I once drove in rush hour traffic in london where a man waved and I stopped the car. I stopped because he was a policeman and he was stopping cars because a pipe had just then burst just past where he was and left a man sized pot hole. A Google car isn't going to stop for every person waving from the sidewalk of a busy street or it will never move. To drive a car properly a human needs knowledge of the world, and real intelligence. The intelligence to ignore, in this case, all pedestrians except one.
Unless the coder anticipated this, and the hundreds of thousands of other eventualities, the self driving cars will crash when their simplistic models of the world fail.
Throwing more hardware at bad software will just make these bad decisions faster.
This is so easy ... To stop a driverless car, a policeman or other emergency worker will have a paging-like system in a holster. That police officer had to be notified of the issue and then get himself into position for his "human intelligence" with the hand wave manoeuvre. LOL
If a Google car can navigate a freeway, today, you don't think humans are capable of technology more sophisticated than a hand wave to get cars to slow down or stop?
I once drove in rush hour traffic in london where a man waved and I stopped the car. I stopped because he was a policeman and he was stopping cars because a pipe had just then burst just past where he was and left a man sized pot hole. A Google car isn't going to stop for every person waving from the sidewalk of a busy street or it will never move. To drive a car properly a human needs knowledge of the world, and real intelligence. The intelligence to ignore, in this case, all pedestrians except one.
Unless the coder anticipated this, and the hundreds of thousands of other eventualities, the self driving cars will crash when their simplistic models of the world fail.
Throwing more hardware at bad software will just make these bad decisions faster.
I bet our anonymous Apple employee here was referring to a new project at Apple so enticing that it would "give Tesla a run for its money" in the talent recruitment wars, a project likely not related to automobiles.
This is so easy ... To stop a driverless car, a policeman or other emergency worker will have an paging-like system in a holster. That police officer had to be notified of the issue and then get himself into position for his "human intelligence" with the hand wave manoeuvre. LOL
If a Google car can navigate a freeway, you don't think humans are capable of technology more sophisticated than a hand wave to get cars to slow down or stop?
Would make a shit getaway car.
I just gave one example of a situation which has to be thought about. It's "easy" because you've be given the problem. Nevertheless your solution is human intervention. Now we have to buy all police men and all government officials these devices, hope they all work, and hope the officials use them fairly and competantky and don't stop all traffic all the time for minor events. We also have to worry about security of the devices or their protocols so they can't be used by criminals, hackers or teenagers.
Jumping from "cars will slowly become more intelligent" to the singilarity is a pretty remarkable leap.
Recorded Human history is around 7000 years give or take? Is that long enough for AI to become a problem? If not in our lifetime, a reasonable period of uninterrupted tech advances make AI more likely.
Will self driving cars enhance or detract from the advent of artifical general intelligence? What about self landing rockets? Direct brain interfaces?
Yeah, because you're reaction times are better than an always-on computerized system.
Your selective quoting is not doing you any favours. I was replying to a guy who answered my thought experiment about how a driverless car couldn't recognise a cop flagging it down. he du ggested giving police forces devices to stop my car. I welcome electronic aids but want control in the final say. I want to drive as I am cleverer than a machine.
Let the jury note that solipism is not replying cogently to any of the technical or legal points put to him but is engaging in ad hominems, presuming the conclusion, strawman arguments and selective quoting.
I welcome electronic aids but want control in the final say.
So you're afraid of not having control. Common, but irrational. Pray tell, what electronic aids do you turn off right now because you think you can do a better job than the computer?
I want to drive as I am cleverer than a machine.
Of course you are¡ :rolleyes:
Let the jury note that solipism is […] replying cogently to […] selective quoting.
Maybe try getting a username right before attempting to slander.
Google's LiDAR-equipped driverless vehicles have been safely navigating U.S. roads since 2012.
The company is now in talks with several major companies including Mercedes-Benz, General Motors, Continental Automotive Systems, Nissan, Toyota, Audi, and Volvo.
Yet testing for driverless vehicles is open in just three states: Nevada, Florida, and California.
And while the California DMV may have already released regulations on the testing of autonomous vehicles, it missed a deadline for a second set of regulations required by Jan 1. The DMV cited “safety concerns” for the delay.
