The celebrated Green-Tao theorem states that the primes contain arbitrarily long arithmetic progressions. In this talk, I will give some background and explain the ideas of proof, including some recent simplifications due to Conlon, Fox, and myself. The main idea is a transference principle, which allows one to transfer a result about dense sets (in this case, Szemerédi's theorem on arithmetic progressions in dense subsets of integers) to one about subsets of sparse pseudorandom sets. The talk will be accessible to people working in algorithms/complexity.