The Anatomy
of AI
How to look inside a neural network and understand what it is doing, drawn as an anatomical atlas. We start from a single neuron and build up, one plate at a time, to the research of 2026. No background needed.
Can we understand what a neural network is doing by looking inside it?
That question is called interpretability, and it is central to making AI safe. This course follows the story of how people have tried to answer it.
All 32 lectures are finished. Each one is a deck of live, step-by-step figures you can touch, with narrated explainer films. Every result in it was re-run by us from scratch, and wherever our numbers differ from a paper's, the plate says so ("In our run"). Each lecture also links to its storyboard and its full experimental findings.