In the end, she [Janelle Shane] concludes: "1. The neural network really likes brown, beige, and grey; 2. The neural network has really, really bad ideas for paint names." Possibly the neural network needed better parameters, but really, who can argue with results like these?
