A couple of things stand out for me from the article, one is that this sort of aerial image to detailed map was one of the things the US Government was doing with satellite images in the 70's, and with Google's compute resources if it this seems like the kind of 'side' use that would be encouraged. It is certainly a more useful use of compute resources than computing cryptohashes :-).
The second thing was that Google owns all of their data. When I was there Maps was pretty new, Google was digesting the acquisition of the Keyhole folks (who became Google Earth) and was complaining bitterly about how restrictive the usage rights were for the map data from a major mapping company. They kicked off a program to become independent of third parties for map data and to have the most accurate data in the world.
And if you turn the problem into an algorithm that can be parceled up between machines, you have way more of that than people. It's the one thing that the previous generation of computer companies didn't get, when computers are cheap having 50,000 of them isn't all that much different than having 5,000 especially if you can code up ways for them to route around damage. I think Maps is a good indication of what you can do when you think like that.
> And if you turn the problem into an algorithm that can be parceled up between machines, you have way more of that than people. It's the one thing that the previous generation of computer companies didn't get
And some open source data initiatives. Open Street Map, for all it's glory, is surprisingly anti-automation. There's a lot of stuff that could probably be done with programs that they refuse because it's too hard to peer review the massive resulting changesets. Which is unfortunate, because there's a huge potential for processing images from places like Planet to provide new data and annotations that we'll just straight up miss out on.
Detailed maps are certainly valuable. Getting those from satellites is impressive, and requires high-resolution imagery as well as lots of computing power.
We can get some satellite imagery for free from weather satellites, but it's nowhere near the resolution needed for street maps.
Sadly you are stuck buying it from a satellite company (expensive) or buying your own satellites (more expensive). My feeling is that 'Maps' is Google's insurance policy against the erosion of search advertising margin.
I could easily see Google making turn by turn or map detail a "small" $4.99/month service rather than having it be free.
The second thing was that Google owns all of their data. When I was there Maps was pretty new, Google was digesting the acquisition of the Keyhole folks (who became Google Earth) and was complaining bitterly about how restrictive the usage rights were for the map data from a major mapping company. They kicked off a program to become independent of third parties for map data and to have the most accurate data in the world.
And if you turn the problem into an algorithm that can be parceled up between machines, you have way more of that than people. It's the one thing that the previous generation of computer companies didn't get, when computers are cheap having 50,000 of them isn't all that much different than having 5,000 especially if you can code up ways for them to route around damage. I think Maps is a good indication of what you can do when you think like that.