A human driver generally would have had the intuition the the walker would or could have walked into the path.
Yeah? Is that why 5,000+ pedestrians were killed by cars in the US in 2015, while this is the first time a pedestrian was struck and killed by an autonomous vehicle since they started on roads in 2009?
10 million *carefully selected* miles with optimal conditions
Seriously, do at least the minimum research. You'll find out, for example, that the reason they test in San Francisco is because of the very high density and the number of obstacles they run into per mile (construction zones, vehicles improperly parked on narrow streets, streetlights without power, jaywalkers, cyclists being cyclists, emergency vehicles, inclement weather, etc etc etc). They even specifically compare the conditions and "value" between center San Francisco and the Phoenix suburbs, which are wildly different. In fact, you can even find a handy table where they list the frequency of encountering these obstacles in Phoenix vs. San Francisco if you're into the whole data thing instead of just trying to make statements that you don't know are true. Carefully selected miles in optimal conditions? You made that claim, you really want to try to back it up now? Go see how many cities Uber operates these vehicles in. Really, expecting someone to do the absolute bare minimum of research before they try to act like an authority shouldn't be asking too much.
About the fatality, that happened at 10pm at an intersection with a theater which lets out a ton of people when the show ends, with a lot of traffic turning there and people walking across the street to the garage, with a bus stop and light rail stop next to it also. Yeah, that's the optimal condition in which to do a driving test, that's totally the reason they chose that time and location.
I asked an honest question if they are looking ahead at intersections.
1) it sounds like you were making an assumption; 2) why are you asking that question on Slashdot? Is it rhetorical, and you're implying the answer is no and that helps your argument, or are you honestly looking for a response from an automatic driving engineer at Uber? Because so far you've just been speculating. I guess you're going to go with the "I'm just asking questions" excuse when I start challenging your assumptions.
If they are gathering information about what is happening down the road then all my comments are moot.
I'm glad we agree that all of your comments are moot. Unless you honestly believe that they are not gathering information about what is happening down the road and that the sensors are there just for people to look at.
This researcher you're talking about, is that the same developer who exposed this? Yeah, let's fuck that guy over all we possibly can, that will definitely be a win for privacy.
It's kind of telling when your company needs to do a "damage control session" because the public finally figured out what your business model is.
Here's a hint: if there's a company with a market cap of almost $500 billion, and you don't know what their product is, you're probably the product. Cambridge Analytica is the customer, they buy you.
I would be more interested in hearing about what data they actually got and what they paid for it, I want to know more about that market value.
Despite the number of human accidents, the fact is that there are millions of cars on the road being driving by humans NOT getting in accidents and they are doing something right.
Waymo and Uber combined have over 5 million autonomous miles driven on public roads since 2009 (that includes only 1 city in which Uber operates, I don't know the numbers for all of the rest; we could be talking about 10 million miles). This has resulted in 1 pedestrian death. There is no reality where this means that autonomous driving is more unsafe than human driving.
If an automated car is not gathering all information about an intersection from the visibility point on then AI is failing us, and will continue to kill people.
What information do you have about the information gathered by the vehicles? Are you just making assumptions about what they are or are not doing, or do you have actual insight? Do you think you have more insight than the people actively working on the problems, or are you just going to second-guess everything that you imagine they're working on?
Furthermore, people like you will continue to shrug their shoulders and wonder why.
Oh, is that what I'm doing, just wondering why? I've been to that intersection many times, I bet I know exactly what happened. I think that a rational response would be to study the interaction, the inputs, and the decisions by the computer and determine 1) if a human would be expected to respond any differently, and if so, adjust the system; 2) if the system failed to react appropriately. Yanking all autonomous vehicles off all roads is not a rational response, it's a knee-jerk panic response by people who don't understand what happened in the first place. If a plane crashes we don't ground all flying vehicles until we figure out why that one crashed.
People literally have the same problem, as attested to the over 5,000 pedestrian deaths due to crashes in the US in 2015. So, we can conclude that human drivers are not ready.
