QCD undergoes a transition from the confined phase (hadron gas) to a deconfined quark-gluon plasma at a temperature of about 155 MeV. This phenomenon can be investigated in relativistic heavy-ion experiments and studied theoretically, using e.g. simulations of QCD discretised on a space-time lattice. In this talk, I will review some aspects of QCD at nonzero temperature, with an emphasis on results obtained by our lattice QCD collaboration. Very recently, machine learning has been introduced as a new tool to study lattice field theory. In the final part, I will present some results of applications of machine learning to phase transitions in statistical field theory.