Why Self-Driving Cars Should Never Be Fully Autonomous (roboticstrends.com)

An anonymous reader writes: David Mindell, an MIT professor, says self-driving cars should never be fully autonomous. "There's an idea that progress in robotics leads to full autonomy. That may be a valuable idea to guide research but when automated and autonomous systems get into the real world, that's not the direction they head. We need to rethink the notion of progress, not as progress toward full autonomy, but as progress toward trusted, transparent, reliable, safe autonomy that is fully interactive: The car does what I want it to do, and only when I want it to do it." Mindell writes, "Google's utopian autonomy is a more brittle, less functional solution than a rich, human-centered automation."

Comments

Re:What if I don't want to own a car?

[radish]

Data, please? People make this claim all the time, but given that there are over a billion trips a day in the US and only around 120 fatalities, I'd say humans drivers pretty much have this thing down. The fact that people can make it around in their cars in myriad weather conditions, successfully navigate unfamiliar terrain, and quickly respond to sudden changes in circumstances (kid darting out in front of them) speaks volumes to how good human drivers are.

So I'm going to try. Putting aside fatalities, as the Google cars have not been involved in any, there were approx 5.5m traffic accidents in the US in 2010. Taking your number of 1b trips per day, we get a figure of ~66k miles per accident. According to Google they have been involved in 11 accidents over 1.7 million miles which is ~154k miles/accident. Now this is a combination of fully automatic and driver assisted miles, so the comparison isn't exact, but it's pretty safe to assume the computer is at least as good as your average driver. And maybe twice as good.

I watched a Google self-driving car cross an intersection this weekend (in Austin). It was moving very cautiously and then slowed down to a walking pace on the other side of the intersection, leaving a trail of human-driven cars stuck in the intersection while it decided to turn down a side street.

That may be evidence of it driving badly. Or it may be evidence of it driving well, because it was responding to a potential danger that the human drivers didn't see or didn't care about. Remember - if we're saying we want the computers to drive better than humans we have to accept that they will at times drive differently.