OK, in 2015 5,376 pedestrians were killed by cars in the US, which averages to about one death per 1.6 hours. For autonomous cars, Waymo alone (this doesn't include Uber or Lyft) has driven over 4 million miles on public roads. I'm not real sure how to translate that to hours, but it took them about 6 months to do the last million miles. Uber did 1 million miles across 30,000 trips in 1 year in "Steel City" (Pittsburgh maybe?). So autonomous cars have driven over 5 million miles at a minimum, and we have 1 pedestrian death. My quick math shows me that 1 death over 5 million miles is somewhat less than 1 death per 1.6 hours.
What "the girl" are you talking about, the one in the video I linked to above? The one who actually got hit by a person, who did not might well have seen her? Or one of the other hundreds or thousands of people who get hit by humans every day? I hope you're not talking about the one who got squished next to Marquee Theater by Uber, because when a show lets out there, there are people running all over the street (right next to a bus stop, 4 or 5 story parking garage, and light rail stop - AKA, obstacles).
If an AI is not looking that far ahead, it might end up with no time to stop.
Well, if my aunt had balls, she might end up being my uncle. All of the speculation is a fun exercise and all, but let's go ahead and assume that the people who have been working on autonomous driving for the last decade haven't been sitting around doing nothing all day.
Also, I like how you're assuming the best for humans, who "might well have" seen someone 300 feet in front of them and they're getting ready to respond to them. Have you seen actual human drivers? Often they have no idea where they're heading or where they're supposed to go, much less anything going on around them. But then, just like you assume the best for humans, you assume the worst for AI, like the AI is sitting there texting its friend instead of watching the road. Come back to reality. Yes, in the reality you're describing where humans are always aware of everything within 100 yards, and AI just straight up doesn't see a person right in front of them, it's safer to keep AI off the road. That is not the reality in which we live though.
Frankly, it doesn't need to look any farther than the distance required to stop. And, again, adding sensors to a vehicle does not magically affect the friction coefficient on the tires and give the car braking abilities that defy physics.
The recent introduction which caused people to be hit by cars? I'm not sure a recent introduction is to blame for that, that's been happening since there have been cars. What we are reading about here is the first time, in history as far as I know, that a vehicle driving autonomously has hit and killed a person. Now, look up how many people died on Saturday in collisions between pedestrians and human drivers. I'll wait.
If the sensors were not adequate for the conditions, I would expect a criminal vehicular homicide case against Uber.
So then, if Uber does not get convicted in a criminal vehicular homicide case then we can conclude that you have no idea what you're talking about.
Seriously, why are you going on and on like you have data which showed that the sensors just straight up failed to see this person? Why have you latched onto that as being definitely 100% the reason this happened?
If the sensors are blind to humans crossing the road
If that were true, the vehicles would have never been on the road at all. Sensors that aren't blind are kind of a pre-requisite, so you must be talking out of your ass.
Right, and people on bikes should stop at all stop signs and red lights and otherwise follow the rules of the road. Imaginary land is fun. And there is no evidence that the car did not react appropriately. Wait for the video before deciding who's at fault.
Don't be an idiot, it's not even close to the level of drunk driving. The system has a reaction time far better than people, and reaction time is one of the major factors of drunk driving. It's not even close. Look at stuff like this, and imagine if that white BMW was a bus or large truck that no sensor could see through. How is it going to avoid hitting her, should it just ram cars next to it and hope that slows it enough? What about the passengers in those other cars? I mean at some point you need to put accountability where it belongs, and when people run out into the street without looking and get their asses hit, they are accountable for that. It does not matter what the nature of the vehicle is which actually hit them, whether it is a Soviet Lada, or a motorcycle, or a bicyclist, or a Ford F-150, or a Volvo with sensors all over it. The person who got hit is still at fault. And if you look at that video and think "well, if that car had magic sensors then that must reduce the braking distance from 120 feet to 10 feet" then you're an idiot.
Autonomous cars still have to follow the laws of physics, and if the car is going the speed limit at 40mph, on a street with no room to swerve, and some idiot who isn't paying attention sprints out into the roadway 15 feet in front of that car, tell me math major, how long does it take the car to get to the person? What is the minimum stopping distance of a 3,000 pound car traveling at 40mph with good tires and a dry road surface? Is it 15 feet? Do autonomous cars get to ignore physics? If they don't, and that person gets hit once the car travels the 15 feet to get to them, whose fault is it?
