But are there some things that self-driving vehicles simply can't do? Tasks that humans will always perform better?
Indeed, but Google is working on a fix for one of them.
There are several areas where computers have a hard time keeping up with us primates. They're iffy when it comes to pattern recognition, their lack of aggression can be dangerous, and they lack a moral compass, which could pose problems when faced with life-or-death decisions (like crashing into an embankment and killing the "driver" or running into a group of people).
Autonomous cars also have another shortcoming, one that has directly to do with pedestrians.
At some point, we've all found ourselves standing on a street corner, wanting to cross but unsure whether the car waiting at the stop sign has seen us. We might wave a bit, try to catch the driver's eye, and then, once she's seen us and smiled or nodded or honked, we zip across to the other side.
How do you do that when there's no driver, though? Patent #9,196,164 that's now owned by Google suggests a way.
Google envisions a screen or signing system located on the side of an autonomous vehicle. As you can see from the drawing submitted with the patent application, that screen might tell a pedestrian that the car is preparing to accelerate and to stay put. Or, it could give a thumbs-up, saying that it's a-okay to cross.
What's fascinating about this patent isn't so much what it does, though that's undoubtedly important. Instead, it's the step-by-step description of the computer's analytical process -- a process that humans take for granted. If you have time, skim the background and brief summary of the patent, found about a third of the way down the front page. No wonder Google, Tesla, and others are fighting for top engineering talent.