Recorded 17 October 2024. Mayank Mehta of the University of California, Los Angeles, presents "Dynamics of brain's deep network" at IPAM's Theory and Practice of Deep Learning Workshop. Abstract: Our brain is still the gold standard for abstraction and generalization, the ultimate goal for AGI. How does the brain's deep network process visual information to create abstract ideas such as events and facts? What the biophysical and emergent principles? While many theories have been proposed, ranging from Hopfield's associative nets to Hinton's error back prop, direct experimental observations and tests of these theories in vivo have remained elusive. Recent advances in neurotechnologies, from virtual reality for rodents to large scale recordings of single neuron activities have overcome these challenges and provided novel insights about the dynamics of the brain's deep network with unprecedented accuracy, which I will discuss. Relevant papers: https://www.nature.com/articles/s41586-021-03989-z https://elifesciences.org/reviewed-preprints/85069 Learn more online at: https://www.ipam.ucla.edu/programs/workshops/workshop-ii-theory-and-practice-of-deep-learning/?tab=overview