Another example is Turkish Airlines 1951 at Amsterdam in 2009. Automatic approach, a problem with the radio altimeter caused the autothrottles to go to idle thrust at 2000 ft. The plane slowed down but the pilots reacted much too late. The plane stalled and crashed.
What's most important: the verdict was that this was pilot error. Even though the approach was being flown on autopilot (hey, doesn't that word mean that the plane will do everything automatically and the pilots can watch Harry Potter?), the pilots took the blame. And deservedly so.
Autopilot does not mean "sit back, relax and watch a movie".
And it's a lot safer if someone gets incapacitated, for example when having a stroke. The car keeps its lane, gradually slows down and activates the hazard lights.
Nope. They can land, but cannot take off. I know it seems counterintuitive, since landing is more difficult than taking off. But they want a pilot to have his or her hands on the controls so they can react quickly if an engine fails or some other unforseen event happens, requiring a split second decision between continuing the take-off and aborting. An engine failure during the approach is less critical due to the lower power setting.
Also, there is less of a need for automated take-offs as humans are capable of taking off in relatively low visibility, much lower than that required for landing.
A Tesla motor produces 362 horsepower and weighs 70 pounds. Even if a permanent magnet motor can be lighter by a factor of 2, it's not going to affect performance that much. The weight of the batteries, and the efficiency of the motor, are much more important. And for range, steady state low load performance is most important as most long range driving is done on a highway. Induction motors are better there. Permanent magnets might be better for city driving, but that's not really the limiting factor for range for most people unless you're driving a taxi.
Even lies, for example claims that the second accident was related in any way to autopilot ("according to police reports", while the actual police reports did not mention autopilot and there's no reason to believe it was on)
So they actually store the text of people's mistyped passwords (not as a hashed value) which might be only a single character different from their actual passwords? That doesn't seem like a very good idea if anyone gets hold of that database.
First of all, only one of the accidents resulted in fatalities. Not two.
The first accident was really caused by the truck crossing the road in front of oncoming traffic. It was a high trailer with lots of space underneath, so the Tesla's radar went underneath and did not see it. Neither did the camera which had trouble with the light truck on a light sky background and possibly confused it with an overhead sign. The driver was distracted and never even applied the brakes, not even at the very last moment before impact. That probably confirms that the truck really was hard to see. Also, Tesla owner's manual says the system should only be used on well marked highways without intersections.
The second accident probably did not even have autopilot on. Media reports everywhere were claiming that the police report said it was on, but it turns out the police report did not even mention autopilot at all. At this point, there is no reason to believe AP was on.
The third one appears to indeed be "caused" by autopilot while it was being used on a badly marked country road it was not designed for. The driver ignored just about every Tesla recommendation about the autopilot (it's still experimental, you should remain vigilant and be ready to take over at any time, and only use it on well marked highways without intersections). Dark country roads at 5 mph above the speed limit should be a definite no-no for anyone with common sense.
Once again, the first accident was the only one where someone got killed. Tesla's passive safety systems are outstanding. Not long ago there was a crash in the Netherlands where a truck agressively changed lanes without noticing the Tesla to his left, the collision was so violent that the Tesla rolled over two and a half times (which is quite extraordinary given its low center of gravity). Both occupants survived with only minor injuries. The driver had glass in his leg which came from a glass bottle in the cabin. His son has a cut on his leg that turned out to have come from his iPad flying around. They were extremely pleased with how safe the car was.
Yeah, before you know it they'll have another referendum, saying that disconnecting the internet will keep the foreigners out and save 100 billion pounds that will go straight to the NHS.
Airbus is pretty similarly limited. But then again, how often do you get high winds in foggy conditions? It only happens extremely rarely and it's not really worth the extra effort to develop and certify the system. Airlines can just divert to another airport or cancel the flight.
This is just sloppy reporting. A Tesla crashed, must be another autopilot case, don't bother to actually check the facts. Even make up nonexisting statements from police reports, why not.
The actual police report does not mention autopilot, contrary to what some news reports are claiming. Tesla has not been able to review the logs yet because the antenna got damaged in the accident. They have been trying to contact the owner in order to get access to the logs but so far have not been able to reach them.
At this point, it looks like this accident has nothing to do with autopilot.
I'm pretty sure they will do better than me on those particular intersections. A human driver who has never been there will have to slow down and go "WTF? Which light is for which road?". An automatic system will have these relatively rare intersections hard-coded into its database so it knows exactly which light is which.
Lots of accidents are caused by people who claim to have been blinded by the sun. Happens all the time. People just assume the road will be clear. Why would anyone be crazy enough to stand right in front of the sun?
