Two AI-based science assistants succeed with drug-retargeting tasks

On Tuesday, the journal Nature unveiled two groundbreaking studies focused on AI technologies aimed at aiding scientists in formulating and testing hypotheses. One of these, developed by Google and referred to as Co-Scientist, operates on the premise of a "scientist in the loop," which emphasizes the essential role of human judgment in guiding the AI's processes. The second system, created by the nonprofit organization FutureHouse, takes a bolder step by employing an AI that can analyze biological data from specific experimental setups. While Google’s Co-Scientist claims applicability in fields like physics as well, both systems predominantly showcase their capabilities in processing biological data, primarily tackling straightforward hypotheses—such as determining if a particular drug is effective. Rather than aiming to replace scientists or the scientific methodology, these AI systems are designed to assist with the challenging task of sifting through vast quantities of information that would otherwise overwhelm human researchers. Both tools, classified as agentic, function in the background by interfacing with separate resources to enhance their analytical capabilities. Despite some differences between the two systems, they share a common goal: to address the overwhelming influx of scientific literature. The digital age has led to a surge in the number of academic journals and published studies, making it increasingly difficult for researchers to stay informed about advancements in their fields. For instance, a researcher focused on eye development might overlook important findings related to similar signaling systems being studied in kidney research, making cross-disciplinary awareness a significant challenge.