While it very well may be safer, the data doesn't exist to prove it.
The NHSTA disagrees with you. Their last investigation into an autopilot incident specifically looked very closely at data along specific routes and found a 40% reduction in crashes.
Well, right. But that makes it look even better. If the system is only used in 50% of driving conditions and eliminates 40% of ALL accidents, then it must be preventing a HUGE percentage of the accidents that were going to happen in that 50% of all driving.
This is especially true if accidents become more common as driving conditions deteriorate, which I take as a given.
I guess it becomes less true if Teslas are generally driven less in those deteriorating conditions. I have no real reason to believe this is so, but it wouldn't surprise me to find out that some Tesla owners pamper their cars and only take them out when it's nice.
No, that is not sound statistical logic. Off highway driving, where people would tend to not use Auto-pilot, is where you would expect more accidents. Highway driving has less accidents per mile than city or off-highway driving.
Unfortunately that data is not normalized for only driving conditions where Autopilot is used, highway driving in clear conditions, no rain, snow, fog, no adverse traffic pattern's,etc.
It's not clear from the article whether the requests to take the wheel were the normal periodic "I'm going to make sure you're paying attention" requests, or a "Hey, I don't know how to handle this situation, you need to do it!" request.
Which might indicate an intricacy of the human factors challenges with driver assist. We also don't know if the driver was actually watching the road, just didn't have a tight grip on the wheel. Lots to be learned over a long period of time, and much to early to make reliable claims as to safety benefits.
You've seen thousands of news stories about Tesla crashes? Because there have been thousands of Tesla crashes.
When the NHTSA last investigated, they found that the rate of crashes with Autopilot on was 40% less than with it off. Hence they supported its further use.
Again, just like Musk, you abused statistics. Was it 40% less "fatal or serious injury' crashes under the same circumstances? that is highway driving in good conditions, no snow, rain, or wet roads or fog? No adverse traffic patterns?
Electric cars and vehicles are definitely the way of the future. Especially when we get to the point where there are AI cars that can run as shuttles to and from regularly scheduled buses.
But, in the case of Tesla, I don't really get why anybody is giving them money to reserve a car that hasn't yet been built and may not be built for quite some time. I have noticed a lot more Teslas around here lately, but they're going to be really popular with poor people during the upcoming revolution as they'll be the second people stoned, just after those Prius drivers.
One is that there is no risk. You can get your money back, so no commitment required. One other is that only the first "x" number of buyers can get the tax credit, so best to get on the list. Of course, Tesla is also a popular brand and people with the money to spare want them.
Maybe some of that is because Musk demands attention. You can't just limit it to the good stuff. Also, he makes safety claims he can't support with data.
There is special interest in autonomous driving, so therefore it will get attention. Would you rather have apathy?
A hell of a post. But lets step back and ask questions about this accident. For example, one that stands out to me is, "the Tesla knew it was approaching an non-moving object, and it did not stop, or veer? Why?
Its a perfectly good question. The Tesla manual clearly states that it cannot be relied upon to stay in the traffic lane or avoid obstacles in its path. But you'd think it would recognize such an obstacle. Why did it veer at all?
I agree these would be better numbers for grading Tesla.
You can't make the comparison using news stories though, because when a Tesla crashes without Autopilot engaged it's just another car crash. It doesn't make national news. This makes the numbers appear slanted against Autopilot.
I don't know. It seems every fatal Tesla crash is already a story before they determine if Auto Pilot was engaged or not. This story is one of those.
Six seconds is pushing the brake to the floor and being at a complete stop, with plenty of room to spare; even more time to slightly turn the wheel to avoid hitting something 150m away and not moving.
Driver not paying attention AT ALL.
Six seconds is the amount of time the driver was not engaging the steering wheel. We do not know how much time there was after the car veered out of the traffic lane and before it hit the barrier, it my have been much less.
It amazes me that often people don't recognize that driving a car is a potentially extremely dangerous activity. 100% attention is required at all times, particularly since other drivers often do things they shouldn't do.
Unfortunately, nobody can claim they pay 100% attention at all times and be telling the truth. Everybody has a moment when they are distracted, good drivers quickly re-engage their minds.
Which is why we need to be very careful about technologies that give people a false sense of confidence that they can take their attention of of driving for longer periods of time.