OK. And maybe the people who insist that autonomous cars should never, ever, ever have even the smallest possibility to ever hit a person, ever, should likewise come back to reality.
If this person has made the same maneuver many times before in front of human cars and did not die then what she did this time is not dangerous and stupid... then this car screwed up with a factor of driving and that is all there is to it.
That's a lot of "ifs" and conclusions when you have zero information about what actually happened. What if the video gets released and this person jumped out 3 feet in front of the car from behind a bus? What's "all there is to it" in that case? Why even bother with posts like this when you have zero information beyond the fact that a car hit a person?
Isn't that one way that the system learns? Are you suggesting that we remove all autonomous vehicles, and have everyone still subject to the failures of human drivers, until we can guarantee with 100% certainty that an autonomous vehicle will never hit a person in any situation? What if, in order to avoid a collision with a human, the car rams the vehicle next to it off the road, is that really a better outcome? What if hitting one person is in fact the lesser of evils?
It's the person's fault for not being detected by a sensor?
Not for not being detected, they probably were detected, but detected too late. If someone sprints from behind an obstacle, like a parked truck or stopped bus, onto a roadway without looking for traffic, then yes, regardless of the nature of the vehicle that hits them, it's their fault. There are loads of videos online where you see exactly this thing happen, where the vehicle had no chance to stop or even react.
You can't pay taxes legally if you're not a legal resident.
That terminology isn't correct. My wife had a social security number and paid her taxes while she was living here on a non-resident visa (H-1B). So, she was not a legal resident. Resident is a legal term.
That's exactly right, in fact just driving to work this morning I heard the daily reminder that everyone should mistreat all undocumented workers. There are a lot of absolutes like that which always apply to 100% of people, for example all anonymous posters are idiots.
Why would the memo say anything? Consider the source. This is a summary of evidence, authored by Republicans. They voted to release their own summary, and they also voted not to release a Democrat summary of the same evidence, and they voted not to release the actual evidence. These votes were on party lines, the Republicans outnumbered the Democrats so things the Republicans wanted to be released got released and things the Democrats wanted to be released did not get released. So, if you want to answer the question on why the memo says anything, the answer is because that's what Republicans want people to hear. It doesn't mean it's true. In fact, the FBI has said the memo is not accurate. But we don't know what the evidence is and we don't know how the Democrats saw it, we only have one point of view.
If you want to be an unsympathetic dick head and personally attack me and everyone else in chronic pain you better be fucking prepared to be called out on it instead of being a whiny little bitch because you can't stand a hard discussion about facts.
Easy there pal, I'm not the dude fucking your wife. I'm just suggesting that surgery is generally the long-term answer for chronic pain. I'm not aware of that opinion being some kind of unsympathetic personal attack on everyone in chronic pain omg, it's just a simple statement that for a lot of people in chronic pain, a surgical procedure is probably better than a lifetime of opiates. Not exactly a fringe opinion or anything to get your panties in a twist about. And, if you think this is me being whiny, well I don't know what to tell you.
About the rest of your post - you're obviously not interested in a discussion. You're super pissed about something that has nothing to do with me and I guess this is your outlet, so hope you feel better now. If not, maybe a couple dozen pills per day will fix that problem of yours.
The solution to a drug problem isn't to attack the only legitimate source of pain relief for hundreds of patients and get it shut down.
We're in total agreement on that. Maybe you don't understand me or my position. Regardless, glad you got your little rant out.
I can parse English just fine. Whoever wrote that sentence has trouble constructing it though. Your example is equally stupid. "interesting that..." makes perfect sense, "hurt that..." does not, unless it's followed by a noun, as in "it must hurt that little brain of yours to think".
How about "It makes by brain hurt because you cannot parse English", or even more correctly, "Your inability to parse English makes my brain hurt." Of course, why that would make your brain hurt is your own problem.
The phrase "it makes their brains hurt that they like being traitors" is just stupid. Does their brain hurt because they like being traitors? Does being a traitor make their brain hurt? Does the author have any fucking clue what they're even trying to say? Who the fuck knows?
Slashdot Mirror
A human driver generally would have had the intuition the the walker would or could have walked into the path.
