The growing limitations of centralized AI, including increasing demands on scalability, energy efficiency, communication bandwidth, and latency, are driving a rapid proliferation of AI-enabled devices, e.g., sensors, cameras, wearables, robots, and mobile systems. As intelligence moves closer to where data are generated, new paradigms are needed to enable learning and decision-making across highly distributed and resource-constrained environments. This talk presents new research vision on sustainable decentralized intelligence, aiming to establish the foundations of AI systems that learn continuously from distributed data while operating under strict resource constraints. The talk will discuss recent advances at the intersection of Federated Learning, Continual (Lifelong) Learning, and Reservoir Computing, three complementary paradigms that enable collaborative, adaptive, and energy-efficient intelligence without relying on centralized data processing. Together, these approaches pave the way for scalable and sustainable AI infrastructures for future cyber-physical systems, smart environments, and distributed autonomous networks, while opening new opportunities for interdisciplinary collaboration.

Paolo Dini is leading researcher within CTTC/CERCA, where he coordinates the activities of the Sustainable AI research unit. Previously he was with University of Sannio (Italy) as assistant professor (2005). He holds a PhD in Information and Communication Technology by the University of Rome “La Sapienza” (Italy). His research interests encompass efficient and distributed machine learning, distributed control, sustainable computing and networking, and cyber-physical systems. He has been actively involved in several EU initiatives in these fields, coordinating some of them, such as MSCA SCAVENGE, MSCA Greenedge, EIC SONATA. SCAVENGE and GREENEDGE have received the recognition of "EU Success Stories" by the European Commission. His contributions are documented in 100+ scientific publications in journal and proceedings of international peer-reviewed conferences. He is Associated Editor of the IEEE Open Journal of Computer Society and member of the ELLIS Society (European Laboratory for Learning and Intelligent Systems).