Foundations of Neurosymbolic AI

Explore the foundational concepts of neurosymbolic AI in this 51-minute lecture that bridges neural networks and knowledge representation reasoning (KRR) approaches. Begin by examining the two primary traditions of artificial intelligence and understanding their respective failures before diving into contemporary examples like Amazon Guardrails and DeepMind AlphaEvolve. Discover the cognitive psychology foundations underlying neurosymbolic AI through dual process theory, including insights from Kahneman's "Thinking Fast and Slow," while also examining critiques of this theoretical framework. Learn about working memory and cognitive architectures that inform neurosymbolic approaches, then delve into the core principles of neurosymbolic AI itself. Study Kautz's taxonomy for categorizing neurosymbolic systems and explore the limitations of this classification system alongside Van Bekkum's design patterns. Gain comprehensive understanding of how symbolic methods like logic programming intersect with deep learning to advance toward artificial general intelligence goals.