The concept of probably approximately correct (PAC) learning has been foundational in computational learning theory.
"It is worth living for, to find out what is true."