Memory

Abstract

Abstract This chapter focuses on the views of the Ibn Sina (Avicenna) on memory. Like other empiricists since Aristotle, Ibn Sina held that memory plays a critical role in human abstraction. In particular, Ibn Sina emphasized the role of affect in allowing memory to influence rational decision-making, by allowing agents to recognize non-perceptible consequences of actions for instrumental choice. The chapter ties Ibn Sina’s ideas about memory to successful models in deep learning that make use of deep reinforcement learning, such as DeepMind’s Episodic Controller architecture, which can play Atari games at superhuman levels of performance. It also reviews classic ideas about the role of memory-simulating modules in neural network architectures, such as the need to overcome catastrophic interference.