Actually, the vast majority of posts here are sensible, excluding yours, and you do your damned best to bring down the IQ of this forum. There isn't a sane person who believes self-driving card will be the norm anytime soon. It's a pipe dream that has many more obstacles besides the technology itself.
Yes the 200 million cars on the road that are not self driving how do you makes sure one of em doesn't drive into you?
My business partner owns a BMW 5. Biggest piece of junk and he regrets getting it. Like clockwork, every three months something leaks/breaks/blows and off to the dealer it goes. The only thing reliable about BMW is their warranty. They break down so often and the price to repair it on each occurrence (had it not been under warranty) easily and effortlessly goes into the thousands of dollars.
I own a $23K BMW (motorcycle) that brand new has broken down so many times, that when I the time came out of frustration discuss a potential trade-in just to get rid of it and move on, the dealer refused to pay any money for a trade in. They conceded that my year/model bike had so many problems that no one would buy it, even though it looks brand new. I realized then that the only reasons BMW's cost so much is that they have to sell them at such a high price to factor in the repairs they will certainly require during the warranty period.
I can understand some semblance of relevance if the comparison was between an imperfect human and a perfect machine. But, between an imperfect human and an imperfect machine? Really?
C'mon...
Yes it was serious.
The repeated implication is that a car AI would be imperfect in many situations and would therefore be unacceptable, while usually implying humans are always superior in the given situations - which may, or may not, be the case. My point is that humans are not perfect - which is tolerated - but there seems to be a general implication that an AI that was less than perfect would not be tolerated.
Why not measure imperfect machine performance against imperfect human performance? If the machine's level of imperfect performance is statistically better, that would certainly make them worth considering.
An ECU is very simple. Ford used a 16 bit microcontroller with 16K of RAM for their PCMs up until the mid-2000s. The software fundamentally looks up a value from a table and writes it into a hardware register. No comparison at all with computer vision and all sorts of non-linear algorithms.
Brilliant avoidance of the topic of system redundancies and fail-safes.
I hope all this Apple car talk is as (in)accurate as all that erroneous rumor talk of Apple building a TV !
Perhaps it's simply an Apple control box,
for that impractical Tesla over $100k electric car, to add to the car collections of the overly rich.
Of course it's not a car . These apple folks are so caught up in their own insanity they would belive an appl aircraft Is coming
It's just some navigation software/0nboard control system doohickey and there were pics of A test rig outed a few weeks ago. The reality distortion field is really something to behold. Turns normal people stupid.
My business partner owns a BMW 5. Biggest piece of junk and he regrets getting it. Like clockwork, every three months something leaks/breaks/blows and off to the dealer it goes. The only thing reliable about BMW is their warranty. They break down so often and the price to repair it on each occurrence (had it not been under warranty) easily and effortlessly goes into the thousands of dollars.
I own a $23K BMW (motorcycle) that brand new has broken down so many times, that when I the time came out of frustration discuss a potential trade-in just to get rid of it and move on, the dealer refused to pay any money for a trade in. They conceded that my year/model bike had so many problems that no one would buy it, even though it looks brand new. I realized then that the only reasons BMW's cost so much is that they have to sell them at such a high price to factor in the repairs they will certainly require during the warranty period.
Hey stop it , I own a Bmw 3 series, now 15 years old 260k miles and it has been one of the most reliable cars I have ever owned. One or two data points doesn't justify you trashing of an entire brand
Of course it's not a car . These apple folks are so caught up in their own insanity they would belive an appl aircraft Is coming
It's just some navigation software/0nboard control system doohickey and there were pics of A test rig outed a few weeks ago. The reality distortion field is really something to behold. Turns normal people stupid.
Just curious how a self driving car would handle potholes, black ice, puddles. A big deal in the American Northeast.
My guess is that a properly programmed driving computer would be better able to recognize conditions that lead to black ice, including sensors to measure air temperature very near the ground, plus would have access to recent weather reports. The car computer would know if temperatures dropped below freezing last night; you on the other hand have to read a weather report in the morning and maybe it'll register in your head that black ice might be an issue. Or maybe it doesn't register, you're too focused on feeding the kids and getting them dressed for school.
