Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only! Grab it Explore the concept of algorithmic alignment in neural network architecture design through this insightful lecture by Petar Veličković. Delve into the fundamental question of what makes neural networks better or worse at fitting certain tasks. Examine various mathematical approaches used to address this question, with a focus on algorithmic alignment. Learn how this approach equates fitting a task to capturing the computations of an algorithm, drawing from diverse branches of mathematics and computer science. Discover the speaker's favorite works in algorithmic alignment and their potential implications for future intelligent systems. Gain valuable insights into this cutting-edge research area presented at the SPCL_Bcast #46 event, recorded on March 21, 2024, at the Scalable Parallel Computing Lab, ETH Zurich.