Meanwhile, Google's driverless vehicles have driven over 700,000 miles without getting into a single accident not caused by a human driver (one driverless vehicle was rear-ended, and another crashed while under human control).
For perspective, the average U.S. driver has an accident every 165,000 miles, according to data from the Federal Highway Administration, meaning the technology has proven about four times safer than human drivers.
It's pretty difficult to hide under the guise of “safety concerns” with data like that.
Well, until I've seen the traffic mix and what operations those cars are attempting (and the road conditions) it is hard to assess that its 4 times safer. But, considering the number of accidents due to plain distraction, tiredness, impaired driving, it wouldn'T be surprising that at least 2-3 times safer would be the norm no matter what Google did.
Ha ha. The singularity is an extreme joke (didn't you get that).
Infinities (yes you have more than one) make everything possible.
Don't believe everything Ray says, he is insane you know.
Self driving cars will creap in gradually over the next 15-20 years. By 2040, 80% of cars on the road will be self-driven. Yes, that soon. Ironically, it may be because cars with full assist have less accidents (and thus cost less to insure).
A few options :
a) They provide all sort of services impossible now. Sharing car rides will become incredibly easy, car routes could be optimized to pick up several near passengers needing a ride in the viccinity. That would enable the speed of cars with the savings of public transport.
b) Another option, smaller pod cars could go get individuals and then join together to become much faster trains of cars on the freeway. Maybe this "train" would be driven by a powerful, more efficient, lead vehicule, that relay themselves on freeway stretches getting power from central lines. That way you get the efficiency of public transport and the end point comfort of the individual vehicule (that scenario though i NOT in 25 years :-). Doing so, it would even be possible to have different lanes for tje "trains" with different speeds for local traffic versus long distance traffic. Inner traffic of those road trains could go to 120mph.
Also, you could combine direct local public transport (a) with those road trains (b) to create extremely, very high density, high speed transport. Maybe having self-adjusting standing seats in the local public transport so when they join the trains, they can get to a very high speed because of sparser traffic 120-140mph. With something like that, many people that take cars now wouldn't mind spending very little on transport that'S just as fast as their cars. Of course, some would still want to pay more for the individual pods. Everyone could do so if they wish. Options are good.
I totally agree with you foggyhill about this development. I think it's obvious it will be something like that. It will become very impractical and costly to drive your own car, plus it will take you for ever to get anywhere ...and be quite risky. When all the other vehicles out there exchange info about their positions and driving-intents in real time, you will be nothing but a roadblock sitting their in your manually navigated SUV. What will your excuses be then?
Of course, liability will be directed to the owners of the vehicles, like in most countries already now. BTW, who's liable today if your car suddenly starts rolling down the street by itself and then hits a kid?
<span style="line-height:1.4em;">I totally agree with you foggyhill about this development. I think it's obvious it will be something like that. It will become very impractical and costly to drive your own car, plus it will take you for ever to get anywhere ...and be quite risky. When all the other vehicles out there exchange info about their positions and driving-intents in real time, you will be nothing but a roadblock sitting their in your manually navigated SUV. What will your excuses be then?</span>
Of course, liability will be directed to the owners of the vehicles, like in most countries already now. BTW, who's liable today if your car suddenly starts rolling down the street by itself and then hits a kid?
The manufacturer if the handbrake is faulty. The owner if it isn't and he didn't engage.
Same with brakes. You can't blame drivers for faulty brakes. Another reason why this won't work.
When we have control over something we can be held responsible. If not we can't be. Without significant changes to the law -- changes which would be thrown out in court -- we would be effectively taxied about by the software. In fact driverless cars would, if they worked at all would probably be hires. No point in showing off in a car you can't accelerate or control. What are you paying for.
The software will either way be a service agreement to convoy you safely from a-b using external and internal tech. If it fails to be safe of course the manufacturer will be sued. If a car goes off the road now and there are no manufacturing defects the blame lies with the driver. If there are manufacturing defects responsible for the crash the survivors can sue.
