85-426 Reverse Engineering Brains and Machines
9 units, Carnegie Mellon University
Biological and artificial intelligence emerge from relatively simple computational principles, yet the representations these systems encode are not directly accessible without careful scientific inquiry. This course explores how brains and machines can be reverse engineered to reveal what they represent, how they compute, and why they behave as they do. Using tools such as representational similarity analysis (RSA), we will study the structural and functional alignment between biological and artificial systems and how sensory information becomes neural code. The course treats brains and modern neural networks as complementary objects of study, emphasizing shared principles, key differences, and the limits of interpretability. Drawing on neuroscience, cognitive science, and machine learning, students will examine methods for interrogating internal structure, including behavioral testing, neural recordings, model comparison, and computational probes. We will also compare perceptual illusions in biological systems with adversarial perturbations in artificial systems. By examining where these models align with, or fail to match, biological behavior and neural activity, students will gain insight into what it takes to build more brain-like AI. This course prepares students to develop their own studies of internal representations and to critically evaluate claims about brain and machine intelligence.
Prerequisites: 15-112
Offered: Fall 2026
Courseweave needs JavaScript to show sections, workloads and planning tools.