The adaptation of infrastructure to events including hydrological and weather extremes, which may be exacerbated by climate change, largely depends on high-resolution climate projections decades and centuries into the future. The downscaling of a wide range of general circulation models (GCMs) is imperative to understanding climate change at a local scale. In parallel of developing frameworks allowing for credible validation, the statistics and machine learning communities are working on specific methods for covariate selection and prediction on various climate applications. State-of-the-art machine learning methods that are pushing the boundaries in fields such as computer vision, natural language processing, and speech recognition should be carefully considered for incorporation into statistical downscaling (SD). More specifically, the deep learning models being developed, focusing on dimensionality reduction and spatiotemporal data, have potential to improve over the current generation of SD models. Utilizing multitask learning (MTL) to take advantage of spatially dependent observations may allow for more generalizable models.