The radar is problematic for stationary objects. It can't really distinguish between a stationary car and stationary traffic signs, stationary asphalt, etc... It's easier to detect a car that's going 3 km/h than it is to detect a stationary car. This explains why the autopilot usually stops just fine when the cars in front come to a stop (it can continue to track them all the way to zero), but will sometimes keep going if the car in front of you leaves your lane while the cars further ahead are standing still. You really do need the camera there (and they really ought to be using a stereo camera).
On top of that, the radar signal probably went right underneath the trailer. The beam has to be relatively narrow because otherwise it would be even worse at discriminating between cars and background. But that means it will occasionally miss things.
How can a web server crash if serving web pages is one of the main purposes of its existence?
It was a high tractor trailer with lots of open space underneath. The radar beam went right underneath (and so did most of the car). Also, the trailer was a light color against a light sky, so the camera missed it as well. And, crucially, even the driver failed to notice it because he did not brake at all, not even in the last fraction of a second before he hit that enormous impossible-not-to-notice hunk of metal.
The fact that the system is not perfect and still makes occasional mistakes, does not negate the fact that it's still the most advanced such system available in production cars today. Other constructors have similar systems in the pipeline but are a little more hesitant releasing them to the public just yet. Elon just barges ahead to get as much real life data as possible to accelerate further development. You can debate whether or not that's a wise choice, but that's another matter.
There have been a few cases where the autopilot failed to brake for another car, and the driver reacted too late. But in all those cases, the driver did end up using the brakes and hitting the other car at a relatively low speed.
In this case, apparently, the car did not brake at all. Either the driver had some sort of lapse of consciousness (stroke, seizure, asleep,...) or he was too distracted by his Harry Potter movie to notice the huge truck in front.
Or maybe the truck really was hard to see against the bright background, and the human driver did no better than the AI. Certainly not while looking at a movie.
TCAS is only a warning system and that's unilkely to change any time soon.
Airplanes on autopilot will happily fly into other airplanes, mountains, buildings, etc... TCAS and GPWS will give aural and visual warnings, but that's it. If the pilots ignore those warnings, the autopilot will just continue on its path.
By the way, TCAS relies on transponders in other aircraft that broadcast their position and altitude. Cars don't have those.
Slashdot Mirror
The radar missed the truck all right. The top of the car didn't.
It could, and has even been tested already. But they decided not to actually certify and install them.
So then Google knows who works at those bases...
Oh, so the US government cannot see it then? That will put the Chinese' mind at ease.
Another example is Turkish Airlines 1951 at Amsterdam in 2009. Automatic approach, a problem with the radio altimeter caused the autothrottles to go to idle thrust at 2000 ft. The plane slowed down but the pilots reacted much too late. The plane stalled and crashed.
What's most important: the verdict was that this was pilot error. Even though the approach was being flown on autopilot (hey, doesn't that word mean that the plane will do everything automatically and the pilots can watch Harry Potter?), the pilots took the blame. And deservedly so.
Autopilot does not mean "sit back, relax and watch a movie".
He's dead because he was watching a movie instead of the road. He didn't even brake at the last minute.
And it's a lot safer if someone gets incapacitated, for example when having a stroke. The car keeps its lane, gradually slows down and activates the hazard lights.
Nope. They can land, but cannot take off. I know it seems counterintuitive, since landing is more difficult than taking off. But they want a pilot to have his or her hands on the controls so they can react quickly if an engine fails or some other unforseen event happens, requiring a split second decision between continuing the take-off and aborting. An engine failure during the approach is less critical due to the lower power setting.
Also, there is less of a need for automated take-offs as humans are capable of taking off in relatively low visibility, much lower than that required for landing.
A Tesla motor produces 362 horsepower and weighs 70 pounds. Even if a permanent magnet motor can be lighter by a factor of 2, it's not going to affect performance that much. The weight of the batteries, and the efficiency of the motor, are much more important. And for range, steady state low load performance is most important as most long range driving is done on a highway. Induction motors are better there. Permanent magnets might be better for city driving, but that's not really the limiting factor for range for most people unless you're driving a taxi.
Even lies, for example claims that the second accident was related in any way to autopilot ("according to police reports", while the actual police reports did not mention autopilot and there's no reason to believe it was on)
So they actually store the text of people's mistyped passwords (not as a hashed value) which might be only a single character different from their actual passwords? That doesn't seem like a very good idea if anyone gets hold of that database.
First of all, only one of the accidents resulted in fatalities. Not two.
The first accident was really caused by the truck crossing the road in front of oncoming traffic. It was a high trailer with lots of space underneath, so the Tesla's radar went underneath and did not see it. Neither did the camera which had trouble with the light truck on a light sky background and possibly confused it with an overhead sign. The driver was distracted and never even applied the brakes, not even at the very last moment before impact. That probably confirms that the truck really was hard to see. Also, Tesla owner's manual says the system should only be used on well marked highways without intersections.
The second accident probably did not even have autopilot on. Media reports everywhere were claiming that the police report said it was on, but it turns out the police report did not even mention autopilot at all. At this point, there is no reason to believe AP was on.