In several articles about this accident, Musk goes on with stats about the safety of autonomous driving. I understand why Musk wants to make it clear the driver was negligent in his use of Auto-pilot, but he should not be making unsupported safety claims regarding autonomous driving, nor conflating them with Tesla Auto pilot safety. While it very well may be safer, the data doesn't exist to prove it. Tesla Auto-pilot is, per Tesla, used only on limited access highways when there is good visibility. It is not used in rain, snow, or fog. It is not to be used where traffic conditions are changing rapidly. So the comparable safety data should be limited to sedans traveling on limited access highways in nice weather with stable traffic conditions. Furthermore, the comparison should be based on number of fatal accidents and not number of deaths, which is higher where there are more people in a vehicle. A car that crashes with 4 people in it is not for times more dangerous to drive than a car that crashes with only one person.
In addition, the data should account for accidents not caused by the sedan, such as a tractor trailer suddenly coming across the median a taking out a car, or other 'unpreventable' incidents that neither an autonomous or human controlled car could avoid.
Comparison against total highway deaths is apples and oranges.
Musk is a smart man, smart enough to know how to use statistics properly. I believe he is quite aware his claims are not supported by existing data. It is disappointing and unnecessary. If I want to abuse statistics, I'd say the data clearly shows that on the particular day of this accident it was thousands of times safer to be in a human driven vehicle when passing the deficient barrier than in an auto-piloted Tesla.
"One ghost image requires thousands of exposures [each with a custom sandpaper filter], and the x-rays add up. In addition, the more detailed the object—for example, a human body—the more exposures you need. However, Wu says the x-ray intensity per exposure can be made low enough that ghost imaging may come out ahead."
The one thing you certainly don't want to do is sacrifice image quality. X-Rays, even if you get a lot of them, are extremely low risk. Missing something due to a poor image could be a significant risk.
Thank you for contacting Facebook(tm) Support. Have a nice day!
Facebook can gladly provide the names of the authors of those illegal ads. The parties that create those ads should be held liable.
Except for the small detail that the Fair Housing Act makes printing, making, or publishing such ads illegal. Not just making.
Of course there are some "safe-harbors" that publishers can attempt to use use.
Fair point. And those safe harbors 'should' be sufficient if FB shows they've taken reasonable measures, which they seemingly have. If I publish a paper, and someone sneaks in and changes something, or intentionally words something that might not catch the eye of an editor, it is hard to see culpability on my part. There appears to be intent to slip something past FB here. There is clearly no intent on FB's part to publish such ads.
The tablet form factor is not as useful in education as the notebook form factor.
My kids schools use Chromebooks and they are used all the time. Kids need to be able type up reports, do research, move files, etc. No tablet is good for that. Keyboard and mouse/mouspad are the right tools for that job.
proposed model for testing an autonomous car driving amidst normal traffic conditions that does not include actually having it drive among normal road traffic?
Hire people operate vehicles on closed tracks to simulate traffic and specific scenarios.
Or, better yet, orchesate dozens of other self driving cars to move in fixed choreographed patterns that the autonomous vehicle must recognize and react appropriately to.
I'd be interested to see just what has been done by various AV developers with regard to comtrolled scenario testing. Obviously you can't re-create every possible scenario, but you could create hundreds of very likely ones. A person crossing the road in various manners would be part of that testing regimen. Seeing what happens with certain sensors failed might be another. It is certainly something worthy of a question. I assume much has been done but I'd love to see the checklist(s).
At some point though, they need to be out on the roads and get the real world testing.
To me this incident illustrates just how hard it really will be to get fully AV on the road any time soon. I know many are enthusiastic about it, but its still at least 10 years away, IMO, except maybe for very limited functions such as certain defined truck routes.
Slashdot Mirror
While it very well may be safer, the data doesn't exist to prove it.
The NHSTA disagrees with you. Their last investigation into an autopilot incident specifically looked very closely at data along specific routes and found a 40% reduction in crashes.
Please provide a citation. I have not found that.
Nice come back. I see you have no actual point to make.
Well, right. But that makes it look even better. If the system is only used in 50% of driving conditions and eliminates 40% of ALL accidents, then it must be preventing a HUGE percentage of the accidents that were going to happen in that 50% of all driving.
