Of course, this doesn’t always apply

[[email protected]]

A sophisticated program would know that reversing might not be necessary if the self-driving car encountered another self-driving car—in which case, perhaps, the cars could establish wireless communication and exchange intent data. Or if the self-driving car could predict, based on the behavior of others, that the other vehicle was driven by a human, who would need no additional incentive to go first. Or, of course, if reversing would risk a collision.

The broader implications of this technology are that humans, for example, don’t think through every absurd scenario on the road, and so a robot would have to be programmed to do so—an impossible task anyway. (Just consider all the weird things operators have seen during Google car test drives: a man playing the saxophone while driving, a woman chasing a list of qatar cell phone numbers turkey with a mop. What would a robot think—and, more importantly, do—if it encountered such crossings or drivers?)

Robots that learn from experience free engineers from having to hand-code algorithms for situations that no one else could possibly anticipate. But a learning machine also means understanding that not all human drivers will react the same way, and that micro-movements can predict whether a human will act in a particular way (such as speeding up when allowed to pass, or moving to the side to let a car behind them pass). That level of awareness could be key to achieving broader social acceptance of self-driving cars.

“I can’t remember who told me this years ago, but someone once joked that once artificial intelligence works, you can’t call it artificial anymore,” says Julie Shah, head of the Interactive Robotics division at the Massachusetts Institute of Technology. In other words, once self-driving cars are proven to be a reality, they will just be cars (or drivers).

SoftBank's Pepper Robot Can Recognize Human Emotions
SoftBank's Pepper Robot Can Recognize Human Emotions
http://www.softbank.jp/en/robot/
If we go further, there is a paradox in all this that reflects the general direction of development of modern robots. The more diverse the goals that machines solve, the better they will be able to adapt to different types of people - and ultimately to each individual character.