A theoretical framework to formalize AGI-Hard problems

Abstract

The main goal of the Artificial General Intelligence field (AGI) to create "human level intelligence" is known as a very ambitious one (Hut04). On the way to the field development there are many difficult problems to solve, like natural language translation, for example, which seem to share some "hardness" properties. The terms "Al-Complete" and "Al-Hard", by analogy with the terms "NP-Complete" and "NP-Hard" from computational complexity theory (CLRS0I), have been informally used to classify them although there are also works that propose some kind of formal definition (SA07), (vABHL03). This work proposes a theoretical framework with formal definitions to distinguish these problems and discuss its use in practical applications and how their properties can be used in order to achieve improvements in the AGI field.