I'm not saying humans don't make mistakes, but if it wasn't for a majority of humans being sensible, there would be a lot more. For every pedestrian death, humans have driven over 530 million miles successfully.
Tax payers had better not shell out for this until some company gives concrete evidence that they will actually be able to save lives in the long run, meaning they will be able to make this technology affordable and accessible for 95% of all people on the road. For now it hasn't gone beyond a geeks fantasy.
Stop being an apologist. A human driver still doesn't speed past them at full speed under the assumption they will not enter the road. They slow down in order to reduce the damage *in case* they walk into the road. We are told that CPUs can react much quicker than we can, 6 seconds was plenty to react somehow.
Apparently the way they had it was that the computer would drive and the driver would stop it from driving, if needed. That doesn't seem like an obviously ridiculous arrangement
Really? They test experimental rockets and roller coasters and and airplanes without the people around them knowing exactly what they are getting into? Pretty sure they don't launch rockets that might explode in the middle of Phoenix.
There is a big difference between Google having my conversation as one in a billion, and sending that conversation to someone random in my contact list. With that said, agreed that listing to people's conversions ever is pretty creepy. If an iPhone doesn't work for you then there aren't many choices.
We are all going to learn that international diplomacy is nothing like business dealings. Using the same tactics for both will end in failure and shows a real lack of maturity as a president.
I calculated +30% wrong. Should be 7,865,000 accidents a year / 409.408 miles without an accident / with waymo being 7.4% as safe.
Though accidents in the US are very heavily weighted towards winter conditions. "The deadliest states are all in the Upper Midwest or near the Great Lakes: Michigan (83), Pennsylvania (65), Indiana (49) and New York (46) round out the top five." https://www.usatoday.com/story...; which Waymo doesn't do. Until it becomes feasible to roll out of your driveway the morning after an ice storm and have a functional self-driving car without having to roll through a de-icing unit (and wipe everything so it is perfectly clean), the accident number to compare is probably closer to 5.5 million anyway.
This brings us to the national poll of unreported accidents, which finds that 30% of all crashes are unreported https://crashstats.nhtsa.dot.g.... So, fine, let's up that from 5.5 million accidents to 7.15 million accidents. Humans still drive 450,349 miles without an accident. Your link gives me a 404 so I can't look at the document you provided (and I will withhold personal comments). However, this link https://qz.com/1220576/the-rac... indicates Waymo has hit 30.500 miles without an interaction in November 2017. However without much consistency. Also, we know that Google is picking where and when they drive, don't drive in bad weather, constantly check sensors, etc etc. Even doing that, they seem to be 30,500/450,329 = 6.8% as safe as a human.
But these are all situations that a self driving car needs to be able to handle! If the self driving car would have plowed into an animal if there was no intervention, that's an accident. If it would plow into a construction zone, that's an accident. Why would it decide to pull over and stop for no reason? Why can't it handle a flat tire? These are all potential accidents. Are you saying hail messing with lidar wouldn't cause an accident eventually if there was no intervention? These are all very serious situations that would likely cause an accident. Not sure what point you are trying to make.
I suspect that a lot of people grow frustrated with other drivers, and then they equate that to 'that person is going to kill someone'. But really, they are just being annoying, not dangerous.
Humans don't learn collectively, they learn individually. On the other hand, self driving cars do learn collectively. They should learn exponentially faster with the number of cars on the road. It only takes a human a few months of practice on a beginners license to learn how to drive. That demonstrates a huge issue right there.
Why would a safety driver take control if a crash was not to happen? Sure maybe the car turned down the wrong way on a one way road and it happened to be empty at the time, but that should still be considered a crash because of its potential to actually cause one.
You are failing to put driving deaths into perspective against successful numbers. Humans drive 3.22 trillion miles a year and get into around 5.5 million accidents. No self driving car comes close to 585K miles without an accident.
Human driving deaths a year is only even a relevant number because of the huge number of miles they are out there successfully.
Slashdot Mirror
I'm not saying humans don't make mistakes, but if it wasn't for a majority of humans being sensible, there would be a lot more. For every pedestrian death, humans have driven over 530 million miles successfully.
The driver was looking down at their lap because Uber required them to check an ipad mounted on the dash as part of the test procedure.
From what I read above, that is the way Uber set up the car. The driver was checking some sort of console as part of the driving test procedure.
The NTSB says it is ok as long as they report how many miles they drive and how many people they hit in those miles.
