Artificial co-scientist is able to synthesize large amounts of literature and suggest new research plans. Already “predicted” results of an unublished discovery and proposed new targets and medication for fibrosis and leukemia.
Google announced last Tuesday the creation of an artificial intelligence co-scientist (AI) that acts as a virtual contributor to biomedical scientists, who uses advanced reasoning to help them to synthesize large amounts of literature, generate new hypotheses e suggest detailed investigation plans For validation of their hypotheses in various fields of medicine, according to technology.
According to Google, “the system is unique in using a self-play approach to scientific reasoning and the generation of new hypotheses, in which several Gemini models discuss and criticize potential hypotheses over time, in a tournament” .
The objective of the project, which has academic partners from Imperial College London, Stanford and Houston Methodist, is improve the quality of results with the scaling of computing in testing timeis referred to in the publication. This allows the system to generate and hierarize hypotheses and refined and high quality research plans for review, feedback and experimental execution of researchers, according to technology.
“We hope that, instead of decreasing scientific collaboration,” said Vivek Natarajan, Google scientist, quoted by.
According to Google, the hypotheses suggested by the tool showed promising initial results with the validation by academic partners.
For example, for researchers of Imperial College London, the co-scientist of IA proposed a hypothesis that reflected the results of an unublished and validated discovery Experimentally from a new gene transfer mechanism that could help understand the spread of antimicrobial resistance, it reads in the publication.
For Stanford scientists, “proposed new epigenetic targets for liver fibrosisAnd the drugs based on these targets were validated by the university as presenting antifibrotic activity.
In the case of Houston Methodist investigators, “the AI system suggested the reuse of an existing drug for the investigation of acute myeloid leukemiawhich was validated through computational biology and laboratory experiences. ”
Google says additional validation efforts are underway with selected academic and industry partners.
Researchers can use the system through Google.