This is especially true if accidents become more common as driving conditions deteriorate, which I take as a given.
I guess it becomes less true if Teslas are generally driven less in those deteriorating conditions. I have no real reason to believe this is so, but it wouldn't surprise me to find out that some Tesla owners pamper their cars and only take them out when it's nice.
No, that is not sound statistical logic. Off highway driving, where people would tend to not use Auto-pilot, is where you would expect more accidents. Highway driving has less accidents per mile than city or off-highway driving.
Unfortunately that data is not normalized for only driving conditions where Autopilot is used, highway driving in clear conditions, no rain, snow, fog, no adverse traffic pattern's,etc.
It's not clear from the article whether the requests to take the wheel were the normal periodic "I'm going to make sure you're paying attention" requests, or a "Hey, I don't know how to handle this situation, you need to do it!" request.
Which might indicate an intricacy of the human factors challenges with driver assist. We also don't know if the driver was actually watching the road, just didn't have a tight grip on the wheel. Lots to be learned over a long period of time, and much to early to make reliable claims as to safety benefits.
You've seen thousands of news stories about Tesla crashes? Because there have been thousands of Tesla crashes.
When the NHTSA last investigated, they found that the rate of crashes with Autopilot on was 40% less than with it off. Hence they supported its further use.
Again, just like Musk, you abused statistics. Was it 40% less "fatal or serious injury' crashes under the same circumstances? that is highway driving in good conditions, no snow, rain, or wet roads or fog? No adverse traffic patterns?
I assume you can't answer that.
Electric cars and vehicles are definitely the way of the future. Especially when we get to the point where there are AI cars that can run as shuttles to and from regularly scheduled buses.
But, in the case of Tesla, I don't really get why anybody is giving them money to reserve a car that hasn't yet been built and may not be built for quite some time. I have noticed a lot more Teslas around here lately, but they're going to be really popular with poor people during the upcoming revolution as they'll be the second people stoned, just after those Prius drivers.
One is that there is no risk. You can get your money back, so no commitment required. One other is that only the first "x" number of buyers can get the tax credit, so best to get on the list. Of course, Tesla is also a popular brand and people with the money to spare want them.
Maybe some of that is because Musk demands attention. You can't just limit it to the good stuff. Also, he makes safety claims he can't support with data.
There is special interest in autonomous driving, so therefore it will get attention. Would you rather have apathy?
A hell of a post. But lets step back and ask questions about this accident. For example, one that stands out to me is, "the Tesla knew it was approaching an non-moving object, and it did not stop, or veer? Why?
Its a perfectly good question. The Tesla manual clearly states that it cannot be relied upon to stay in the traffic lane or avoid obstacles in its path. But you'd think it would recognize such an obstacle. Why did it veer at all?
I agree these would be better numbers for grading Tesla.
You can't make the comparison using news stories though, because when a Tesla crashes without Autopilot engaged it's just another car crash. It doesn't make national news. This makes the numbers appear slanted against Autopilot.
I don't know. It seems every fatal Tesla crash is already a story before they determine if Auto Pilot was engaged or not. This story is one of those.
Six seconds is pushing the brake to the floor and being at a complete stop, with plenty of room to spare; even more time to slightly turn the wheel to avoid hitting something 150m away and not moving.
Driver not paying attention AT ALL.
Six seconds is the amount of time the driver was not engaging the steering wheel. We do not know how much time there was after the car veered out of the traffic lane and before it hit the barrier, it my have been much less.
It amazes me that often people don't recognize that driving a car is a potentially extremely dangerous activity. 100% attention is required at all times, particularly since other drivers often do things they shouldn't do.
Unfortunately, nobody can claim they pay 100% attention at all times and be telling the truth. Everybody has a moment when they are distracted, good drivers quickly re-engage their minds.
Which is why we need to be very careful about technologies that give people a false sense of confidence that they can take their attention of of driving for longer periods of time.