Computers would be better able to detect a loss of traction as it occurs. After all, many new cars have electronic stability control and traction control systems. These systems have been around for years, just like anti-lock braking systems.
We already know that the existing self-driving cars have optical scans to detect objects, lane lines, etc., so pothole detection should be pretty easy. Moreover, a computer driven car would be able to host a database of local items. A pothole detected by a computer driven car could be uploaded to a central database, and other cars with access to the same database would know there's a pothole at that location.
Google purchased Waze a while back, the excellent crowd-sourced navigation app which has user-contributed traffic reports as well as things like potholes, disabled vehicles, etc. so much of the data is already there. Same thing with black ice, the car would record a black ice incident at a specific location. That location could be flagged as a site that has a tendency to form black ice.
If it was your neighborhood, you might know of these specific areas, but if you were driving in an unfamiliar place, you wouldn't know those spots where black ice forms. That's where a self-driving car would have an advantage.
I always use Waze while driving, not just for the turn-by-turn navigation (I often know where I'm going), but for the crowd-sourced traffic conditions: vehicle stopped on shoulder, debris on roadway, police, traffic cameras, etc.
As for puddles, a computer-driven car would have the optical sensors to detect puddles. The Google self-driving Lexus has been on local roads for a couple of years, so it is very likely that it has encountered inclement weather including rain puddles. Yes, California is in the midst of a drought, but it still rains occasionally.
Many roadways have a tendency to flood in highly specific areas after major rain, so again a computer-driven car would have access to a database of those locations.
My strong belief is that part of self-driving car programs in place is to devise a method for the car to report road conditions it encounters to a central database. Some, like puddles, would be flagged as a temporary condition. Others like potholes would have more permanence, until the database receives multiple confirmations that the pothole is gone (filled in). This is right up Google's alley.
Yeah, the extra bit of editorializing was wholly unnecessary. The late-1990's 5 Series used to be a great car.
And, now we have a four-cylinder as the standard engine in a 5 series and a three-cylinder coming to the 3 series. Both good engines as far as fours and threes are concerned, but it was disappointing to see the disappearance of the standard inline-six.
The repeated implication is that a car AI would be imperfect in many situations and would therefore be unacceptable, while usually implying humans are always superior in the given situations - which may, or may not, be the case. My point is that humans are not perfect - which is tolerated - but there seems to be a general implication that an AI that was less than perfect would not be tolerated.
Why not measure imperfect machine performance against imperfect human performance? If the machine's level of imperfect performance is statistically better, that would certainly make them worth considering.
"Why not measure imperfect machine performance against imperfect human performance?"
Because that would be an insanely ridiculous thing to do, in the context of driving in situations where there is nearly an infinite number of variables. At the end of the day, (most) human being possess something called common sense, and a survival instinct, which is often what is required when the unexpected happens. A car is not a plane that can have a flight program. So yes, until driverless cars are NEARLY perfect, there's no way in hell they can be allowed on the roads with everyone else, even if they are statistically more "perfect" than human beings.
I think the problem with self driving cars arises when there is an accident and the question of liability comes up.
How did we deal that issue when cruise control first appeared? We dealt with it because even though cruise control will automate maintaining a speed limit the driver is still the one in control. It's the same thing with a self-serving car. The next step is adaptive cruise control becoming more common place. Then you have additional automation added but the driver will still be in control… until it becomes a liability for the driver to maintain ultimate control, which may or may not happen in our lifetimes. Looking at aircraft, even with all the automation involved the pilots are still in control.
[VIDEO]
There is also the ethical conundrum - if the self driving car has to make a choice between crashing into a pedestrian on the sidewalk with a child in a stroller, or hitting a single person on a pedestrian crossing - how will it decide? The person on the crossing could be a child and the stroller could be a wheelbarrow. Or in another scenario - how will it choose to kill the passenger it carries or the person blocking the way? How will it asses the risk and if all goes wrong who is to blame?
There will always be an ethical issue to consider, but I don't understand that scenario. This assumes the car simply can't brake. Why can't it?
