Hand-placed toy map. Real embeddings are learned by a model, not placed by a person, and have far more than two numbers.