Machine learning techniques are found to be increasingly useful in analyzing data from large galaxy surveys, solving tasks such as automatic classification of galaxy morphology, estimation of photometric redshifts, outlier recognition, and data compression, just to quote a few examples. With respect to traditional template-fitting techniques, machine learning methods can help optimally harvest information from heterogeneous data sources, limit the biasing impact of model dependence, and improve speed; however, validation is often challenging, and they may be hindered by lack of interpretability. In this talk, I review some applications of machine learning and deep learning to the problem of measuring galaxy physical properties, and highlight what in my opinion are the most significant challenges we need to solve in order to be ready for the next generation of surveys.