Today in grad-school: students pretending to be dolphins to understand actor-critic reinforcement learning.