This review paper provides an overview of active inference, a unifying theory of action and perception. Active inference is based on the idea that sentient behavior relies on the brain's use of internal models to predict, infer, and direct action. The paper traces the evolution of this theory from early ideas on unconscious inference to contemporary understanding of action and perception. It discusses related perspectives, the neural underpinnings of active inference, and future opportunities for development. Key developments include predictive coding models, sequential models for planning, and hierarchical generative models. Active inference has been used to explain aspects of anatomy, neurophysiology, and psychopathology, and to unify psychological theories. The paper highlights the growing impact of active inference across disciplines, including neuroscience, psychology, robotics, and artificial intelligence. It also discusses the conceptual and historical roots of active inference, including ideas from Helmholtz, cybernetics, and Gibson. The paper emphasizes the normative perspective of active inference, which unifies predictive and enactive views of the brain and behavior. It also discusses the application of active inference to cognitive tasks, the importance of precision control, and the potential for extending active inference to multi-agent settings. The paper concludes with the benefits of unification, including the ability to connect different levels of understanding, unify cognitive functions, and provide a framework for understanding sentient behavior. The paper also highlights the potential of active inference in the development of generative AI and the importance of empirical validation of active inference theories.This review paper provides an overview of active inference, a unifying theory of action and perception. Active inference is based on the idea that sentient behavior relies on the brain's use of internal models to predict, infer, and direct action. The paper traces the evolution of this theory from early ideas on unconscious inference to contemporary understanding of action and perception. It discusses related perspectives, the neural underpinnings of active inference, and future opportunities for development. Key developments include predictive coding models, sequential models for planning, and hierarchical generative models. Active inference has been used to explain aspects of anatomy, neurophysiology, and psychopathology, and to unify psychological theories. The paper highlights the growing impact of active inference across disciplines, including neuroscience, psychology, robotics, and artificial intelligence. It also discusses the conceptual and historical roots of active inference, including ideas from Helmholtz, cybernetics, and Gibson. The paper emphasizes the normative perspective of active inference, which unifies predictive and enactive views of the brain and behavior. It also discusses the application of active inference to cognitive tasks, the importance of precision control, and the potential for extending active inference to multi-agent settings. The paper concludes with the benefits of unification, including the ability to connect different levels of understanding, unify cognitive functions, and provide a framework for understanding sentient behavior. The paper also highlights the potential of active inference in the development of generative AI and the importance of empirical validation of active inference theories.