This review paper by Pezzulo, Parr, and Friston provides an overview of active inference, a unifying perspective on action and perception. Active inference posits that sentient behavior depends on the brain's implicit use of internal models to predict, infer, and direct action. The authors trace the evolution of this theory from Helmholtzian ideas on unconscious inference to contemporary understandings of action and perception. Key developments include the formulation of predictive coding models, the use of sequential models for planning and policy optimization, and the importance of hierarchical (temporally) deep internal models. Active inference has been applied to various fields, including neuroscience, psychology, robotics, and artificial intelligence, and has been used to account for aspects of anatomy, neurophysiology, and psychopathology. The paper highlights the benefits of unification in cognitive psychology and neuroscience, emphasizing how active inference can align conceptual terms with formal terms, suggest a unique process theory, unify different levels of understanding, endow constructs with validity, and reconcile theoretical perspectives. The authors also discuss future opportunities, including deeper empirical scrutiny of active inference and its potential to complement the development of Generative AI systems.This review paper by Pezzulo, Parr, and Friston provides an overview of active inference, a unifying perspective on action and perception. Active inference posits that sentient behavior depends on the brain's implicit use of internal models to predict, infer, and direct action. The authors trace the evolution of this theory from Helmholtzian ideas on unconscious inference to contemporary understandings of action and perception. Key developments include the formulation of predictive coding models, the use of sequential models for planning and policy optimization, and the importance of hierarchical (temporally) deep internal models. Active inference has been applied to various fields, including neuroscience, psychology, robotics, and artificial intelligence, and has been used to account for aspects of anatomy, neurophysiology, and psychopathology. The paper highlights the benefits of unification in cognitive psychology and neuroscience, emphasizing how active inference can align conceptual terms with formal terms, suggest a unique process theory, unify different levels of understanding, endow constructs with validity, and reconcile theoretical perspectives. The authors also discuss future opportunities, including deeper empirical scrutiny of active inference and its potential to complement the development of Generative AI systems.