#115 – Dileep George: Brain-Inspired AI

The conversation explores the intersection of neuroscience and artificial intelligence, focusing on the potential for brain-inspired engineering to advance artificial general intelligence. Building functional models of the brain requires moving beyond simple feed-forward neural networks to incorporate feedback connections, lateral connectivity, and dynamic inference. These mechanisms allow systems to model the world, perform causal reasoning, and handle out-of-distribution tasks with minimal training data, as demonstrated by the Recursive Cortical Network. Current large-scale language models face fundamental limitations because they lack a grounded world model and the capacity to simulate physical scenarios. Intelligence emerges from understanding the machinery of the world through perception and motor interaction, suggesting that future progress depends on integrating structured cognitive architectures with the ability to perform iterative, top-down inference rather than relying solely on statistical text compression.