A good learning algorithm should be able to generalize from a limited set of examples to unseen data.
Childhood is the first and last frontier of literature.
The most interesting art happens at the edges of disciplines.
You can't control what you don't measure.