Closed-loop perception: gaps between artificial intelligence and biology

Abstract

Biological perception is achieved via brain–world interactions, predominantly implemented within closed-loop systems that integrate sensory inputs and the motor actions used to acquire them. In contrast, current developments of artificial perception primarily exploit open-loop configurations. Lacking this dynamic interplay, artificial intelligence (AI) remains limited in interpreting real-world data. Here, we review the fundamental gaps between biological perception, as revealed through neuroscience research, and artificial perception, as currently implemented in AI systems. We conclude with two major recommendations for advancing AI: adopting event-based processing and integrating closed-loop architectures.