Yeah? Is that why 5,000+ pedestrians were killed by cars in the US in 2015, while this is the first time a pedestrian was struck and killed by an autonomous vehicle since they started on roads in 2009?
Can the AI spot a car traveling at 50 MPH approaching a 4 way stop and not slowing down?
Yes, any reasonably good AI is always scanning the sides to see what might be coming out of driveways or the side of the road.
10 million *carefully selected* miles with optimal conditions
Seriously, do at least the minimum research. You'll find out, for example, that the reason they test in San Francisco is because of the very high density and the number of obstacles they run into per mile (construction zones, vehicles improperly parked on narrow streets, streetlights without power, jaywalkers, cyclists being cyclists, emergency vehicles, inclement weather, etc etc etc). They even specifically compare the conditions and "value" between center San Francisco and the Phoenix suburbs, which are wildly different. In fact, you can even find a handy table where they list the frequency of encountering these obstacles in Phoenix vs. San Francisco if you're into the whole data thing instead of just trying to make statements that you don't know are true. Carefully selected miles in optimal conditions? You made that claim, you really want to try to back it up now? Go see how many cities Uber operates these vehicles in. Really, expecting someone to do the absolute bare minimum of research before they try to act like an authority shouldn't be asking too much.
About the fatality, that happened at 10pm at an intersection with a theater which lets out a ton of people when the show ends, with a lot of traffic turning there and people walking across the street to the garage, with a bus stop and light rail stop next to it also. Yeah, that's the optimal condition in which to do a driving test, that's totally the reason they chose that time and location.
I asked an honest question if they are looking ahead at intersections.
1) it sounds like you were making an assumption; 2) why are you asking that question on Slashdot? Is it rhetorical, and you're implying the answer is no and that helps your argument, or are you honestly looking for a response from an automatic driving engineer at Uber? Because so far you've just been speculating. I guess you're going to go with the "I'm just asking questions" excuse when I start challenging your assumptions.
If they are gathering information about what is happening down the road then all my comments are moot.
I'm glad we agree that all of your comments are moot. Unless you honestly believe that they are not gathering information about what is happening down the road and that the sensors are there just for people to look at.
This researcher you're talking about, is that the same developer who exposed this? Yeah, let's fuck that guy over all we possibly can, that will definitely be a win for privacy.
It's kind of telling when your company needs to do a "damage control session" because the public finally figured out what your business model is.
Here's a hint: if there's a company with a market cap of almost $500 billion, and you don't know what their product is, you're probably the product. Cambridge Analytica is the customer, they buy you.
I would be more interested in hearing about what data they actually got and what they paid for it, I want to know more about that market value.
Despite the number of human accidents, the fact is that there are millions of cars on the road being driving by humans NOT getting in accidents and they are doing something right.
Waymo and Uber combined have over 5 million autonomous miles driven on public roads since 2009 (that includes only 1 city in which Uber operates, I don't know the numbers for all of the rest; we could be talking about 10 million miles). This has resulted in 1 pedestrian death. There is no reality where this means that autonomous driving is more unsafe than human driving.
If an automated car is not gathering all information about an intersection from the visibility point on then AI is failing us, and will continue to kill people.
What information do you have about the information gathered by the vehicles? Are you just making assumptions about what they are or are not doing, or do you have actual insight? Do you think you have more insight than the people actively working on the problems, or are you just going to second-guess everything that you imagine they're working on?
Furthermore, people like you will continue to shrug their shoulders and wonder why.
Oh, is that what I'm doing, just wondering why? I've been to that intersection many times, I bet I know exactly what happened. I think that a rational response would be to study the interaction, the inputs, and the decisions by the computer and determine 1) if a human would be expected to respond any differently, and if so, adjust the system; 2) if the system failed to react appropriately. Yanking all autonomous vehicles off all roads is not a rational response, it's a knee-jerk panic response by people who don't understand what happened in the first place. If a plane crashes we don't ground all flying vehicles until we figure out why that one crashed.
People literally have the same problem, as attested to the over 5,000 pedestrian deaths due to crashes in the US in 2015. So, we can conclude that human drivers are not ready.
