AI conducts biomedical research independently

Scientific research involves many steps: formulating a question, designing an experiment, collecting data, interpreting it, and reporting findings. Until now, all of those steps required human involvement. The study, published in Science, describes an AI agent that carried out all of these steps autonomously in a biomedical context.

The agent was not only able to search existing literature and generate hypotheses. It could also design experiments, direct their execution, and interpret results without continuous human oversight. The researchers tested this system on biomedical questions, though the available summary does not fully detail which specific experiments were conducted or how results compare to human-led research.

What changes for aging research?

For longevity science, this development is potentially significant. Aging research faces a fundamental problem: there are too many hypotheses and too few resources to test them systematically. An autonomous AI agent that generates hypotheses and runs experiments could, in theory, expand the search space and accelerate the pace of discovery.

But serious questions remain. The reproducibility and reliability of AI-generated research have not yet been tested at scale. Who evaluates the quality of the hypotheses an AI selects to investigate? And how do you prevent such a system from systematically searching in particular directions, at the expense of the unexpected discoveries that humans make by chance?

An early stage with large questions

This is an early proof of concept, published in a leading journal. The authors make no extravagant claims about what the system can do. But the direction is clear: AI-driven scientific practice is growing rapidly and is reaching aging research. How these tools are deployed and who evaluates their outputs are questions science must start answering now.

Read the original article