The recipe for intelligence in natural and artificial systems.

Abstract

Ten years ago, two landmark papers discussed open problems in AI research from the standpoint of neuroscience and cognitive science. Today, many of these problems have been solved, mostly without direct inspiration from the study of the human brain. In this paper, we consider what is still shared between the twin endeavors of understanding natural brains and building artificial ones. We argue that, as well as understanding the "ingredients" of intelligence (its component parts), both fields are now coming to grasp the "recipe" for intelligence (how they fit together). Surprisingly, this recipe has many parallels in biological and artificial systems. Despite rapid progress in AI research, there are many computational problems that remain unsolved. We discuss how these problems may be addressed in natural intelligence.