In the context of computational learning, the concept of 'probably approximately correct' (PAC) learning provides a framework for understanding the efficiency and feasibility of learning algorithms.
I sent my Soul through the Invisible, Some letter of that After-life to spell: And by and by my Soul return'd to me, And answer'd "I Myself am Heav'n and Hell."