OK, in 2015 5,376 pedestrians were killed by cars in the US, which averages to about one death per 1.6 hours. For autonomous cars, Waymo alone (this doesn't include Uber or Lyft) has driven over 4 million miles on public roads. I'm not real sure how to translate that to hours, but it took them about 6 months to do the last million miles. Uber did 1 million miles across 30,000 trips in 1 year in "Steel City" (Pittsburgh maybe?). So autonomous cars have driven over 5 million miles at a minimum, and we have 1 pedestrian death. My quick math shows me that 1 death over 5 million miles is somewhat less than 1 death per 1.6 hours.
OK so the bad news is that an exception was thrown, but the good news is that I caught it. Wait, no, that's also bad news.
What "the girl" are you talking about, the one in the video I linked to above? The one who actually got hit by a person, who did not might well have seen her? Or one of the other hundreds or thousands of people who get hit by humans every day? I hope you're not talking about the one who got squished next to Marquee Theater by Uber, because when a show lets out there, there are people running all over the street (right next to a bus stop, 4 or 5 story parking garage, and light rail stop - AKA, obstacles).
There's no reason to freak out.
If an AI is not looking that far ahead, it might end up with no time to stop.
Well, if my aunt had balls, she might end up being my uncle. All of the speculation is a fun exercise and all, but let's go ahead and assume that the people who have been working on autonomous driving for the last decade haven't been sitting around doing nothing all day.
Also, I like how you're assuming the best for humans, who "might well have" seen someone 300 feet in front of them and they're getting ready to respond to them. Have you seen actual human drivers? Often they have no idea where they're heading or where they're supposed to go, much less anything going on around them. But then, just like you assume the best for humans, you assume the worst for AI, like the AI is sitting there texting its friend instead of watching the road. Come back to reality. Yes, in the reality you're describing where humans are always aware of everything within 100 yards, and AI just straight up doesn't see a person right in front of them, it's safer to keep AI off the road. That is not the reality in which we live though.
Frankly, it doesn't need to look any farther than the distance required to stop. And, again, adding sensors to a vehicle does not magically affect the friction coefficient on the tires and give the car braking abilities that defy physics.
The recent introduction which caused people to be hit by cars? I'm not sure a recent introduction is to blame for that, that's been happening since there have been cars. What we are reading about here is the first time, in history as far as I know, that a vehicle driving autonomously has hit and killed a person. Now, look up how many people died on Saturday in collisions between pedestrians and human drivers. I'll wait.
If the sensors were not adequate for the conditions, I would expect a criminal vehicular homicide case against Uber.
So then, if Uber does not get convicted in a criminal vehicular homicide case then we can conclude that you have no idea what you're talking about.
Seriously, why are you going on and on like you have data which showed that the sensors just straight up failed to see this person? Why have you latched onto that as being definitely 100% the reason this happened?
If the sensors are blind to humans crossing the road
If that were true, the vehicles would have never been on the road at all. Sensors that aren't blind are kind of a pre-requisite, so you must be talking out of your ass.
Right, and people on bikes should stop at all stop signs and red lights and otherwise follow the rules of the road. Imaginary land is fun. And there is no evidence that the car did not react appropriately. Wait for the video before deciding who's at fault.
Don't be an idiot, it's not even close to the level of drunk driving. The system has a reaction time far better than people, and reaction time is one of the major factors of drunk driving. It's not even close. Look at stuff like this, and imagine if that white BMW was a bus or large truck that no sensor could see through. How is it going to avoid hitting her, should it just ram cars next to it and hope that slows it enough? What about the passengers in those other cars? I mean at some point you need to put accountability where it belongs, and when people run out into the street without looking and get their asses hit, they are accountable for that. It does not matter what the nature of the vehicle is which actually hit them, whether it is a Soviet Lada, or a motorcycle, or a bicyclist, or a Ford F-150, or a Volvo with sensors all over it. The person who got hit is still at fault. And if you look at that video and think "well, if that car had magic sensors then that must reduce the braking distance from 120 feet to 10 feet" then you're an idiot.
Autonomous cars still have to follow the laws of physics, and if the car is going the speed limit at 40mph, on a street with no room to swerve, and some idiot who isn't paying attention sprints out into the roadway 15 feet in front of that car, tell me math major, how long does it take the car to get to the person? What is the minimum stopping distance of a 3,000 pound car traveling at 40mph with good tires and a dry road surface? Is it 15 feet? Do autonomous cars get to ignore physics? If they don't, and that person gets hit once the car travels the 15 feet to get to them, whose fault is it?
