I will present a framework and a combined empirical-computational program that explores what cortical neural representation could underlie our intelligent behavior. I will start with showing how probabilistic internal representations could be implemented in the cortex in a sampling-based manner, and how this sampling-based probabilistic approach can explain a wide range of puzzling behavioral observations such as illusions and generalizations, physiological findings such as trial-to-trial variability and correlational patterns in neural responses, as well as provide a functional role to the high level of spontaneous activity in the brain. Next, I provide a confirmation of this proposed framework by demonstrating that as young animals grow, the visually evoked and spontaneous activity in their brains becomes statistically similar, indicating how their internal model gets tuned to the structure of their environment. I will also provide evidence that these changes are related to sensory experience rather than simply to developmental programs. Specifically, I will show that the general development of spontaneous and evoked activities is robust to visual deprivation, but the fine-tuning of their statistical similarity requires normal visual experience. This suggests a continuous ongoing integration of sensory experience into internal models of the visual environment. Time permitting, I will also show how sampling based probabilistic representation in cortical computation gives a new role to time in decision making and in top-down effects.