How did we deal that issue when cruise control first appeared? We dealt with it because even though cruise control will automate maintaining a speed limit the driver is still the one in control. It's the same thing with a self-serving car.
There's an article published today at one of the big news sites indicating that studies predict self-driving cars to have 90% fewer accidents than those fully under a human's control.
People are easily distracted and too often use poor judgement or react emotionally to other drivers. When self-driving cars finally arrive in force a lot of folks here pooh-poohing them may well change their tune.
"Why not measure imperfect machine performance against imperfect human performance?"
Because that would be an insanely ridiculous thing to do, in the context of driving in situations where there is nearly an infinite number of variables. At the end of the day, (most) human being possess something called common sense, and a survival instinct, which is often what is required when the unexpected happens. A car is not a plane that can have a flight program. So yes, until driverless cars are NEARLY perfect, there's no way in hell they can be allowed on the roads with everyone else, even if they are statistically more "perfect" than human beings.
I don't agree with you. You are saying even if the machines can be shown to be better drivers than humans, they still shouldn't be allowed until they are nearly perfect, which is illogical since humans are allowed to drive and be imperfect.
Say the annual road toll in the US is 40 K a year and the AI guided cars are shown to result in 15% fewer fatalities. You are saying it's somehow better to have 40 K dead than 34 K dead?
There's an article published today at one of the big news sites indicating that studies predict self-driving cars to have 90% fewer accidents than those fully under a human's control.
People are easily distracted and too often use poor judgement or react emotionally to other drivers. When self-driving cars finally arrive in force a lot of folks here pooh-poohing them may well change their tune.
I'd say it'll ultimately [help reduce accidents even more] than that, but for now I think people shouldn't focus on whatever sci-fi movie they picture in their heads when we talk about increasing automation in our automobiles.
IF self driving cars become ubiquitous then I don't believe folks will own cars anymore. Transportation will "be a service", whereby you order your vehicle of choice (maybe a nicer one for an evening out, or a larger capacity for hauling the kids to the soccer game, or even a truck or van for moving/hauling). It will come and pick you up, deliver you, and then await your return requests to take you home (perhaps self parking itself or running other errands for other clients instead of parking and waiting). There will be specialized cars for city driving versus urban and rural, all of which will tie into municipal transportation options (such long haul buses, trains, planes) providing seamless long distance and even international travel.
All extremely efficient, safe, reliable, and not unlike a software service on demand.
I, however, will die before I give up my two classic Porsches and the freedom to run the open roads.
Yup. On the off chance that Apple DOES get into designing cars, I can much more easily see it as a service, than something you own and all the headaches that will entail.
One of the most sensible posts to have graced the Apple Insider forums.
Actually, the vast majority of posts here are sensible, excluding yours, and you do your damned best to bring down the IQ of this forum. There isn't a sane person who believes self-driving cars will be the norm anytime soon. It's a pipe dream that has many more obstacles besides the technology itself.
Your opinion flies in the face of the evidence here at AI, where many people have confidently asserted that self-driving will be commonplace in 5-10 years.
They're nuts, but then so are the people who think the Apple Watch will be a success.
Comments
Yes the 200 million cars on the road that are not self driving how do you makes sure one of em doesn't drive into you?
I hope all this Apple car talk is as (in)accurate as all that erroneous rumor talk of Apple building a TV !
Perhaps it's simply an Apple control box,
for that impractical Tesla over $100k electric car, to add to the car collections of the overly rich.
WTF?
That hurts....
My business partner owns a BMW 5. Biggest piece of junk and he regrets getting it. Like clockwork, every three months something leaks/breaks/blows and off to the dealer it goes. The only thing reliable about BMW is their warranty. They break down so often and the price to repair it on each occurrence (had it not been under warranty) easily and effortlessly goes into the thousands of dollars.
I own a $23K BMW (motorcycle) that brand new has broken down so many times, that when I the time came out of frustration discuss a potential trade-in just to get rid of it and move on, the dealer refused to pay any money for a trade in. They conceded that my year/model bike had so many problems that no one would buy it, even though it looks brand new. I realized then that the only reasons BMW's cost so much is that they have to sell them at such a high price to factor in the repairs they will certainly require during the warranty period.
