Machine learning relies heavily on sampling from complex probability distributions. NASA scientists successfully trained the D-Wave 2X system on image data sets in a generative unsupervised learning setting, one of the most difficult paradigms in machine learning.