The serious money is in the data products, I believe. The Google cars work by localizing themselves in very high quality maps obtained by post-processing data obtained by cars that drove the route already at least twice. The maps are checked and corrected by humans, for example to decide where the center of each lane is, and where to look for traffic lights.

The software alone is of limited value without access to the database. The current database of a subset of California's roads is about 20GB. They'll need to freshen it frequently, but renting access to the database is what I think the business model is in the long term.

From a good-for-mankind perspective, probably. From Google's perspective I really don't think so. If they can perfect that technology they will have a great opportunity to be the only company in the world able to license the ability to automate the transportation of people and goods using existing vehicles and roadways. It would be a gold mine.

Well technically, they could open source and still have patents on it. That way they can be the only supplier of autonomous car software for car manufacturers and still allow people to play with the code.