The law says that they are at fault if they hit you from behind.
No law says that.
Oh, and if you haven't heard, the Swoop and Squat fraud has gotten places with previously confusing laws to clarify that the person behind may have a larger burden of proof, but stomping on your brakes with the intent of causing a crash is not just a traffic infringement, but a crime, should they choose to pursue it.
Yes, AI is unpredictable and unknowable to the average driver. The AI follows the rules. The meat-bags around are confused by "signals" and "yielding", and end up steering at the self driving car just to see if it'll manage to avoid them.
None of the cars have ever done that, or even can do that. The human can turn off auto pilot, but the computer can't turn off itself.
Yet, I constantly see the complaints here about these non-existent cars which cause crashes, but turn off before the impact. If they are so bad, why do you have to lie so much to make them look worse?
You'd still compare it to the wrong thing. Crashes per million miles doesn't work when you are comparing a national average to a car used only in a specific city. And so far, everyone I've seen estimating the crashes per million vehicle miles to compare has compared against averages that aren't representative of the conditions the cars are used in.
Only the former. The case of self-driving cars getting into a bad situation then handing control back to the human after a collision was unavoidable so the human could be blamed is only considered a reasonable act by the slashdotters making up strawmen.
And there's solo accidents and chain collisions, so it's not given that there's two parties involved.
So what works out is that some (one-car) crashes are 100% one car's fault, and some chain collisions have smaller shared faults, it would seem that the one-car crashes aren't that rare, so there would be more than a 50% chance of "half" being the fault of the driver in question.
Also, if you consider a tire blowing out to not be driver error, then you don't know how to drive, or how fault is assigned. Most tire failures are from underinflation. That's a maintenance error that is the fault of the maintainer of the car. That's the likely driver, so a blow-out is more likely than not the fault of the driver. Also, a blowout is easily recoverable. It should never result in a crash. Cruising at 70 in a 70 and having a tire explode in a hand grenade-like explosion should result in the driver calmly pulling to the side in a controlled manner. Brakes not used if the blowout was in the front, and freely used if on the back, but not hard, if used.
I've had it happen, and it's startling, but not a problem to anyone who is awake.
Tree across the road is 100% driver fault, unless the tree fell as the driver approached (statistically rounds to "impossible"). If you are driving too fast to stop for a tree, you are doing it wrong.
You have an incorrect understanding of law. The person with the right of way may pass on it. If they wave you on, you now legally have the right of way. If you don't go, or waive someone else on, then it is *you*, not them that is holding people up.
That's not the rule anywhere I've lived. If two cars signal to enter the same lane, the vehicle closest to the median has the right of way. This is not true if the vehicle in the slower lane signaled first and has begun the maneuver. But it's the "in a tie, the faster lane wins" rule.
You can anticipate 99.9% of what happens trivially. How often do you get shot at in the car? When people lead with stupidity like that, it's hard to consider their point. Yes, there will be some unknown things that happen. They will either be ignored (as shooting would be, so long as a critical sensor isn't hit), or met with stopping and waiting for manual-drive (the common "you are driving in a rain storm and the road is washed out ahead of you, what does the car do?" question).
got 90% of the 0.1% of unusual situations, it's perfectly safe to pull off the road and park. Either uploading the parameters to home so a human in the control room can remotely decide the best action, or the human in the car to decide the next action.
Though it's nice that we've moved away from the people who ask "what happens when you are teleported by Harry Potter to a motorway going 150 mph in a 55 mph, while being tailgated by an 18-wheeler, and there's a stopped car 150 feet ahead of you and traffic all around"?" There's no right answer for a human, but if a computer can't solve it, then computers are obviously useless.
what does it do at four-way stop if it was the second car there but the human in the first car is waving it on?
Legally, once waived on, it has the right of way and may go, however, as it didn't establish eye contact with the driver of the other car, it should wait. After 4+k*RND seconds, it proceeds.
has it been tested on drivethru fastfood?
Yes, and it does *not* like fries with that.
if someone is shooting at car, how does the car react?
It dodges each individual bullet, and shoots back, if you have the Thug Life upgrade installed.
does it get up to speed on onramps or does it merge into highway traffic going 45mph?
It tries to match speed, so long as that doesn't interfere with other parameters. Not sure how they are going in CA, with not speeding, and the freeways marked at 55 and cars either going 5 or 85 on them, depending on time of day.
But they don't need to be worked out. As long as you aren't going to cause a problem behind, you stop for it, kid, adult, animal. If you can avoid it with 0 chance of damage to yourself or others, then do so. You don't need to identify it to avoid it.
The problem is the people who think the difference matters.