In the "driverless" car the responsibility will lie with the driver or manufacturer which in this case is the same entity. The car isn't really driverless. It's just human driverless. Don't imagine a car which is driven by Google or Apple software and which is responsible for plowing into school kids because it's mapping was wrong, and the occupients not in control, wouldnf see major class action and possibly criminal suits for Apple/Google and its executives. It would.
Most Europeans drive manual. If automatic gear changes can be resisted then this pie in the sky nonsense can be ignored.
The tech on the car looks a bit more like the kind for driverless cars. The vehicles were apparently registered through a 3rd party company, not directly to Apple. If you look at Google's Street View car, it has a dome of cameras and the cameras are all at the same location so that you can get a 360 degree view:
[VIDEO]
The Google Street View car also doesn't seem to have any rotating devices on it. Their driverless car does:
If those are all cameras then they must be capturing data differently than the Google car. They could be capturing the data similar to how the iOS panoramas are captured. The control on the wheel will give an accurate measure of speed to be able to counter blur and control the shutter. That data might look a bit odd if they projected a single panorama down a street but it would be more efficient to view and they could so some 3D object reconstruction.
If they are sampling depth data with the spinning sensors, they can be building 3D streets. If someone holds an IR camera near the car, they should be able to see dots of light being emitted all over the place. They haven't used the sensors from the Kinect company they bought yet. That would work well with turn by turn directions as you could be watching the street as you move down it continuously and see the signs. Google Maps just moves point to point when it should really behave more like an on-rails first person game. They can eliminate humans from the scene more easily this way too vs blurring them.
The following video shows situations where driverless cars would really suck:
[VIDEO]
They give too much freedom to cyclists. What automated driving would miss out on is aggression. The result of this is shown in the following clip where the previously aggressive bus driver was replaced with a more passive one and stuck at a junction:
[VIDEO]
If the car waits too much for situations to resolve themselves, they could be waiting a long time and frustrating passengers and motorists behind. When a human being moves a car in an aggressive way e.g deciding it's their turn to move ahead, other drivers react to it and it's still safe. Driverless cars would work best if all vehicles were driverless including bicycles and motorcyclists.
"It is thought that driverless cars could eventually save motorists six working weeks a year in driving time, improve safety, reduce congestion and cut emissions. But there are years of rigorous testing ahead before they become an everyday reality."
In some cases they are essential such as for elderly, disabled, blind passengers who want to be more independent.
Environmental factors are another thing to consider. The car can't see an incoming tornado, a huge puddle near a bus stop, icy/muddy roads etc.
Environmental factors are another thing to consider. The car can't see an incoming tornado, a huge puddle near a bus stop, icy/muddy roads etc.
I find that incredibly hard to believe, "can't". My iPhone can tell me right now about construction zones, weather issues, and road conditions. EDIT: the 4th video you posted, demonstrates that Google cars already can anticipate construction detours, what looks like an intoxicated cyclist, necessary lane changes etc. So it can see construction workers and drunk bicyclists and other obstacles in the road, but not giant puddles or tornados. Oooookkkaaayyy.
Comments
It means the world isn't a freeway.
I once drove in rush hour traffic in london where a man waved and I stopped the car. I stopped because he was a policeman and he was stopping cars because a pipe had just then burst just past where he was and left a man sized pot hole. A Google car isn't going to stop for every person waving from the sidewalk of a busy street or it will never move. To drive a car properly a human needs knowledge of the world, and real intelligence. The intelligence to ignore, in this case, all pedestrians except one.
Unless the coder anticipated this, and the hundreds of thousands of other eventualities, the self driving cars will crash when their simplistic models of the world fail.
Throwing more hardware at bad software will just make these bad decisions faster.
You or someone else mentioned the "singularity". Just referring back to that.
No shit, Cumberbatch!
Nope.
If a Google car can navigate a freeway, today, you don't think humans are capable of technology more sophisticated than a hand wave to get cars to slow down or stop?
Would make a shit getaway car.
I just gave one example of a situation which has to be thought about. It's "easy" because you've be given the problem. Nevertheless your solution is human intervention. Now we have to buy all police men and all government officials these devices, hope they all work, and hope the officials use them fairly and competantky and don't stop all traffic all the time for minor events. We also have to worry about security of the devices or their protocols so they can't be used by criminals, hackers or teenagers.
