Keynote Speaker Profile

Abstract: Neurosymbolic AI seeks to bridge the strengths of neural networks and symbolic AI to overcome fundamental limitations of each approach when used in isolation. This talk provides an accessible introduction to the field, clarifying core concepts and distinguishing modern neurosymbolic methods from earlier hybrid attempts -- or perhaps noting how they are not so different. We will critically examine why this integration is urgently needed today: deep learning’s data hunger, lack of robustness, opacity, and struggle with causal or relational reasoning contrast sharply with symbolic AI’s brittleness in uncertain, real-world environments—and neither alone suffices for trustworthy, generalizable AI. The discussion will then pivot to the current state-of-the-art gaps, highlighting key challenges such as enabling gradient-based learning of symbolic structures, scaling complex reasoning, or achieving true generalization, as well as domains well-suited the approach, such as healthcare, robotics, and scientific discovery. This will equip early researchers with both the foundational understanding and awareness of open problems in the field.