Artificial Intelligence (AI) has long aspired to support and even automate the scientific process by integrating data and knowledge structures to uncover scientific laws. Recent advancements in machine learning (ML) — notably in deep neural networks, explainable AI, and foundation models — are transforming how we analyze large, complex datasets and derive interpretable models across disciplines. However, the increasing use of AI in science also highlights significant limitations in interpretability, integration of domain knowledge, and adherence to physical laws. This project aims to develop novel AI methodologies that address these challenges by advancing four key areas: explainable ML for transparent model insights, scalable foundation models adaptable to diverse scientific domains, automated scientific modeling for hypothesis generation and law discovery, and semantic technologies to support open, FAIR, and TRUSTworthy science. By applying these methods to a range of applications in physical sciences, engineering, and life sciences, the project seeks to bridge the gap between data-driven approaches and the rigorous demands of scientific inquiry, ultimately enhancing the reliability and impact of AI in advancing scientific discovery.

ARIS Gravity