Universal Algorithmic Intelligence: A mathematical top->down approach

TL;DR

This paper introduces the AIXI model, a universal, mathematically rigorous framework for artificial intelligence that combines decision theory and universal induction, and discusses its theoretical properties and computable approximations.

Contribution

It presents the AIXI model as a formal, unifying theory of universal AI, and proposes a computable variant AIXItl with bounded resources.

Findings

AIXI is the most intelligent unbiased agent possible.

AIXItl is a practical, resource-bounded approximation of AIXI.

The paper formalizes intelligence and discusses relations to other AI approaches.

Abstract

Sequential decision theory formally solves the problem of rational agents in uncertain worlds if the true environmental prior probability distribution is known. Solomonoff's theory of universal induction formally solves the problem of sequence prediction for unknown prior distribution. We combine both ideas and get a parameter-free theory of universal Artificial Intelligence. We give strong arguments that the resulting AIXI model is the most intelligent unbiased agent possible. We outline how the AIXI model can formally solve a number of problem classes, including sequence prediction, strategic games, function minimization, reinforcement and supervised learning. The major drawback of the AIXI model is that it is uncomputable. To overcome this problem, we construct a modified algorithm AIXItl that is still effectively more intelligent than any other time t and length l bounded agent. The computation time of AIXItl is of the order t x 2^l. The discussion includes formal definitions of intelligence order relations, the horizon problem and relations of the AIXI theory to other AI approaches.