ABSTRACT
Autonomous mental development models all or part of the brain and how a system develops autonomously through interactions with the environments. The most fundamental difference between traditional machine leaning and autonomous mental development is that a developmental program is task nonspecific so that it can autonomously generate internal representations for a wide variety of simple to complex tasks. This chapter discusses why autonomous development is necessary based on a concept called task muddiness. It explains some basic concepts of autonomous development, including the paradigm for autonomous development, mental architectures, developmental algorithm, a refined classification of types of machine learning, spatial complexity, and time complexity. The chapter describes the architecture of spatiotemporal machine that is capable of autonomous development. The biological mental development takes place in concurrence with the body development and they are closely related. In the machine learning literature, there have been widely accepted definitions of learning types, such as supervised, unsupervised, and reinforcement learning.
