“Probably Approximately Correct,” by Leslie Valiant
In Probably Approximately Correct: Nature’s Algorithms for Learning and Prospering in a Complex World, author Leslie Valiant shows how machine learning principles arise in the evolution of life and allow organisms to predict things about their environments without needing to formulate theories first. Valiant, a professor of computer science at Harvard, calls these processes “ecorhithms,” and explains the core computer science principles behind them without burdening nontechnical readers with code. The book has been the topic of considerable discussion in computational biology, due to its argument that Darwin’s theory of evolution is incomplete; Valiant makes the case that Darwin’s theory does not adequately predict the rapid rate at which evolution occurs and ought to be supplemented by additional computational theory.
The book’s title was derived from “probably approximately correct learning,” a machine learning paradigm proposed by Valiant in 1984. Valiant won the Turing Award, computer science’s highest honor, in 2010.