My "rule" is that if it moves, you can hit it without losing control or damaging yourself. if it's moving or sliding, it's light enough to smash through, whether it's a paper bag, a squirrel, a tumbleweed. Now, this doesn't work when the winds are high. I've seen large, solid objects move into or across traffic lanes by wind, when wind is exceptionally high.
Being able to identify the exact object is a red herring. Just identify whether it's solid enough to be damaging, with enough accuracy to be useful.
I don't really care if I'm not legally responsible for the head on collision I had at 60mph, if it could have been avoided by a lane change. I don't even care if the lane change was improper and illegal,
If the "illegal" lane change would prevent a crash and not cause one, it's not illegal, it's also not a lane change. It's an evasive maneuver, and those are explicitly legal everywhere I've ever bothered to read the traffic laws.
(turning right on red, while legal, is technically running the red)
It's technically not "running the red" which implies the illegal act of running a red light. You may legally drive through a red light for a limited number of reasons (depends on jurisdiction). That's not "running the red". That's "obeying the traffic signal."
And in my experience, they find the person 51% at fault and call it 100% because it's quicker and easier. Only at a true 50/50 do they bother to put anything down other than 100/0. But again, I'm sure that varies by jurisdiction, and applies to initial determinations only, if you spend $1,000,000 to debate the matter in court, I'm sure that the split will be more accurate, but the cops who write the reports that are used by insurance without question generally stick to naming one party to be fully at fault.
I agree with the math and statistics, but I think the bad assumption is too provably wrong to make the statistics valid.
If you are testing a self driving car, would you set it on a test track to drive circles for 1,000,000 miles?
No, that'd be boring and not give useful data. Same reason they aren't driving these from LA to NYC every week. Instead, they keep them on city streets, in the worst crash rate areas, seeing how they perform. It's reasonable to compare them against the regular crash rate, but take the rates for the streets they are on, not the average million-mile rates that are heavily skewed by low-crash divided highways.
If you are right, then you are wrong. The sample is "trained professional drivers driving a prototype". The propensity of a self-driving car to influence driver behavior should be examined, but not when it's in trials with professional drivers. The data wouldn't be of any use in that case anyway.
The 6000lb pickups aren't "passenger cars". When your definition is wrong, then your observations will never match reality. So, do you change your definitions, or assert reality is wrong?
When I was a kid, miniaturization was a big thing. There were some tiny piston engines, and miniature turbines. Well, at least as small as they can make it. The electric ones were more limited in the '80s because the batteries were a bit heavier than now. So people were pushing the envelope for IC engines in RC then.
We have an instinctive ability to do that math, an intuitive knack for maths and physics which is quite powerful. If we could apply it to school maths hardly anybody would ever fail.
Nope. It's not math. It's about feedback. You "guess" where it's going to land based on previous experience, and adjust as you see it coming. Try playing volleyball with a strobe light. You can see it, watch the path, but will have great trouble hitting it. Or take an adult with a math degree who hasn't played baseball. Hit a long fly ball towards them. They'll run in, as they'll feel like it'll fall short, but it will fly well over their heads.
Math is unrelated to catching. Catching is a feedback loop, combined with practice.
With what we know about how human brains intuitively work with basic physics
Nope. People don't have any intuitive knowledge of physics.
So unless there is some sort of relationship between the average number of Buzzfeed posts in a given week to seismic activity in Brunei you can't accurately predict earthquakes from our very limited information. So far this hasn't been demonstrated.
Other than things like the article we are discussing now. Where some manner of activity resulted in predictions that were close. Full details aren't available to verify if they are lucky charlatans, or good scientists.
Part of the problem is also language. It is trespassing, but may not be an illegal trespass. You don't need to have "no trespassing" posted. Ever. And in most cases, the sign carries no legal weight. If someone climbs a fence or locked gate to access the property, it's likely illegal trespass (without sign). "No trespass" "posted" and "private property" are all the same thing, and are legally interchangeable, at least in most cases.
If you are opening a closed (but not locked) gate, or wandering on unfenced property, it's hard to commit a trespass.
Also note the wording of your example sign. "No Trespassing" That implies that being there without the sign is trespassing, but not illegal.
You don't have to understand it to predict it. I've had arguments on Slashdot that dogs had to understand physics to catch a ball. Nope, they don't know how to calculate where it's going to land. They just figure it was about that hard, so it'll go about that far, and run and keep looking. You can take information and predict a result without understanding the mechanism of the result.
Slashdot Mirror
The law says that they are at fault if they hit you from behind.
No law says that.