Tax payers had better not shell out for this until some company gives concrete evidence that they will actually be able to save lives in the long run, meaning they will be able to make this technology affordable and accessible for 95% of all people on the road. For now it hasn't gone beyond a geeks fantasy.
Stop being an apologist. A human driver still doesn't speed past them at full speed under the assumption they will not enter the road. They slow down in order to reduce the damage *in case* they walk into the road. We are told that CPUs can react much quicker than we can, 6 seconds was plenty to react somehow.
Apparently the way they had it was that the computer would drive and the driver would stop it from driving, if needed. That doesn't seem like an obviously ridiculous arrangement
You mean... like Autopilot?
Really? They test experimental rockets and roller coasters and and airplanes without the people around them knowing exactly what they are getting into? Pretty sure they don't launch rockets that might explode in the middle of Phoenix.
There is a big difference between Google having my conversation as one in a billion, and sending that conversation to someone random in my contact list. With that said, agreed that listing to people's conversions ever is pretty creepy. If an iPhone doesn't work for you then there aren't many choices.
We are all going to learn that international diplomacy is nothing like business dealings. Using the same tactics for both will end in failure and shows a real lack of maturity as a president.
Oh yes, because international diplomacy is EXACTLY like making a business deal. For America's sake I hope so, but I can't see it ending well.
I calculated +30% wrong. Should be 7,865,000 accidents a year / 409.408 miles without an accident / with waymo being 7.4% as safe.
Though accidents in the US are very heavily weighted towards winter conditions. "The deadliest states are all in the Upper Midwest or near the Great Lakes: Michigan (83), Pennsylvania (65), Indiana (49) and New York (46) round out the top five." https://www.usatoday.com/story...; which Waymo doesn't do. Until it becomes feasible to roll out of your driveway the morning after an ice storm and have a functional self-driving car without having to roll through a de-icing unit (and wipe everything so it is perfectly clean), the accident number to compare is probably closer to 5.5 million anyway.
This brings us to the national poll of unreported accidents, which finds that 30% of all crashes are unreported https://crashstats.nhtsa.dot.g.... So, fine, let's up that from 5.5 million accidents to 7.15 million accidents. Humans still drive 450,349 miles without an accident. Your link gives me a 404 so I can't look at the document you provided (and I will withhold personal comments). However, this link https://qz.com/1220576/the-rac... indicates Waymo has hit 30.500 miles without an interaction in November 2017. However without much consistency. Also, we know that Google is picking where and when they drive, don't drive in bad weather, constantly check sensors, etc etc. Even doing that, they seem to be 30,500/450,329 = 6.8% as safe as a human.
If the wikipedia page is to be believed, it is for all motor vehicles, which includes commercial. Stats are from the NHTSA.
https://en.wikipedia.org/wiki/...
But these are all situations that a self driving car needs to be able to handle! If the self driving car would have plowed into an animal if there was no intervention, that's an accident. If it would plow into a construction zone, that's an accident. Why would it decide to pull over and stop for no reason? Why can't it handle a flat tire? These are all potential accidents. Are you saying hail messing with lidar wouldn't cause an accident eventually if there was no intervention? These are all very serious situations that would likely cause an accident. Not sure what point you are trying to make.
I have Android Auto in my vehicle and it doesn't understand my home address, no matter how much I try.
I suspect that a lot of people grow frustrated with other drivers, and then they equate that to 'that person is going to kill someone'. But really, they are just being annoying, not dangerous.
85,956,061*
3.22 trillion miles driven / 37461 = 859,56,061 miles driven without a death. That's pretty safe.
Humans don't learn collectively, they learn individually. On the other hand, self driving cars do learn collectively. They should learn exponentially faster with the number of cars on the road. It only takes a human a few months of practice on a beginners license to learn how to drive. That demonstrates a huge issue right there.
Why would a safety driver take control if a crash was not to happen? Sure maybe the car turned down the wrong way on a one way road and it happened to be empty at the time, but that should still be considered a crash because of its potential to actually cause one.
You have to realize, there are a great many truckers driving every day all day safely that weight the stats.
Five by my count, just by the people around me I know well enough.
You are failing to put driving deaths into perspective against successful numbers. Humans drive 3.22 trillion miles a year and get into around 5.5 million accidents. No self driving car comes close to 585K miles without an accident.
Human driving deaths a year is only even a relevant number because of the huge number of miles they are out there successfully.
What do you mean barely? In a lot of countries the signals are totally ignored.