Is this even a serious question?
I can understand some semblance of relevance if the comparison was between an imperfect human and a perfect machine. But, between an imperfect human and an imperfect machine? Really?
C'mon...
Yes it was serious.
The repeated implication is that a car AI would be imperfect in many situations and would therefore be unacceptable, while usually implying humans are always superior in the given situations - which may, or may not, be the case. My point is that humans are not perfect - which is tolerated - but there seems to be a general implication that an AI that was less than perfect would not be tolerated.
Why not measure imperfect machine performance against imperfect human performance? If the machine's level of imperfect performance is statistically better, that would certainly make them worth considering.
An ECU is very simple. Ford used a 16 bit microcontroller with 16K of RAM for their PCMs up until the mid-2000s. The software fundamentally looks up a value from a table and writes it into a hardware register. No comparison at all with computer vision and all sorts of non-linear algorithms.
Brilliant avoidance of the topic of system redundancies and fail-safes.
Of course it's not a car . These apple folks are so caught up in their own insanity they would belive an appl aircraft Is coming
It's just some navigation software/0nboard control system doohickey and there were pics of A test rig outed a few weeks ago. The reality distortion field is really something to behold. Turns normal people stupid.
"these apple folks"—what are you doing here?
Just curious how a self driving car would handle potholes, black ice, puddles. A big deal in the American Northeast.
My guess is that a properly programmed driving computer would be better able to recognize conditions that lead to black ice, including sensors to measure air temperature very near the ground, plus would have access to recent weather reports. The car computer would know if temperatures dropped below freezing last night; you on the other hand have to read a weather report in the morning and maybe it'll register in your head that black ice might be an issue. Or maybe it doesn't register, you're too focused on feeding the kids and getting them dressed for school.
Computers would be better able to detect a loss of traction as it occurs. After all, many new cars have electronic stability control and traction control systems. These systems have been around for years, just like anti-lock braking systems.
We already know that the existing self-driving cars have optical scans to detect objects, lane lines, etc., so pothole detection should be pretty easy. Moreover, a computer driven car would be able to host a database of local items. A pothole detected by a computer driven car could be uploaded to a central database, and other cars with access to the same database would know there's a pothole at that location.
Google purchased Waze a while back, the excellent crowd-sourced navigation app which has user-contributed traffic reports as well as things like potholes, disabled vehicles, etc. so much of the data is already there. Same thing with black ice, the car would record a black ice incident at a specific location. That location could be flagged as a site that has a tendency to form black ice.
If it was your neighborhood, you might know of these specific areas, but if you were driving in an unfamiliar place, you wouldn't know those spots where black ice forms. That's where a self-driving car would have an advantage.
I always use Waze while driving, not just for the turn-by-turn navigation (I often know where I'm going), but for the crowd-sourced traffic conditions: vehicle stopped on shoulder, debris on roadway, police, traffic cameras, etc.
As for puddles, a computer-driven car would have the optical sensors to detect puddles. The Google self-driving Lexus has been on local roads for a couple of years, so it is very likely that it has encountered inclement weather including rain puddles. Yes, California is in the midst of a drought, but it still rains occasionally.
Many roadways have a tendency to flood in highly specific areas after major rain, so again a computer-driven car would have access to a database of those locations.
My strong belief is that part of self-driving car programs in place is to devise a method for the car to report road conditions it encounters to a central database. Some, like puddles, would be flagged as a temporary condition. Others like potholes would have more permanence, until the database receives multiple confirmations that the pothole is gone (filled in). This is right up Google's alley.
Yeah, the extra bit of editorializing was wholly unnecessary. The late-1990's 5 Series used to be a great car.
And, now we have a four-cylinder as the standard engine in a 5 series and a three-cylinder coming to the 3 series. Both good engines as far as fours and threes are concerned, but it was disappointing to see the disappearance of the standard inline-six.
Yes it was serious.
