Slashdot Mirror
How Autonomous Cars' Safety Features Clash With Normal Driving
An anonymous reader writes: Google's autonomous cars have a very good safety record so far — the accidents they've been involved in weren't the software's fault. But that doesn't mean the cars are blending seamlessly into traffic. A NY Times article explains how doing the safest thing sometimes means doing something entirely unexpected to real, human drivers — which itself can lead to dangerous situations. "One Google car, in a test in 2009, couldn't get through a four-way stop because its sensors kept waiting for other (human) drivers to stop completely and let it go. The human drivers kept inching forward, looking for the advantage — paralyzing Google's robot." There are also situations in which the software's behavior may be so incomprehensible to human passengers that they end up turning it off. "In one maneuver, it swerved sharply in a residential neighborhood to avoid a car that was poorly parked, so much so that the Google sensors couldn't tell if it might pull into traffic."
They are confused by BAD driving. People in general really really suck at driving and a computer will have problems with that.
Do not look at laser with remaining good eye.
Amber means stop you can do so safely:
https://www.gov.uk/government/...
It does not mean try to squeeze through because you think you have time. I know people who've had tickets for running the amber, despite being across the line before it went red.
Actually, autonomous cars are programmed to exceed the speed limit by up to 10 mph. This is done because Google deems it safer than driving at the speed limit and being slower than the other cars on the road.
http://gizmodo.com/googles-autonomous-car-is-programmed-to-speed-because-i-1624025227
http://www.bbc.com/news/technology-28851996
The NYT hates new car tech, especially EVs and robots. I seems to be in the pocket of some big vested interests (oil presumably, maybe other auto manufacturers who are falling behind). Remember the infamous Tesla Model S review by that Broder guy, where he did everything in his power to make it fail, exceeding the speed limit and slow-cooking himself with the heater etc.
This is just another hit-piece against autonomous cars. It might even be out to trash Tesla again, since they are introducing autopilot.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
The other get out for hitting a driver in front is if said driver pulled into your safe braking distance.
So if some idiot pulls out of a junction without looking and you go into the back of them as a result it is not your fault. Another one would be someone overtaking pulling in and then slamming the brakes on (the last one is often done as part of an insurance fraud).
There are a whole bunch of others as well, though they can be hard to prove if you don't have a dashboard camera.
Not your job to prevent others from speeding. In fact in most cases the law is written forcing you to yield the left lane to the speeder as they are traveling faster then you.
It is more dangerous for you to go slowly in the left lane and force the speeders to pass you on the right, then it is for you to just get the hell over.
And dogs and cats and raccoons and moose and cows and trees and power poles and snow and rain and potholes and road construction and miscellaneous debris and so on and so on ad infinitum.
If you are not allowed to question your government then the government has answered your question.
That's not really the take-away from the article. The issue is that even with this limited amount of real-world usage, the cars are running into scenarios that the programmers didn't anticipate and the software handles the scenario poorly as a result - more poorly than humans would. That is the issue. That is the limitation of the self-driving car. There will *always* be unanticipated events when driving - more so when there are more automated cars on the roads. The automated cars will handle those situations worse, on average, than human drivers do because human cognition is far, far deeper and more robust than the programs operating these cars. This is also typical of Google engineering. Build something that works for a particular situation, claim that you are geniuses, watch is fail in the real world that is more complicated than their Ivory Tower view of reality, but care not one bit because you make your money off the advertising associated with the thing rather than the thing itself.
Every time I've heard an expert (usually a college professor with a background in computer science, robotics, or automation) discuss existing self-driving cars (the Google car is almost always mentioned as an example), the experts always describe self-driving cars as something more highly programmed and rule-bound than actually autonomous.
They rely less on machine vision and more on extremely detailed and high-resolution saved maps versus driving the road they see in front of them. Sensors are used to determine hazards, but more for avoidance than some kind of self-guided navigation decisions.