The brain as a blueprint: a survey of brain-inspired approaches to learning in artificial intelligence

Abstract

Inspired by key neuroscience principles, deep learning has driven exponential breakthroughs in developing functional models of perception and other cognitive processes. A key to this success has been the implementation of crucial features found in biological neural networks: neurons as units of information transfer, non-linear activation functions that enable general function approximation, and complex architectures vital for attentional processes. However, standard deep learning models rely on biologically implausible error propagation algorithms and struggle to accumulate knowledge incrementally. While, the precise learning rule governing synaptic plasticity in biological systems remains unknown, recent discoveries in neuroscience could fuel further progress in AI. Here I examine successful implementations of brain-inspired principles in deep learning, current limitations, and promising avenues inspired by recent advances in neuroscience, including error computation, propagation, and integration via synaptic updates in biological neural networks.