Data-driven paradigms are well-known and salient demands of future wireless communication. Empowered by big data and machine learning, next-generation data-driven communication systems will be intelligent with the characteristics of expressiveness, scalability, interpretability, and especially uncertainty modeling, which can confidently involve diversified latent demands and personalized services in the foreseeable future. In this paper, we review and present a promising family of nonparametric Bayesian machine learning methods, i.e., Gaussian processes (GPs), and their applications in wireless communication due to their interpretable learning ability with uncertainty. Specifically, we first envision three-level motivations of data-driven wireless communication using GPs. Then, we provide the background of the GP model in terms of covariance structure and model inference. The expressiveness of the GP model is introduced by using various interpretable kernel designs, namely, stationary, non-stationary, deep, and multi-task kernels. Furthermore, we review the distributed GP with promising scalability, which is suitable for applications in wireless networks with a large number of distributed edge devices. Finally, we provide representative solutions and promising techniques that adopting GPs in wireless communication systems.