Search Results

Neurosymbolic Artificial Intelligence (NeSy) promises to combine the scalability and adaptability of neural networks with the precision and interpretability of logic-based reasoning systems. Despite significant theoretical advances and growing interest in the field, the development of NeSy systems that seamlessly integrate symbolic and sub-symbolic elements at scale for real-world practical applications remains an open challenge. This Dagstuhl Seminar brought together leading researchers from diverse fields including machine learning, logic, knowledge representation, knowledge discovery, robotics, healthcare, and natural sciences to chart a roadmap for advancing NeSy systems from synthetic benchmarks to practical deployment. The seminar addressed four key challenges: (I) developing realistic benchmarks and standardized evaluation frameworks for NeSy systems, (II) identifying and advancing applications in robotics, healthcare, and scientific discovery, (III) creating open-source libraries to lower entry barriers and accelerate development, and (IV) creating accessible educational materials to broaden participation in the field. Through structured discussions, breakout sessions, and collaborative planning, participants developed concrete action plans and deliverables, including a planned Communications of the ACM article, open-source library development (ULLER, DeepLog), educational resources, and improvements to the field’s Wikipedia entry. This report documents the seminar’s presentations, discussions, and findings, and presents strategic recommendations for realizing the promise of Neurosymbolic AI in real-world applications.