Easier to stop a car myself. Thanks.
Yeah, because you're reaction times are better than an always-on computerized system.
Recorded Human history is around 7000 years give or take? Is that long enough for AI to become a problem? If not in our lifetime, a reasonable period of uninterrupted tech advances make AI more likely.
Will self driving cars enhance or detract from the advent of artifical general intelligence? What about self landing rockets? Direct brain interfaces?
Your selective quoting is not doing you any favours. I was replying to a guy who answered my thought experiment about how a driverless car couldn't recognise a cop flagging it down. he du ggested giving police forces devices to stop my car. I welcome electronic aids but want control in the final say. I want to drive as I am cleverer than a machine.
Let the jury note that solipism is not replying cogently to any of the technical or legal points put to him but is engaging in ad hominems, presuming the conclusion, strawman arguments and selective quoting.
So you're afraid of not having control. Common, but irrational. Pray tell, what electronic aids do you turn off right now because you think you can do a better job than the computer?
Of course you are¡ :rolleyes:
Maybe try getting a username right before attempting to slander.
Common sense, reason, logic, history, every industry Apple has ever entered…
Refusing to trust a technology that is nowhere near 1. safe 2. complete 3. reliable ? refusing to trust industrial technology.
They here and they work.
Google's LiDAR-equipped driverless vehicles have been safely navigating U.S. roads since 2012.
The company is now in talks with several major companies including Mercedes-Benz, General Motors, Continental Automotive Systems, Nissan, Toyota, Audi, and Volvo.
Yet testing for driverless vehicles is open in just three states: Nevada, Florida, and California.
And while the California DMV may have already released regulations on the testing of autonomous vehicles, it missed a deadline for a second set of regulations required by Jan 1. The DMV cited “safety concerns” for the delay.
Meanwhile, Google's driverless vehicles have driven over 700,000 miles without getting into a single accident not caused by a human driver (one driverless vehicle was rear-ended, and another crashed while under human control).
For perspective, the average U.S. driver has an accident every 165,000 miles, according to data from the Federal Highway Administration, meaning the technology has proven about four times safer than human drivers.
It's pretty difficult to hide under the guise of “safety concerns” with data like that.
Well, until I've seen the traffic mix and what operations those cars are attempting (and the road conditions) it is hard to assess that its 4 times safer. But, considering the number of accidents due to plain distraction, tiredness, impaired driving, it wouldn'T be surprising that at least 2-3 times safer would be the norm no matter what Google did.
Ooooookay... Goodbye.
Self driving cars will creap in gradually over the next 15-20 years. By 2040, 80% of cars on the road will be self-driven. Yes, that soon. Ironically, it may be because cars with full assist have less accidents (and thus cost less to insure).
A few options :
a) They provide all sort of services impossible now. Sharing car rides will become incredibly easy, car routes could be optimized to pick up several near passengers needing a ride in the viccinity. That would enable the speed of cars with the savings of public transport.
b) Another option, smaller pod cars could go get individuals and then join together to become much faster trains of cars on the freeway. Maybe this "train" would be driven by a powerful, more efficient, lead vehicule, that relay themselves on freeway stretches getting power from central lines. That way you get the efficiency of public transport and the end point comfort of the individual vehicule (that scenario though i NOT in 25 years :-). Doing so, it would even be possible to have different lanes for tje "trains" with different speeds for local traffic versus long distance traffic. Inner traffic of those road trains could go to 120mph.
Also, you could combine direct local public transport (a) with those road trains (b) to create extremely, very high density, high speed transport. Maybe having self-adjusting standing seats in the local public transport so when they join the trains, they can get to a very high speed because of sparser traffic 120-140mph. With something like that, many people that take cars now wouldn't mind spending very little on transport that'S just as fast as their cars. Of course, some would still want to pay more for the individual pods. Everyone could do so if they wish. Options are good.
Of course, liability will be directed to the owners of the vehicles, like in most countries already now. BTW, who's liable today if your car suddenly starts rolling down the street by itself and then hits a kid?
The manufacturer if the handbrake is faulty. The owner if it isn't and he didn't engage.