OK. And maybe the people who insist that autonomous cars should never, ever, ever have even the smallest possibility to ever hit a person, ever, should likewise come back to reality.
If this person has made the same maneuver many times before in front of human cars and did not die then what she did this time is not dangerous and stupid... then this car screwed up with a factor of driving and that is all there is to it.
That's a lot of "ifs" and conclusions when you have zero information about what actually happened. What if the video gets released and this person jumped out 3 feet in front of the car from behind a bus? What's "all there is to it" in that case? Why even bother with posts like this when you have zero information beyond the fact that a car hit a person?
Isn't that one way that the system learns? Are you suggesting that we remove all autonomous vehicles, and have everyone still subject to the failures of human drivers, until we can guarantee with 100% certainty that an autonomous vehicle will never hit a person in any situation? What if, in order to avoid a collision with a human, the car rams the vehicle next to it off the road, is that really a better outcome? What if hitting one person is in fact the lesser of evils?
It's the person's fault for not being detected by a sensor?
Not for not being detected, they probably were detected, but detected too late. If someone sprints from behind an obstacle, like a parked truck or stopped bus, onto a roadway without looking for traffic, then yes, regardless of the nature of the vehicle that hits them, it's their fault. There are loads of videos online where you see exactly this thing happen, where the vehicle had no chance to stop or even react.
You can't pay taxes legally if you're not a legal resident.
That terminology isn't correct. My wife had a social security number and paid her taxes while she was living here on a non-resident visa (H-1B). So, she was not a legal resident. Resident is a legal term.
That's exactly right, in fact just driving to work this morning I heard the daily reminder that everyone should mistreat all undocumented workers. There are a lot of absolutes like that which always apply to 100% of people, for example all anonymous posters are idiots.
Why would the memo say anything? Consider the source. This is a summary of evidence, authored by Republicans. They voted to release their own summary, and they also voted not to release a Democrat summary of the same evidence, and they voted not to release the actual evidence. These votes were on party lines, the Republicans outnumbered the Democrats so things the Republicans wanted to be released got released and things the Democrats wanted to be released did not get released. So, if you want to answer the question on why the memo says anything, the answer is because that's what Republicans want people to hear. It doesn't mean it's true. In fact, the FBI has said the memo is not accurate. But we don't know what the evidence is and we don't know how the Democrats saw it, we only have one point of view.
If you want to be an unsympathetic dick head and personally attack me and everyone else in chronic pain you better be fucking prepared to be called out on it instead of being a whiny little bitch because you can't stand a hard discussion about facts.
Easy there pal, I'm not the dude fucking your wife. I'm just suggesting that surgery is generally the long-term answer for chronic pain. I'm not aware of that opinion being some kind of unsympathetic personal attack on everyone in chronic pain omg , it's just a simple statement that for a lot of people in chronic pain, a surgical procedure is probably better than a lifetime of opiates. Not exactly a fringe opinion or anything to get your panties in a twist about. And, if you think this is me being whiny, well I don't know what to tell you.
About the rest of your post - you're obviously not interested in a discussion. You're super pissed about something that has nothing to do with me and I guess this is your outlet, so hope you feel better now. If not, maybe a couple dozen pills per day will fix that problem of yours.
The solution to a drug problem isn't to attack the only legitimate source of pain relief for hundreds of patients and get it shut down.
We're in total agreement on that. Maybe you don't understand me or my position. Regardless, glad you got your little rant out.
I can parse English just fine. Whoever wrote that sentence has trouble constructing it though. Your example is equally stupid. "interesting that..." makes perfect sense, "hurt that..." does not, unless it's followed by a noun, as in "it must hurt that little brain of yours to think".
How about "It makes by brain hurt because you cannot parse English", or even more correctly, "Your inability to parse English makes my brain hurt." Of course, why that would make your brain hurt is your own problem.
The phrase "it makes their brains hurt that they like being traitors" is just stupid. Does their brain hurt because they like being traitors? Does being a traitor make their brain hurt? Does the author have any fucking clue what they're even trying to say? Who the fuck knows?