Cosmology has no predictive power without a theory of initial conditions. Because of the singularity theorems of Penrose and myself, many people assume that the initial state is necessarily of trans-Planckian curvature. We have no ideas of how to formulate initial conditions in such situations. String theory, at least in the form we know it, is based on perturbations about flat space, and so would break down along with classical general relativity. However, the singularity theorem relevant to cosmology, though not that for black holes, depends on the strong energy condition: TabVaVb 1 2 TVaVa for any timelike or null vector Va. This is always satisfied by gauge fields, but can be violated locally by scalar fields. It is therefore possible for the universe either to bounce or to approach a de Sitter state in the past. Such non-singular solutions form only a small subset of the space of all scalar field gravity solutions, but I shall show that the no boundary condition implies that they provide the dominant contribution to the present state. The curvature of the universe need never have been at the Planck level, and the birth of the universe can have been entirely within the semi classical domain. String theory is not necessary for cosmology.
展开▼