Same with brakes. You can't blame drivers for faulty brakes. Another reason why this won't work.
When we have control over something we can be held responsible. If not we can't be. Without significant changes to the law -- changes which would be thrown out in court -- we would be effectively taxied about by the software. In fact driverless cars would, if they worked at all would probably be hires. No point in showing off in a car you can't accelerate or control. What are you paying for.
The software will either way be a service agreement to convoy you safely from a-b using external and internal tech. If it fails to be safe of course the manufacturer will be sued. If a car goes off the road now and there are no manufacturing defects the blame lies with the driver. If there are manufacturing defects responsible for the crash the survivors can sue.
In the "driverless" car the responsibility will lie with the driver or manufacturer which in this case is the same entity. The car isn't really driverless. It's just human driverless. Don't imagine a car which is driven by Google or Apple software and which is responsible for plowing into school kids because it's mapping was wrong, and the occupients not in control, wouldnf see major class action and possibly criminal suits for Apple/Google and its executives. It would.
Most Europeans drive manual. If automatic gear changes can be resisted then this pie in the sky nonsense can be ignored.
[VIDEO]
The Google Street View car also doesn't seem to have any rotating devices on it. Their driverless car does:
http://www.extremetech.com/extreme/181508-googles-self-driving-car-passes-700000-accident-free-miles-can-now-avoid-cyclists-stop-for-trains
[VIDEO]
[VIDEO]
If those are all cameras then they must be capturing data differently than the Google car. They could be capturing the data similar to how the iOS panoramas are captured. The control on the wheel will give an accurate measure of speed to be able to counter blur and control the shutter. That data might look a bit odd if they projected a single panorama down a street but it would be more efficient to view and they could so some 3D object reconstruction.
If they are sampling depth data with the spinning sensors, they can be building 3D streets. If someone holds an IR camera near the car, they should be able to see dots of light being emitted all over the place. They haven't used the sensors from the Kinect company they bought yet. That would work well with turn by turn directions as you could be watching the street as you move down it continuously and see the signs. Google Maps just moves point to point when it should really behave more like an on-rails first person game. They can eliminate humans from the scene more easily this way too vs blurring them.
The following video shows situations where driverless cars would really suck:
[VIDEO]
They give too much freedom to cyclists. What automated driving would miss out on is aggression. The result of this is shown in the following clip where the previously aggressive bus driver was replaced with a more passive one and stuck at a junction:
[VIDEO]
If the car waits too much for situations to resolve themselves, they could be waiting a long time and frustrating passengers and motorists behind. When a human being moves a car in an aggressive way e.g deciding it's their turn to move ahead, other drivers react to it and it's still safe. Driverless cars would work best if all vehicles were driverless including bicycles and motorcyclists.
https://uk.news.yahoo.com/driverless-cars-set-testing-across-britain-020405967.html
"It is thought that driverless cars could eventually save motorists six working weeks a year in driving time, improve safety, reduce congestion and cut emissions.
But there are years of rigorous testing ahead before they become an everyday reality."
In some cases they are essential such as for elderly, disabled, blind passengers who want to be more independent.
Environmental factors are another thing to consider. The car can't see an incoming tornado, a huge puddle near a bus stop, icy/muddy roads etc.
[IMG ALT=""]http://forums.appleinsider.com/content/type/61/id/55287/width/500/height/1000[/IMG]
Mapping seems like the more obvious route for Apple but it's likely driverless cars will appear in some form.
Indeed; a self-driving car is quite probably going to wash out its brakes.
Environmental factors are another thing to consider. The car can't see an incoming tornado, a huge puddle near a bus stop, icy/muddy roads etc.
I find that incredibly hard to believe, "can't". My iPhone can tell me right now about construction zones, weather issues, and road conditions. EDIT: the 4th video you posted, demonstrates that Google cars already can anticipate construction detours, what looks like an intoxicated cyclist, necessary lane changes etc. So it can see construction workers and drunk bicyclists and other obstacles in the road, but not giant puddles or tornados. Oooookkkaaayyy.
So it can see construction workers and drunk bicyclists and other obstacles in the road, but not giant puddles or tornados.
Yes, because they’re completely different concepts.