This podcast discusses how optimization makes sense for single agents in robotics trying to optimize some objective function, and how there is a branch of optimization that explicitly looks for saddle points, which is uncommon since most optimization problems dislike them.