Given a potential function, which represents a repulsive interaction, the energy of a configuration of points is defined by summing a corresponding penalty for each pair of points. The goal is to minimize this energy over all configurations in Euclidean space with a fixed number of points per unit volume. Siegel's mean value theorem gives the average value of the energy over all lattices of determinant 1. Lower bounds on the energy can be proved by the linear programming method. In joint work with Henry Cohn, we show that the lower bound and the average are within a factor $1+o(1)$ as the spatial dimension grows, provided the potential is a Gaussian and not too steep. In particular, lattices are close to optimal in high dimensions. The limiting case of a very steep potential is related to sphere packing, where it is far from understood how close to optimal lattices may be. Time allowing, we will also discuss the work of Cohn-Kumar-Miller-Radchenko-Viazovska giving an exact solution to the linear program in dimension 8 and 24.