Analogical Reasoning and Similarity as a Foundation for Artificial General Intelligence

This paper deals with the relation of similarity as it is employed in analogical reasoning, and within this framework we will examine its role as a basis for essential capabilities required for Artificial General Intelligence (AGI). The paper has two main objectives: first, to examine and analyze the basic capabilities required to achieve artificial general intelligence. Second, to show that analogical reasoning underlies each of these capabilities, and as such it is a fundamental ability upon which we would want to develop systems with general intelligence.

For the purpose of this discussion, I will adopt the view that developing artificial general intelligence requires developing capabilities essential for cognitive tasks. There are several such basic capabilities on which there is broad consensus in the literature: learning, creativity, understanding, and reasoning of various types. The paper’s claim is that analogical reasoning based on similarity underlies all of these capabilities.

For creativity, I will show its connection to associative thinking and how analogical reasoning and similarity underlie our methods for making creative discoveries and forming associative connections. In learning, I will examine how analogical reasoning and similarity underlie both transfer learning and autonomous learning. Regarding understanding, I will adopt the view that understanding constitutes knowledge of dependencies, where counterfactual reasoning is essential. I will show that analogical reasoning forms the foundation of such reasoning. For reasoning, I will demonstrate how analogical reasoning and similarity underlie common forms of inductive reasoning as well as abductive reasoning.