This talk presents a vision for a new scientific paradigm: AI-driven continuous discovery, in which artificial intelligence serves as an active partner throughout the research cycle, from data acquisition to decision-making and experimental feedback. Central to this idea is the move toward closed-loop, self-improving discovery systems that continuously learn from diverse scientific inputs. The presentation introduces interconnected initiatives under the umbrella concept of ScienceAI, which includes systems that orchestrate scientific workflows across disciplines. GENIUS focuses on physics and materials science, automating simulation protocols and enhancing reasoning through AI and knowledge graphs. In catalysis, AI4CAT emphasizes closed-loop discovery using Bayesian optimization and AI-guided experimentation. BioAI applies these concepts to life sciences and sustainable agriculture, leveraging multimodal data to accelerate discovery. A key aspect is SABER-X, which bridges the gap between advanced characterization tools and routine lab access by integrating AI with X-ray science for improved material characterization. PIPO introduces a methodological framework to enhance AI reasoning and verification, fostering trustworthy scientific AI. Collectively, these initiatives represent a shift from static predictions to dynamic, self-refining discovery engines. The future of science lies in integrated ecosystems where models, data, workflows, instruments, and human expertise interact in real time, marking the emergence of AI-driven continuous discovery across various scientific fields.
