Abstract: These lectures discuss an S-matrix approach to quantum gravity, and its relation to more local spacetime approaches.Prominent among the problems of quantum gravity are those of unitarity and observables.In a unitary theory with solutions approximating Minkowski space, the S-matrix (or, in four dimensions, related inclusive probabilities) should be sharply formulated and physical.Features of its perturbative description are reviewed.A successful quantum gravity theory should in particular address the questions posed by the ultrahigh-energy regime.Some control can be gained in this regime by varying the impact parameter as well as the collision energy.However, with decreasing impact parameter gravity becomes strong, first eikonalizing, and then entering the regime where in the classical approximation black holes form.Here one confronts what may be the most profound problem of quantum gravity, that of providing unitary amplitudes, as seen through the information problem of black hole evaporation.Existing approaches to quantum gravity leave a number of unanswered questions in this regime; there are strong indications that new principles and mechanisms are needed, and in particular there is a good case that usual notions of locality are inaccurate.One approach to these questions is investigation of the approximate local dynamics of spacetime, its observables, and its limitations; another is to directly explore properties of the gravitational S-matrix, such as analyticity, crossing, and others implied by gravitational physics.