Machine learning techniques are ubiquitous in the digital economy as well as in the physical sciences. Recently, several uses of machine learning techniques have been proposed in the field of lattice QCD or, more broadly, lattice field theory. This is a rather natural development, since lattice QCD is one of the most computationally intensive areas of scientific research. Simulations with physical quark masses consume hundreds of millions of CPU hours and also generate large amounts of data. Machine learning promises to greatly improve the efficiency of these calculations in ways that are only beginning to be explored. In turn, the unique features and demands of lattice field theory calculations, in particular symmetries such as gauge invariance, and demands of guaranteed exactness, offer compelling research opportunities in machine learning with broad applicability.