The science of understanding what happens inside AI is called interpretability.

This demo is a beginner-friendly introduction to interpretability that explores an AI model called Gemma 2 2B. It also contains interesting and relevant content even for those already familiar with the topic.