Google TechTalk, given in 2006 by Sebastian Thrun
What are robotic cars? An autonomous vehicle that can drive on ordinary roads. They offer comfortable workspace, face-to-face, and they bring Moore's law to transportation, as soon as transportation becomes a software problem the pace of change will rapidly increase. They park, deliver and refuel themselves. Three new terms: robo-taxi, whistlecar and deliver-bot.
The DARPA grand challenge competitions was surprising, teams with very small budgets came up with amazing results, despite the almost total failure of the first competition.
When? Some technology is already here, many predict as early as 2020. But barrier may be more legal and social than technically. It does require breakthroughs, but noting compared to general AI. Needs cheaper hardware and heavily parallel hardware.
Why is it such a great thing? Accidents, human drivers kill 45,000 people a year, over a million people worldwide are killed in traffic accidents, and far more people are injured. We also don't realise how much infastructure we've given over to cars, there are six car parking places for every car in the US.
Self-delivery is almost as self-driving. Here we come to the robo-taxi (and the whistlecar which self-delivers but doesn't self-drive), which you summon and then takes you on the trip. The advantage is that you get the right vehicle for the trip. A 10-mile range electric trike allows you to pickup a truck or a van.
Energy is today's hot-button issue, so lets look at the electric car. Who killed the electric car? Conspiracy theories aside, the battery killed it. People won't tolerate the cost, limited range, long recharge times and poor availability of fast recharge. We can make an efficient short-range cars, but people don't want them.
Robots don't care how convenient recharging is, a robot car refuels/recharges itself and stations don't need to be on-route or close. This enables experimental fuels. You don't care about the range of your taxi, just that it will get your where you are going.
A lot of suggestions for futurist transport have actually been around for 40 years, unlike a lot of these robot cars use the existing infrastructure and can be bought be private individuals, and crucially by early adopters and alpha geeks. No matter how attractive the centrally planned system sounds, it won't out compete a system that grows from the bottom.
More advantages. No parking, no congestion, you can read and work in transit. You don't need a license and you don't need to own a car. It's cheaper and safer for accidents. That's by definition by the way, we' shouldn't allow them on the streets until they're safer for accidents than human driven cars.
You can already buy a lot of this technology off the shelf, self-parking, auto-spacing cruise control, lane departure prevention, road side reading, auto-braking and coming soon from VW; lane following, passing and parking.
We have to apply the School of Fish test, put a whole bunch of cars on a track and challenge people to drive amoungst them and crash into one. If the robot cars pass the test, people shouldn't be able to able to touch them, they'll just get out of the way.
Who will be behind the move to robot cars? The companies that want to sell them, accident victim, the alcohol companies, the environmentalist (once they see the advantages)...
What will stand in the way? The law, fear of the unknown, liability, terrorists, the technological challenges. What about software recalls? What happens when your car gets disabled when the manufacturer issues a safety advisory on the current firmware.
...and we're done.