This document summarizes an AI presentation on using machine learning models to predict TV audiences and recommend TV content. It discusses how viewing habits are shifting to digital platforms and how AI can help with audience prediction, segmentation, and content scheduling/recommendations by analyzing programming data and identifying important features like content types and external events. The models showed improved prediction accuracy when incorporating detailed programming data as features compared to only past viewership or coarse programming categories. Feature selection identified content and events as important predictive characteristics.