Back to Main List
Back to Book List

Notes and Reflections on Books and Media

by Hannah Leitheiser

Master Algorithum

The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World

Pedro Domingos




The various strategies for machine learning include artificial neural networks that roughly simulates brain cells, inductive logic programming that finds probable logical assertions, support vector machines that learn by analogy, and bayesian networks which work on probabilities. I'm just beginning to study this, so know very little myself, but it seems researchers are finding some main approaches can be represented by the same framework, leading to the search for a master algorithm which would unify artificial intelligence, much as physicists search for unifying theories of quantum mechanics and general relativity.

Presumably this algorithm would solve the AI-complete problem -- the problem of creating human-level general intelligence. Every problem in machine learning that must learn from many domains or learn contextual understanding seems to be roughly the same problem and will probably involve roughly the same computer code.