Nature papers describe Google and FutureHouse AI science assistants

In short: Two papers in Nature describe AI systems that help scientists read research faster and propose testable ideas, including ways to reuse existing drugs.

What happened

Nature published two papers about AI tools built to support scientific work, not replace scientists. One is Google’s “Co-Scientist,” which uses Google’s Gemini model and keeps researchers involved at each step. Scientists set the goal, the system searches the research literature, proposes possible explanations and drug ideas, and humans review what to test.

The second tool is from FutureHouse, a nonprofit, and its system is called “Robin.” Like Co-Scientist, it searches and summarizes large numbers of papers quickly, then ranks possible explanations for a disease. FutureHouse says Robin reviewed 551 papers in about 30 minutes, which they estimate would take a person about 540 hours.

A key difference is that Robin also includes a component that can evaluate results from certain common lab tests. For example, it can help analyze flow cytometry (a lab method that counts and measures cells, like a fast sorting machine) and RNA-seq (a way to measure which genes are active). In the work described, both tools suggested existing drugs to test, one for acute myeloid leukemia and the other for a form of macular degeneration.

Why it matters

Modern science produces more papers than any one person can keep up with. These tools aim to act like a very fast research assistant that reads widely and points out useful connections, then leaves final decisions and careful checking to human experts. The papers also highlight a risk, AI can make up incorrect citations, so access to reliable literature sources and human review still matter.

Source: Arstechnica