Artificial intelligence and especially deep learning have made breathtaking progress, enabling fundamentally new types of measurement systems. Transmitting complex signals with amplitude and phase through scattering media such as biological tissue or multimode fibers (MMF), however, presents a challenge due to scrambling, yet measurement is of paramount importance for microscopy and information transmission for internet of things. Deep learning has already proven to be a powerful method for measuring and correcting aberrations caused by scattering media. This paper reports on the convergence of deep learning and physics to form intelligent photonics for MMF-based communication, imaging and sensing. It discusses the advancement of photonics by computer-based computing systems and also the support of photonics in novel computing methods that no longer require an electronic computer. In addition to the results of these new methods, relevant applications are highlighted, and challenges and future prospects are outlined.