A virtual laboratory with AI to accelerate discoveries

Scientific research is accelerating with the ' Virtual Lab ,' the first virtual laboratory with artificial intelligence systems that emulate human researchers: programmed with different scientific skills , they form a multidisciplinary team capable of reasoning about complex problems , developing research strategies , and conducting virtual experiments , producing results that can then be validated in real laboratories . The platform, which aims not to replace scientists but to enhance their work , was presented in the journal Nature by researchers at the Chan Zuckerberg Biohub San Francisco and Stanford University. In the Virtual Lab, the human user creates an artificial intelligence agent that acts as the principal investigator, who in turn assembles and directs a team of AI agents that emulate specialized research roles typical of scientific laboratories, such as virologists, computer scientists, or bioengineers; there's even a critic , whose job is to question hypotheses to avoid 'hallucinations.'

Once the team is ready, the human researcher proposes a scientific question and then monitors meetings in which the various AI agents exchange ideas to advance the research. The agents are managed by a large linguistic model (LLM) that provides them with scientific reasoning and decision-making capabilities. As a testbed , the Virtual Lab was used to design nanoantibodies (small proteins that function like antibodies) directed against the new variants of the SarsCoV2 virus : in a short time, 92 were developed , two of which proved particularly effective at binding the viral Spike protein.

"This is the first demonstration of autonomous AI agents actually solving a challenging research problem, from start to finish," says Stanford's James Zou. "What was once a crazy science fiction idea is now reality ," comments John Pak of the Chan Zuckerberg Biohub in San Francisco, unconcerned about the risk of AI supplanting humans in scientific research. " The Virtual Lab has given us more work , in a sense, because it has given us more ideas to test . If AI can produce more testable hypotheses , that means more work for everyone ."