The test of our progress is not whether we add more to the abundance of those who have much; it is whether we provide enough for those who have too little.
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.