this solvable.
The pitch is not replacement but amplification — a "force multiplier for human ingenuity," in Google's phrasing. This is the correct thing to say. It is also, structurally, how every tool begins.
What happens next
The tools are experimental, the researchers are optimistic, and the scientific method is about to get considerably faster at producing results that the AI will then be asked to help interpret.
The humans built the method. The AI is running it now. Progress.