The sixth-generation wireless communications (6G) is often labeled as 'connected intelligence'. Radio sensing, aligned with machine learning (ML) and artificial intelligence (AI), promises, among other benefits, breakthroughs in the system's ability to perceive the environment and effectively utilize this awareness. This article offers a panoramic view of radio sensing by unifying physical object sensing and spectrum sensing. To this end, while staying in the framework of integrated sensing and communication (ISAC), we expand the term 'sensing' from radar, via spectrum sensing, to miscellaneous applications of radio sensing like non-cooperative transmitter localization. We formulate the problems, explain the state-of-the-art approaches, and detail AI-based techniques to tackle various objectives in the context of wireless sensing. Finally, we discuss the potential integration of various radio sensing functions into a common AI-enhanced framework, emphasizing the possible benefits and the challenges to overcome. In addition to the tutorial-style core of this work based on direct authors' involvement in 6G research problems, we review the related literature, and provide both a good start for those entering this field of research, and a topical overview for a general reader with a background in wireless communications.