Oh, and if you haven't heard, the Swoop and Squat fraud has gotten places with previously confusing laws to clarify that the person behind may have a larger burden of proof, but stomping on your brakes with the intent of causing a crash is not just a traffic infringement, but a crime, should they choose to pursue it.
Yes, AI is unpredictable and unknowable to the average driver. The AI follows the rules. The meat-bags around are confused by "signals" and "yielding", and end up steering at the self driving car just to see if it'll manage to avoid them.
None of the cars have ever done that, or even can do that. The human can turn off auto pilot, but the computer can't turn off itself.
Yet, I constantly see the complaints here about these non-existent cars which cause crashes, but turn off before the impact. If they are so bad, why do you have to lie so much to make them look worse?
You'd still compare it to the wrong thing. Crashes per million miles doesn't work when you are comparing a national average to a car used only in a specific city. And so far, everyone I've seen estimating the crashes per million vehicle miles to compare has compared against averages that aren't representative of the conditions the cars are used in.
Only the former. The case of self-driving cars getting into a bad situation then handing control back to the human after a collision was unavoidable so the human could be blamed is only considered a reasonable act by the slashdotters making up strawmen.
And there's solo accidents and chain collisions, so it's not given that there's two parties involved.
So what works out is that some (one-car) crashes are 100% one car's fault, and some chain collisions have smaller shared faults, it would seem that the one-car crashes aren't that rare, so there would be more than a 50% chance of "half" being the fault of the driver in question.
Also, if you consider a tire blowing out to not be driver error, then you don't know how to drive, or how fault is assigned. Most tire failures are from underinflation. That's a maintenance error that is the fault of the maintainer of the car. That's the likely driver, so a blow-out is more likely than not the fault of the driver. Also, a blowout is easily recoverable. It should never result in a crash. Cruising at 70 in a 70 and having a tire explode in a hand grenade-like explosion should result in the driver calmly pulling to the side in a controlled manner. Brakes not used if the blowout was in the front, and freely used if on the back, but not hard, if used.
I've had it happen, and it's startling, but not a problem to anyone who is awake.
Tree across the road is 100% driver fault, unless the tree fell as the driver approached (statistically rounds to "impossible"). If you are driving too fast to stop for a tree, you are doing it wrong.
You have an incorrect understanding of law. The person with the right of way may pass on it. If they wave you on, you now legally have the right of way. If you don't go, or waive someone else on, then it is *you*, not them that is holding people up.
That's not the rule anywhere I've lived. If two cars signal to enter the same lane, the vehicle closest to the median has the right of way. This is not true if the vehicle in the slower lane signaled first and has begun the maneuver. But it's the "in a tie, the faster lane wins" rule.
You can anticipate 99.9% of what happens trivially. How often do you get shot at in the car? When people lead with stupidity like that, it's hard to consider their point. Yes, there will be some unknown things that happen. They will either be ignored (as shooting would be, so long as a critical sensor isn't hit), or met with stopping and waiting for manual-drive (the common "you are driving in a rain storm and the road is washed out ahead of you, what does the car do?" question).
got 90% of the 0.1% of unusual situations, it's perfectly safe to pull off the road and park. Either uploading the parameters to home so a human in the control room can remotely decide the best action, or the human in the car to decide the next action.
Though it's nice that we've moved away from the people who ask "what happens when you are teleported by Harry Potter to a motorway going 150 mph in a 55 mph, while being tailgated by an 18-wheeler, and there's a stopped car 150 feet ahead of you and traffic all around"?" There's no right answer for a human, but if a computer can't solve it, then computers are obviously useless.
what does it do at four-way stop if it was the second car there but the human in the first car is waving it on?
Legally, once waived on, it has the right of way and may go, however, as it didn't establish eye contact with the driver of the other car, it should wait. After 4+k*RND seconds, it proceeds.
has it been tested on drivethru fastfood?
Yes, and it does *not* like fries with that.
if someone is shooting at car, how does the car react?
It dodges each individual bullet, and shoots back, if you have the Thug Life upgrade installed.
does it get up to speed on onramps or does it merge into highway traffic going 45mph?
It tries to match speed, so long as that doesn't interfere with other parameters. Not sure how they are going in CA, with not speeding, and the freeways marked at 55 and cars either going 5 or 85 on them, depending on time of day.
But they don't need to be worked out. As long as you aren't going to cause a problem behind, you stop for it, kid, adult, animal. If you can avoid it with 0 chance of damage to yourself or others, then do so. You don't need to identify it to avoid it.
The problem is the people who think the difference matters.
My "rule" is that if it moves, you can hit it without losing control or damaging yourself. if it's moving or sliding, it's light enough to smash through, whether it's a paper bag, a squirrel, a tumbleweed. Now, this doesn't work when the winds are high. I've seen large, solid objects move into or across traffic lanes by wind, when wind is exceptionally high.
