Newer Tesla Model S cars will be able to basically do the same thing this summer with its auto-pilot 'lane holding' firmware update.
It may not be obvious to those who aren't paying close attention to the advancement of self-driving technology, but driving hundreds of miles on a highway is actually fairly easy for today's AI and requires only a basic sensor stack (GPS, HD camera with IR for nighttime, 600 ft radar sensor up front, and a slew of sonar sensors for close up decisions). Lane holding and traffic-aware cruise control together basically can take you 99% of the way to any city from any other on the interstate -- and this is likely coming to a car near you (not just $100k luxury cars) in the next 5 years.
Beyond the highway though, is where the current technology falls apart... The difference between maintaining a lane on the highway and driving in a suburban neighborhood is orders of magnitude in complexity. Google's super-fancy Lidar-based "driverless" cars still have tons of trouble navigating in cities and suburban settings. I'm not even going to approach the topic of weather, but to illustrate some of the challenges, in a city you might run into:
* Roads with inadequate, faded or absent lanes markings.
* Intersections with no stop/yield signs or broken/flashing traffic lights
* Vast distances with no speed limit signs
* Random and unpredictable people, animals and inanimate objects crossing or blowing across the road (e.g. a raccoon, kid on bike, plastic bag, paint bucket, police officer with hand up each may require a completely different reaction from the driver and the inappropriate reaction could put occupants or pedestrians in serious danger.)
All of these are relatively easy for people to navigate, but pose significant challenges for AI. I predict that all-weather, door-to-door, autonomous driving is closer to 10-20 years away -- perhaps 10 years for high-end vehicles and 20 years for your run-of-the-mill Toyota Corolla, etc... (Think of the rollout of GPS navigation or airbags.)