Slashdot Mirror
A Self-Driving Uber Car Went the Wrong Way On a One-Way Street in Pittsburgh (qz.com)
An anonymous reader writes: Uber driver Nathan Stachelek was pulled off to the side of the road when he saw the self-driving car turn the wrong way. It was the night of Sept. 26 and the car he had spotted, one of the autonomous Ford Fusions that Uber is testing in Pittsburgh, Pennsylvania, was heading through the city's Oakland neighborhood, just steps from the center of campus for the University of Pittsburgh. Stachelek watched the car turn off Bates Street and onto Atwood, a one-way road, going in the wrong direction. From a distance he couldn't tell whether the car was driving itself, or its human operator had made a mistake. Stachelek took out his phone in time to shoot a brief video of Uber's vehicle backing up and driving away, then uploaded it to Facebook. "Driverless car went down a one way the wrong way," he wrote. "Driver had to turn car around."
In all fairness, for self-driving cars to live up to the claims that proponents are making, they can't do this.
Murphy's Law of Computing #8:
To screw up is human, to screw up royally requires a computer.
The issue has never been one self driving car screwing up vs one driver screwing up - the machine will eventually beat the human there (and arguably it already has). The issue is that one mistake on a map update or some defect in the algorithm, and the possibility that you'll have cars full of people driving over the edge of an incomplete bridge for hours on end. That's always been my personal reluctance for enthusiastically embracing self driving cars.
I don't think anyone anywhere is claiming that self driving cares will be perfect.
That's just stupid to expect.
Lowering the 100,000+ deaths per year in the world due to humans driving is the actual goal.
There are two types of people in the world: Those who crave closure
At least they don't drive in the Lexus Lanes when they are only supposed to be used by Lexus owners and carpools....
he couldn't tell whether the car was driving itself, or its human operator had made a mistake
I do not block ads. I do block third party scripts.
We need regulating bodies and driverless car makers to agree on standards and zoning.
A driverless car has sensors, not eyes and spatial awareness. It has GPS and map data not a sense of direction.
If the data fed to the car says it can turn into oncoming traffic (and there are no vehicle so the sensors don't alert some wannabe AI) it will turn. Any human that might make the error will very quickly notice they are going the wrong way without the need for cars. the might notice how (most) cars are parked facing in a certain direction or road markings that give clues like "no entry" and the corresponding road markings.
Car AI cannot yet read these properly. Forget reading in time or when it's raining and the sign is slightly eroded or placed at an odd angle.
A human can spot a branch handing on power lines dangling in the wind, a sensor designed to avoid collisions with other cars cannot.
I'm certain that driverless cars will get much better and will very quickly be safer than a human driver despite these and other faults BUT to make it all so much safer we need approved zones. Like zoning for congestion or weight/height limits.
Car manufacturers will know that in these specific zones/highways they can expect a rather predictable set of road conditions. A human can drive the car out of some odd city intersection with angry aggressive drivers in rush-hour then switch to autopilot for that boring and predictable 100 mile highway journey. (Or not if you like that sort of driving)
When a driverless car can navigate A to B across a busy city in India it might be ready to do away with zoning but until then it's simply necessary and I believe it's just a matter of time until zoning happens.
A 'singular oddity' is an event that cannot be explained and only happens when you are alone.
Again, mistakes will happen. Self driving will not be perfect. To expect such is stupidity.
They are already an order of magnitude safer than the average driver.
They have already gone beyond their initial goals.
There are two types of people in the world: Those who crave closure
"There are literally millions of things that decent, off-the-shelf sensors can detect---things that humans cannot perceive, either due to sensory or attention limitations."
Yes that still does not reduce the GP's argument that there are also many problems that a computer-operated vehicle cannot perceive either. The best solution still seems to be a combination of the two: a human driver, and sensors/warnings/etc to augment him/her
For the one-way street example, sure. But what happens when someone incorrectly marks the bridge under construction as a passable road and several hundred commuters plunge off the end of it like lemmings?