You want to try out a machine-learning TensorFlow example notebook with the fewest configuration steps on GCP. What do you do?