There exists the "entropy problem" of the early universe, that is, why did the universe begin with an extremely low entropy and how did it evolve into such high entropy at late times? It has been long believed that inflation cannot be the solution since it requires an extremely low entropy to ever occur. However, we point out that since the inflation is always accompanied with a horizon, the correct probability of inflation is associated with the quantum entanglement entropy, which should in principle be larger than what considered previously. This motivates us to reexamine the issue by computing the evolution of the cosmological entanglement entropy in the early universe. We invoke a toy model of nonlinear generalized Chaplygin gas (GCG), which has the advantage of providing a smooth and unitary transition between the inflation epoch and the radiation dominant era. We found that soon after the onset of the inflation, the total entanglement entropy rapidly decreases to a minimum, and it rises monotonically afterwards throughout the remainder of the inflation and the radiation epochs. This indicates that the universe does not need to begin with an extremely low entropy; its smallness can be naturally induced by the dynamics of inflation itself. We believe that our computation largely captures the essential feature of entropy evolution and can provide us insights beyond the toy model.