Maybe the 400 years referes to the number of people-years used to manually clean up map information?
Although frankly speaking, if anything 400 man-years seems very low for worldwide mapping.
edit: I... I skimmed the article the first time, but... The author actually claims that it's a data quality problem, not an algorithm problem. Huh?!
second edit: I guess my only real contention is that "data quality" is a fake idea.
Apple has had trouble with Big Data, in the parlance, but iCloud is still running. They've made some pretty dumb mistakes with their online services over the years, but they still do happen to be syncing data (if even just push notifications) to something on the order of a hundred million devices.
Catching up to where Google Maps is today will take somewhere between 10 days and 400 years. I'll bet good money that they have this nailed within a 95% tolerance before next year's new hardware model.
This is apart from MapMaker, and their dev teams, and their software improvements.
For instance, if you look at directions in India now, they're awesome, since they tell you landmarks to look for on the way [pass by * on left], at the intersections, etc. http://goo.gl/maps/b8w15 for instance.
My guess is 95% will take 3-4 years - once they accept they have a problem.
On the other hand, this actually is one of the rarer problems where throwing 10,000 contract workers from India for 5 months could solve much of the problem, especially since Google has already paved the way on the harder parts.
What I really want to see is a serious quality comparison between Apple and OSM.
edit: And the million dollar question... How much of Google's map data was correlated/verified against iPhone positioning data?
I can see how that would help reveal issues, but I don't see how it would help much in fixing them.
Fixing map data is.