The success of WMAP was made possible by two factors: (1) the precise measurement of CMB anisotropies; and, (2) the precise theory of linear perturbations. While surveys of galaxies have been playing an important role in constraining cosmology for three decades, it has become increasingly clear that the current theory of the galaxy power spectrum is not adequate for interpreting the accurate observational data from the planned galaxy surveys, e.g., HETDEX, WFMOS, CIP, etc. Can we do better? In this talk I argue that higher-order perturbation theory is going to make galaxy surveys at high redshifts, i.e. z > 2, extremely powerful in determining cosmological parameters, such as the shape of the primordial power spectrum (tilt and running index) as well as the neutrino mass, which cannot be measured very well by CMB data alone. Specifically, I shall review the latest results from our on-going efforts to model and correct non-linearities in the power spectrum resulting from: (a) non-linear gravitational evolution; (b) non-linear bias; and, (c) non-linear redshift space distortion. The results, thus far, are very promising.