Being able to identify the exact object is a red herring. Just identify whether it's solid enough to be damaging, with enough accuracy to be useful.
I don't really care if I'm not legally responsible for the head on collision I had at 60mph, if it could have been avoided by a lane change. I don't even care if the lane change was improper and illegal,
If the "illegal" lane change would prevent a crash and not cause one, it's not illegal, it's also not a lane change. It's an evasive maneuver, and those are explicitly legal everywhere I've ever bothered to read the traffic laws.
(turning right on red, while legal, is technically running the red)
It's technically not "running the red" which implies the illegal act of running a red light. You may legally drive through a red light for a limited number of reasons (depends on jurisdiction). That's not "running the red". That's "obeying the traffic signal."
And in my experience, they find the person 51% at fault and call it 100% because it's quicker and easier. Only at a true 50/50 do they bother to put anything down other than 100/0. But again, I'm sure that varies by jurisdiction, and applies to initial determinations only, if you spend $1,000,000 to debate the matter in court, I'm sure that the split will be more accurate, but the cops who write the reports that are used by insurance without question generally stick to naming one party to be fully at fault.
How is that different from today? I see that behavior regularly.
The reason is that when I am in an accedent, my first priority is not if I wasd right, but if I was safe.
There have been zero injury accidents involving the sum of all self driving cars on the planet. That sounds pretty "safe" so far.
I agree with the math and statistics, but I think the bad assumption is too provably wrong to make the statistics valid.
If you are testing a self driving car, would you set it on a test track to drive circles for 1,000,000 miles?
No, that'd be boring and not give useful data. Same reason they aren't driving these from LA to NYC every week. Instead, they keep them on city streets, in the worst crash rate areas, seeing how they perform. It's reasonable to compare them against the regular crash rate, but take the rates for the streets they are on, not the average million-mile rates that are heavily skewed by low-crash divided highways.
If you are right, then you are wrong. The sample is "trained professional drivers driving a prototype". The propensity of a self-driving car to influence driver behavior should be examined, but not when it's in trials with professional drivers. The data wouldn't be of any use in that case anyway.
The 6000lb pickups aren't "passenger cars". When your definition is wrong, then your observations will never match reality. So, do you change your definitions, or assert reality is wrong?
And someone operating it outside their ToS would be using it illegally, thus reducing the liability for the corporate overlords.
When I was a kid, miniaturization was a big thing. There were some tiny piston engines, and miniature turbines. Well, at least as small as they can make it. The electric ones were more limited in the '80s because the batteries were a bit heavier than now. So people were pushing the envelope for IC engines in RC then.
We have an instinctive ability to do that math, an intuitive knack for maths and physics which is quite powerful. If we could apply it to school maths hardly anybody would ever fail.
Nope. It's not math. It's about feedback. You "guess" where it's going to land based on previous experience, and adjust as you see it coming. Try playing volleyball with a strobe light. You can see it, watch the path, but will have great trouble hitting it. Or take an adult with a math degree who hasn't played baseball. Hit a long fly ball towards them. They'll run in, as they'll feel like it'll fall short, but it will fly well over their heads.
Math is unrelated to catching. Catching is a feedback loop, combined with practice.
With what we know about how human brains intuitively work with basic physics
Nope. People don't have any intuitive knowledge of physics.
So unless there is some sort of relationship between the average number of Buzzfeed posts in a given week to seismic activity in Brunei you can't accurately predict earthquakes from our very limited information. So far this hasn't been demonstrated.
Other than things like the article we are discussing now. Where some manner of activity resulted in predictions that were close. Full details aren't available to verify if they are lucky charlatans, or good scientists.
Part of the problem is also language. It is trespassing, but may not be an illegal trespass. You don't need to have "no trespassing" posted. Ever. And in most cases, the sign carries no legal weight. If someone climbs a fence or locked gate to access the property, it's likely illegal trespass (without sign). "No trespass" "posted" and "private property" are all the same thing, and are legally interchangeable, at least in most cases.
If you are opening a closed (but not locked) gate, or wandering on unfenced property, it's hard to commit a trespass.
Also note the wording of your example sign. "No Trespassing" That implies that being there without the sign is trespassing, but not illegal.
You don't have to understand it to predict it. I've had arguments on Slashdot that dogs had to understand physics to catch a ball. Nope, they don't know how to calculate where it's going to land. They just figure it was about that hard, so it'll go about that far, and run and keep looking. You can take information and predict a result without understanding the mechanism of the result.