The repeated implication is that a car AI would be imperfect in many situations and would therefore be unacceptable, while usually implying humans are always superior in the given situations - which may, or may not, be the case. My point is that humans are not perfect - which is tolerated - but there seems to be a general implication that an AI that was less than perfect would not be tolerated.
Why not measure imperfect machine performance against imperfect human performance? If the machine's level of imperfect performance is statistically better, that would certainly make them worth considering.
"Why not measure imperfect machine performance against imperfect human performance?"
Because that would be an insanely ridiculous thing to do, in the context of driving in situations where there is nearly an infinite number of variables. At the end of the day, (most) human being possess something called common sense, and a survival instinct, which is often what is required when the unexpected happens. A car is not a plane that can have a flight program. So yes, until driverless cars are NEARLY perfect, there's no way in hell they can be allowed on the roads with everyone else, even if they are statistically more "perfect" than human beings.
How did we deal that issue when cruise control first appeared? We dealt with it because even though cruise control will automate maintaining a speed limit the driver is still the one in control. It's the same thing with a self-serving car. The next step is adaptive cruise control becoming more common place. Then you have additional automation added but the driver will still be in control… until it becomes a liability for the driver to maintain ultimate control, which may or may not happen in our lifetimes. Looking at aircraft, even with all the automation involved the pilots are still in control.
[VIDEO]
There will always be an ethical issue to consider, but I don't understand that scenario. This assumes the car simply can't brake. Why can't it?
People are easily distracted and too often use poor judgement or react emotionally to other drivers. When self-driving cars finally arrive in force a lot of folks here pooh-poohing them may well change their tune.
EDIT: Found the reference:
http://consumerist.com/2015/03/05/study-self-driving-vehicles-could-eliminate-90-of-car-accidents-in-united-states/
I'd hope it's more in the Panemera territory, but that's just me.
"Why not measure imperfect machine performance against imperfect human performance?"
Because that would be an insanely ridiculous thing to do, in the context of driving in situations where there is nearly an infinite number of variables. At the end of the day, (most) human being possess something called common sense, and a survival instinct, which is often what is required when the unexpected happens. A car is not a plane that can have a flight program. So yes, until driverless cars are NEARLY perfect, there's no way in hell they can be allowed on the roads with everyone else, even if they are statistically more "perfect" than human beings.
I don't agree with you. You are saying even if the machines can be shown to be better drivers than humans, they still shouldn't be allowed until they are nearly perfect, which is illogical since humans are allowed to drive and be imperfect.
Say the annual road toll in the US is 40 K a year and the AI guided cars are shown to result in 15% fewer fatalities. You are saying it's somehow better to have 40 K dead than 34 K dead?
I'd say it'll ultimately [help reduce accidents even more] than that, but for now I think people shouldn't focus on whatever sci-fi movie they picture in their heads when we talk about increasing automation in our automobiles.
edit: Clarification.
IF self driving cars become ubiquitous then I don't believe folks will own cars anymore. Transportation will "be a service", whereby you order your vehicle of choice (maybe a nicer one for an evening out, or a larger capacity for hauling the kids to the soccer game, or even a truck or van for moving/hauling). It will come and pick you up, deliver you, and then await your return requests to take you home (perhaps self parking itself or running other errands for other clients instead of parking and waiting). There will be specialized cars for city driving versus urban and rural, all of which will tie into municipal transportation options (such long haul buses, trains, planes) providing seamless long distance and even international travel.
All extremely efficient, safe, reliable, and not unlike a software service on demand.
I, however, will die before I give up my two classic Porsches and the freedom to run the open roads.
Yup. On the off chance that Apple DOES get into designing cars, I can much more easily see it as a service, than something you own and all the headaches that will entail.
One of the most sensible posts to have graced the Apple Insider forums.
Actually, the vast majority of posts here are sensible, excluding yours, and you do your damned best to bring down the IQ of this forum. There isn't a sane person who believes self-driving cars will be the norm anytime soon. It's a pipe dream that has many more obstacles besides the technology itself.
Your opinion flies in the face of the evidence here at AI, where many people have confidently asserted that self-driving will be commonplace in 5-10 years.
They're nuts, but then so are the people who think the Apple Watch will be a success.