This is an example of projecting current trends and projecting them far into the future. Those of you who remember Popular Mechanics from the 50's can recall flying cars, nuclear energy too cheap to meter, atomic powered planes, domed cities on the moon, etc. Sure, the urbanization trend will continue for a decade or two. But if you project some trends we already know about: an information based economy, ubiquitous very high speed internet, and robotics, the need for super-megalopolises is less apparent. My SO and I live in a town of 7K with a nearby city of 50K, and LA and SF are 4 hours away. If high speed intercity rail became as common as pre-1940's passenger rail, we would be half an hour from fast transit to anywhere. I dont see a need to live with 20M other folks.
These trends, projected far into the future, evoke points of conversation along the lines of "The Singularity" which as been a recent hot topic here on HN.
You touch on a great notion that for some trends, projection makes relative sense (especially in IT). Trends like population seem a little less sound to me; these could be easily interrupted by so many destructive factors (disease, politics, genocides...) that it seems somewhat senseless to plot.
I could be wrong though, and the population trends could be more stable than anything.
Predictions of population growth have confounded nearly everyone back to Malthus. The interesting trend is that the middle class seem to reproduce at a rate too low to sustain the population. Japan and Italy below the rate to sustain their populations. By 2100 the global population will probably be heading downward from a peak a few decades earlier.
No one really wants to throw a global cataclysm into the stew: nuclear war, pandemic, Yellowstone eruption, asteroid strike.
I remember reading predictions from a hundred years ago. (No, I'm not that old.) It missed all the mini-singularities, even the ones that had been already invented, like the automobile and the telephone.
