Every supervised recognition problem in machine learning requires a algorithm's memory of data. Every story or poem a human writes requires a human's memory of living in the world. At the intersection of human and machine memory, new opportunities arise for creativity. In this talk we'll see battling image recognizers remembering both the Parthenon and the crowded front of an American refrigerator; we'll walk thru a story generation graph visually; and we'll write poems using a photograph's colors and a word2vec maze. The magical transformation common in all these AI toys is making the image textual and the textual visual. The algorithms' memories will do all the hard work: we'll just invite them to guide us through the palace.