This graph is smoothed out first by the logarithmic nature, and second by the fact that "estimates for the 1st century suggest uncertainty of the order of 50%"

This graph is smoothed out first by the logarithmic nature, and second by the fact that "estimates for the 1st century suggest uncertainty of the order of 50%" (and that's only 2000 years ago). The graph does not show us that anybody thinks the human population was flat before a certain point.