Image-based phenotyping with genomic correlation is opening a new window to the groundbreaking and innovative research field by combining genetic and morphologic data. The natural extensions of the works with visual review of CT scans by radiologists are applications of deep learning and artificial intelligence. We introduce the main concepts and core ideas and briefly review the history and current state-of-the-art artificial intelligence (AI) and deep learning (DL) applications in clinical and research radiology. Unsupervised and supervised general approaches are reviewed. We spell out the technical, knowledge discovery, regulatory, and ethical challenges for the wider and seamless adoption of AI in clinical radiology practice. We advocate for parallel development of deterministic and probabilistic approaches to supplement the rapid advancements in deep learning and simulation technique applications which often serve as “black box” with limited clinical insight behind the results they provide. The current applications of deep learning in chest imaging are also summarized.