Entropy and initial conditions in cosmology

I discuss the Boltzmann-Penrose question of why the initial conditions for cosmology have low entropy. The modern version of Boltzmann's answer to this question, due to Dyson, Kleban and Susskind, seems to imply that the typical intelligent observer arises through thermal fluctuation, rather than cosmology and evolution. I investigate whether this can be resolved within the string landscape. I end with a review of the suggestion that Holographic Cosmology provides a simpler answer to the problem. This paper is a revision of unpublished work from the spring of 2006, combined with my talk at the Madrid conference on String theory and Cosmology, Nov 2006.