The third one appears to indeed be "caused" by autopilot while it was being used on a badly marked country road it was not designed for. The driver ignored just about every Tesla recommendation about the autopilot (it's still experimental, you should remain vigilant and be ready to take over at any time, and only use it on well marked highways without intersections). Dark country roads at 5 mph above the speed limit should be a definite no-no for anyone with common sense.
Once again, the first accident was the only one where someone got killed. Tesla's passive safety systems are outstanding. Not long ago there was a crash in the Netherlands where a truck agressively changed lanes without noticing the Tesla to his left, the collision was so violent that the Tesla rolled over two and a half times (which is quite extraordinary given its low center of gravity). Both occupants survived with only minor injuries. The driver had glass in his leg which came from a glass bottle in the cabin. His son has a cut on his leg that turned out to have come from his iPad flying around. They were extremely pleased with how safe the car was.
Yeah, before you know it they'll have another referendum, saying that disconnecting the internet will keep the foreigners out and save 100 billion pounds that will go straight to the NHS.
Airbus is pretty similarly limited. But then again, how often do you get high winds in foggy conditions? It only happens extremely rarely and it's not really worth the extra effort to develop and certify the system. Airlines can just divert to another airport or cancel the flight.
This is just sloppy reporting. A Tesla crashed, must be another autopilot case, don't bother to actually check the facts. Even make up nonexisting statements from police reports, why not.
The actual police report does not mention autopilot, contrary to what some news reports are claiming. Tesla has not been able to review the logs yet because the antenna got damaged in the accident. They have been trying to contact the owner in order to get access to the logs but so far have not been able to reach them.
At this point, it looks like this accident has nothing to do with autopilot.
I'm pretty sure they will do better than me on those particular intersections. A human driver who has never been there will have to slow down and go "WTF? Which light is for which road?". An automatic system will have these relatively rare intersections hard-coded into its database so it knows exactly which light is which.
This sounds like a very short-sighted proposition as it consumes a resource that could be put to far better use for lunar colonisation.
They'll figure that out once they run out.
Lots of accidents are caused by people who claim to have been blinded by the sun. Happens all the time. People just assume the road will be clear. Why would anyone be crazy enough to stand right in front of the sun?
The radar is problematic for stationary objects. It can't really distinguish between a stationary car and stationary traffic signs, stationary asphalt, etc... It's easier to detect a car that's going 3 km/h than it is to detect a stationary car. This explains why the autopilot usually stops just fine when the cars in front come to a stop (it can continue to track them all the way to zero), but will sometimes keep going if the car in front of you leaves your lane while the cars further ahead are standing still. You really do need the camera there (and they really ought to be using a stereo camera).
On top of that, the radar signal probably went right underneath the trailer. The beam has to be relatively narrow because otherwise it would be even worse at discriminating between cars and background. But that means it will occasionally miss things.
But uh you had better stare blankly out the windshield for all six hours of your drive...
Six hours? In a Tesla?
How can a web server crash if serving web pages is one of the main purposes of its existence?
It was a high tractor trailer with lots of open space underneath. The radar beam went right underneath (and so did most of the car). Also, the trailer was a light color against a light sky, so the camera missed it as well. And, crucially, even the driver failed to notice it because he did not brake at all, not even in the last fraction of a second before he hit that enormous impossible-not-to-notice hunk of metal.
The fact that the system is not perfect and still makes occasional mistakes, does not negate the fact that it's still the most advanced such system available in production cars today. Other constructors have similar systems in the pipeline but are a little more hesitant releasing them to the public just yet. Elon just barges ahead to get as much real life data as possible to accelerate further development. You can debate whether or not that's a wise choice, but that's another matter.
Even in larger planes, all they do is warn ("traffic... traffic" or "terrain.. terrain... pull up!"). It's up to the pilot to take over.
There have been a few cases where the autopilot failed to brake for another car, and the driver reacted too late. But in all those cases, the driver did end up using the brakes and hitting the other car at a relatively low speed.
In this case, apparently, the car did not brake at all. Either the driver had some sort of lapse of consciousness (stroke, seizure, asleep,...) or he was too distracted by his Harry Potter movie to notice the huge truck in front.
Or maybe the truck really was hard to see against the bright background, and the human driver did no better than the AI. Certainly not while looking at a movie.
TCAS is only a warning system and that's unilkely to change any time soon.
Airplanes on autopilot will happily fly into other airplanes, mountains, buildings, etc... TCAS and GPWS will give aural and visual warnings, but that's it. If the pilots ignore those warnings, the autopilot will just continue on its path.
By the way, TCAS relies on transponders in other aircraft that broadcast their position and altitude. Cars don't have those.
Fortunately they still have a lot of immigrants from India doing entrepreneuring.