A physical selfie stick extends the user's reach, enabling the creation of personal photos that include more of the background scene. Conversely a quadcopter can capture photos at distances unattainable for the human, but teloperating a quadcopter to a good viewpoint is a non-trivial task. This paper presents a natural interface for quadcopter photography, the Selfie Drone Stick that allows the user to guide the quadcopter to the optimal vantage point based on the phone's sensors. The user points the phone once, and the quadcopter autonomously flies to the target viewpoint based on the phone camera and IMU sensor data. Visual servoing is achieved through the combination of a dense neural network object detector that matches the image captured from the phone camera to a bounding box in the scene and a Deep Q-Network controller that flies to the desired vantage point. Our deep learning architecture is trained with a combination of real-world images and simulated flight data. Integrating the deep RL controller with an intuitive interface provides a more positive user experience than a standard teleoperation paradigm.