In several articles about this accident, Musk goes on with stats about the safety of autonomous driving. I understand why Musk wants to make it clear the driver was negligent in his use of Auto-pilot, but he should not be making unsupported safety claims regarding autonomous driving, nor conflating them with Tesla Auto pilot safety. While it very well may be safer, the data doesn't exist to prove it. Tesla Auto-pilot is, per Tesla, used only on limited access highways when there is good visibility. It is not used in rain, snow, or fog. It is not to be used where traffic conditions are changing rapidly. So the comparable safety data should be limited to sedans traveling on limited access highways in nice weather with stable traffic conditions. Furthermore, the comparison should be based on number of fatal accidents and not number of deaths, which is higher where there are more people in a vehicle. A car that crashes with 4 people in it is not for times more dangerous to drive than a car that crashes with only one person.
In addition, the data should account for accidents not caused by the sedan, such as a tractor trailer suddenly coming across the median a taking out a car, or other 'unpreventable' incidents that neither an autonomous or human controlled car could avoid.
Comparison against total highway deaths is apples and oranges.
Musk is a smart man, smart enough to know how to use statistics properly. I believe he is quite aware his claims are not supported by existing data. It is disappointing and unnecessary. If I want to abuse statistics, I'd say the data clearly shows that on the particular day of this accident it was thousands of times safer to be in a human driven vehicle when passing the deficient barrier than in an auto-piloted Tesla.
"One ghost image requires thousands of exposures [each with a custom sandpaper filter], and the x-rays add up. In addition, the more detailed the object—for example, a human body—the more exposures you need. However, Wu says the x-ray intensity per exposure can be made low enough that ghost imaging may come out ahead."
The one thing you certainly don't want to do is sacrifice image quality. X-Rays, even if you get a lot of them, are extremely low risk. Missing something due to a poor image could be a significant risk.
Is it possible that some studies are done purely to spark discussions on places like this and otherwise serve no purpose and provide no value at all?
Next they'll ban defensive language and we won't be able to say anything.
Solution: don't post illegal ads
Thank you for contacting Facebook(tm) Support. Have a nice day!
Facebook can gladly provide the names of the authors of those illegal ads. The parties that create those ads should be held liable.
Except for the small detail that the Fair Housing Act makes printing, making, or publishing such ads illegal. Not just making.
Of course there are some "safe-harbors" that publishers can attempt to use use.
Fair point. And those safe harbors 'should' be sufficient if FB shows they've taken reasonable measures, which they seemingly have. If I publish a paper, and someone sneaks in and changes something, or intentionally words something that might not catch the eye of an editor, it is hard to see culpability on my part. There appears to be intent to slip something past FB here. There is clearly no intent on FB's part to publish such ads.
But hey, sometimes it just depends on the judge.
Provide a platform that lets people commit tax evasion and see how long you avoid prison.
Craigslist. Ebay, Amazon. All allow you to avoid paying sales taxes.
problem solved.
Solution: don't post illegal ads
Thank you for contacting Facebook(tm) Support. Have a nice day!
Facebook can gladly provide the names of the authors of those illegal ads. The parties that create those ads should be held liable.
I guess he needs to bone up on his company security policy and business model.
Well, he can't exactly bring the Russians with him to explain how things work.
The tablet form factor is not as useful in education as the notebook form factor.
My kids schools use Chromebooks and they are used all the time. Kids need to be able type up reports, do research, move files, etc. No tablet is good for that. Keyboard and mouse/mouspad are the right tools for that job.
IPads make great PR though.
I think that was Waymo that set up the town.
Uber has the technology already developed to punish jaywalkers.
proposed model for testing an autonomous car driving amidst normal traffic conditions that does not include actually having it drive among normal road traffic?
Hire people operate vehicles on closed tracks to simulate traffic and specific scenarios.
Or, better yet, orchesate dozens of other self driving cars to move in fixed choreographed patterns that the autonomous vehicle must recognize and react appropriately to.
I'd be interested to see just what has been done by various AV developers with regard to comtrolled scenario testing. Obviously you can't re-create every possible scenario, but you could create hundreds of very likely ones. A person crossing the road in various manners would be part of that testing regimen. Seeing what happens with certain sensors failed might be another. It is certainly something worthy of a question. I assume much has been done but I'd love to see the checklist(s).
At some point though, they need to be out on the roads and get the real world testing.
To me this incident illustrates just how hard it really will be to get fully AV on the road any time soon. I know many are enthusiastic about it, but its still at least 10 years away, IMO, except maybe for very limited functions such as certain defined truck routes.