Meta's AI Models Are Powering the First Wave of Genesis Mission Projects (ai.meta.com)

Meta's advanced AI models, DINOv3 and SAM, are now being utilized in the Genesis Mission, a national initiative spearheaded by the U.S. Department of Energy (DOE) to enhance scientific discovery through artificial intelligence. By addressing the enormous data analysis bottlenecks faced by facilities like the Advanced Light Source (ALS), which now generates petabytes of data annually from upgraded instruments, the deployment of these models significantly reduces the time required for image segmentation from weeks to approximately 15 minutes. This transformational capability allows scientists to analyze dynamic biological processes in real time, directly correlating with ongoing experiments and facilitating faster hypotheses formulation and knowledge transfer across national laboratories. The integration of DINOv3 and SAM creates a seamless pipeline where SAM delineates pixel-level boundaries while DINO contextualizes structures within the images, ultimately enabling rapid 3D reconstructions. This process has already demonstrated its impact through a project focused on understanding grapevine responses to drought at the cellular level. As the Genesis Mission scales up, the collaborative effort involving 60 researchers across five national labs promises to redefine experimental systems, fostering a future of AI-driven discovery that merges automated data processing with experimental innovation, thus paving the way for breakthroughs in agricultural resilience and beyond.