Virtual lab powered by “AI scientists” greatly enhances biomedical research

In an effort to use artificial intelligence (AI) to automate scientific discovery, researchers are building multiple “AI scientists” (scientists) who can collaborate to achieve goals set by human researchers. We created a virtual laboratory that combines a large-scale language model with defined roles.

The system, described in a preprint posted to bioRxiv last month, is able to engineer antibody fragments called nanobodies that can bind to the virus that causes COVID-19, and uses these structures to We were able to propose nearly 100 of them within the time it took to do all the calculations. -Human Research Group.

Researchers have built an “AI scientist” — what can it do?

“These virtual laboratory AI agents have been shown to be highly capable of performing many tasks,” said study co-author and computational biologist at Stanford University in California. James Zou said. “We are very excited to explore the potential of virtual labs across a variety of scientific disciplines.”

Yanjun Gao, who studies AI healthcare applications at the University of Colorado Anschutz Medical Campus in Aurora, said the experiment “represents a new paradigm of using AI as a collaborator rather than just a tool.” says. But human input and monitoring remain important, she added. “At this stage, I don’t think we can fully trust AI to make decisions.”

Multidisciplinary AI

Scientists around the world are using large-scale language models (LLMs) to speed up research, including the creation of “AI scientists” who can perform parts of the scientific process, from generating hypotheses to designing experiments to drafting papers. ). However, most research focuses on applying LLM to a narrow range of experiments, rather than exploring its potential in interdisciplinary research, Zou says. He and his colleagues established a virtual lab to combine expertise from different fields.