AI Tools Help Scientists Find New Uses for Existing Drugs

Scientists just got two powerful new assistants that could help them find medical breakthroughs hiding in plain sight.

Google and nonprofit FutureHouse each released AI systems this week that tackle a growing problem in science: there's simply too much research for any human to read. With scientific journals exploding online, crucial discoveries in one field often go unnoticed by researchers who could benefit from them.

These new tools work differently than you might expect. Instead of replacing scientists, they run searches in the background, connecting dots across thousands of papers while researchers focus on actual experiments.

FutureHouse's system, called Robin, analyzed 551 research papers in just 30 minutes. A human scientist would need 540 hours to do the same work. The AI then suggested theories about eye disease and identified existing drugs that might help treat macular degeneration.

Google's Co-Scientist took a similar approach with leukemia, combing through medical literature to find drugs already approved for other conditions that might work against cancer cells. Scientists stayed involved throughout, reviewing and prioritizing the AI's suggestions before testing them in the lab.

The results showed real promise. Several drugs identified by Google's system proved effective against certain types of leukemia cells. That's typical for cancer research, where different tumors respond to different treatments.

The Ripple Effect

The broader impact goes beyond individual discoveries. An eye researcher might miss important findings about cell signaling happening in kidney studies, simply because nobody has time to read everything. These AI tools excel at finding those hidden connections.

Both systems avoid a common AI pitfall by constantly checking their work against actual published research. This prevents them from making up plausible-sounding but fake discoveries, something earlier AI models struggled with.

The tools also improve over time. Google designed its system to work with future, more advanced AI models as they become available. FutureHouse used specialized reading tools that create both quick summaries and deep analyses of scientific papers.

Neither group claims their AI can dream up revolutionary new theories or replace human creativity. Instead, they're targeting what they call "low-hanging fruit," valuable connections that exist but remain undiscovered because knowledge gets trapped in separate fields.

The approach works for any science, though both teams focused their first tests on biology and medicine. Google mentioned its system could also help physics researchers but provided few details.

Scientists reviewed and approved every suggestion before testing began. The human experts had full access to the research the AI used, ensuring transparency and safety throughout the process.

This collaboration between artificial and human intelligence points toward a future where researchers spend less time searching and more time discovering.