Machine learning algorithms have given us powerful new tools for computational drug design. Going beyond machine learning regression methods for predicting drug binding and activity, generative AI methods can create new potential drug molecules. These methods can simultaneously explore novel ligand chemical space while optimizing contacts with the protein drug binding pocket. I will describe our work with two different classes of generative AI algorithms, hybrid generative adversarial networks and deep reinforcement learning. The primary remaining challenge is designing molecules that are synthetically tractable.