[Chapter 2] Exercise 4 prepare and predict pipeline

Not sure if this is the correct place for this question, but is there any point in practice to a pipeline that both prepares the data and then makes the prediction, as in Exercise 4?

From what I understand in the book, a pipeline causes fit_transform to be called for each transformer on the data sequentially - does this not mean that if you were to use this pipeline for the test set, it would fit the transformers and model to the test set, which is not what we want?