8 ms·

I think more impressively, even in this seemingly worst case scenario where everything lined up as poorly as it could have, the car was still able to react quit

I think more impressively, even in this seemingly worst case scenario where everything lined up as poorly as it could have, the car was still able to react quite aggressively and the situation ended only with minor scratches it seems? Honestly, if anything, this is probably proof that self-driving is the future, as I would bet any human would've most definitely done much worse in the same scenario.

gnicholas 3y ago
I wonder if a human driver would have (for better or worse) swerved to avoid hitting the cyclist. That is, the Waymo might be programmed to only execute certain types of evasive maneuvers, whereas a human would instinctively do other things as well. Of course, if there were someone in the vehicle's blind spot, it's possible this would have resulted in a worse accident, since there would have been no time for a human driver to double check.
- petre 3y ago
  In a recent accident that happened in my country, a human driver swerved to avoid a cyclist that was quickly switching lanes without signaling in order to cross from one side to another. The swerving car went over the sidewalk and killed a pedestrian. You should draw your own conclusions.
  - viraptor 3y ago
    This is my pet peeve with people being so sure online about their abilities of avoiding things on the road. Yeah, maybe you could avoid it with your super human reflexes, but there physically no way you could check that side of your car, register what's there, decide whether you can move that way and then did it. The kneejerk avoidance reaction may be extremely dangerous on its own. (Also in Oz, rolling your car when you try to swerve to avoid a kangaroo is so common that you're normally taught to just break instead)
    - panopticon 3y ago
      I swerved into another lane to avoid another car cutting me off during rush hour once. It all happened so fast, and I will always remember how incredibly lucky I was that no one was in the lane I moved into.
    - cgriswald 3y ago
      Do people really not mentally track the vehicles around them? I remember in driver’s ed even before we were behind the wheel to have “exit strategies” for situations exactly like this. Cars don’t just appear magically next to you.
      - viraptor 3y ago
        > Cars don’t just appear magically next to you. They don't, but: 1. Your perception is not perfect. 2. A truck passing you means there was a blind spot and you have no idea what was there. 3. There are areas with many lanes merging / splitting, where it takes most of your attention to know where you're going; little left to track everything around. Tracking your surrounding is great. Accepting that you're going to make mistakes, adjusting for that and also allowing assistive tech is even better.
      - gnicholas 3y ago
        People try to, but are imperfect at doing so. For example, most people do not closely track what is happening 2 lanes away. But someone can move from your blind spot 2 lanes away into your blind spot right next to you. Also, if something potentially dangerous/important is happening to your left, you will focus your attention there and not be aware of things happening to your right, even if you normally would have seen those things had you been looking straight ahead. I try to know what is going on around me, for the reason you describe (exit strategies), but I'm aware that sometimes I simply do not know what is to my left — and would not want to swerve there. Having seen friends driving Teslas, I can see how the surround-view could be useful for situations like this. However, I have also seen that cars appear/vanish for no reason, and it is clearly not reliable at this point.
- shadowgovt 3y ago
  Broadly speaking (and this is a very broad brush, details may differ), autonomous vehicles will resolve collision avoidance by doing something legal. Swerving is not legal and it's not legal because you might hit something you can't see. On the other hand, an autonomous vehicle maintains perpetual 360° situational awareness so it already knows whether it's safe to swerve.
  - lozenge 3y ago
    Can you provide a source that swerving to prevent a collision is not legal?
  - gnicholas 3y ago
    > Swerving is not legal and it's not legal because you might hit something you can't see. If I'm driving down the road and someone pushes a kid in front of my car, I'm very likely to swerve, even if I'm not 100% sure if there is someone next to me. If I hit a car next to me and cause damage/injure people, there is a good chance that a jury would find in my favor because a "reasonable person" (the relevant legal standard in a civil case) would seek to avoid the certain death of a faultless child, even if it means a chance of injury to passengers in a vehicle that might be nearby. I wouldn't swerve if I thought a bicyclist were there, but I would likely swerve into a vehicle, which affords occupants significant protection. And I certainly would not be found guilty of a crime ("beyond a reasonable doubt", the standard in a criminal trial) because it is not criminally negligent to swerve when faced with the certainty of killing a child. Under the "defense of self/others" doctrine, you can do many things that would normally be illegal, if you are doing them to prevent the imminent death of yourself or someone else. (These analyses are not super detailed, but IAAL.)
    - shadowgovt 3y ago
      AVs aren't doing near enough calculus to make a decision like that; they have obstacles and they avoid them. Their avoidance can allow for lane departure (that 360 situational awareness is handy), but they aren't going to solve a trolley problem in realtime if boxed in. The thing is... Neither are people. A kid darts out on front of you and you won't have time to make a conscious decision. The maneuver you make will be too fast for rational thought, and your brain will explain to itself why it did that later. ... Incidentally, this is one of the things that scares me about human drivers. Put a human on the stand to testify and most will say they never saw that kid, no matter what physical process happened to their eyes and neurons. I think most even actually believe it. In any case, I'm glad that in this incident they'll be able to do a full NHTSA workup, pull logs, and reprogram the machine to avoid an accident like this in the future. Can't do that to a human driver.
zx8080 3y ago
Thare's a, probably not very obvious, situation when human could have no such accident in the same scenario. At all. Preventing an accident does not leave any trace. No one knows about the scenario if there's no accident. How many similar (and others) accidents are being prevented by human drivers daily, and no one knows about them, at all? We just don't know. That's why I take any stats of "human vs AI" driving with a grain of salt.
- CityOfThrowaway 3y ago
  We can compare this statistically by looking at accidents per 1,000,000 miles driven of autonomous cars vs humans. Over long mileage, the rate of prevented accidents shows up in the data as a lower overall accident rate. I don't have the stats at hand but Waymo does great on this stat. Now, that's not to say that Waymo would avoid any specific accident better than a human. But from a public health standpoint, that's not really the right way to think about it anyways.
  - talldatethrow 3y ago
    IMO these are a slightly fake stat. You're comparing to overall stats, that include the drunk, the high, the sleep deprived, and the crazy wild aggressive wreckless. If you compare to my mother driving, the injuries per mile are higher in autonomous. You can argue that I'm making an unfair distinction since those people exist, but I'd say the people driving under those conditions are commiting a crime and shouldn't be counted the same.
    - resolutebat 3y ago
      You're making a classic is-ought fallacy here. Drunk, high, sleep deprived or "crazy wild aggressive reckless" human drivers ought not to exist, but they do, and thus it's absolutely fair to compared autonomous drivers against them.
      - eurleif 3y ago
        If GP's mother switches from driving for herself to having Waymo drive for her, her risk of injury will increase. Therefore, if her goal is to avoid injury, it would be counterproductive for her to make that switch. That conclusion does not require any is-ought fallacy.
qwertox 3y ago
Yet it makes me wonder how the car would react if it came to a stop on top of the leg of the cyclist or in some situation where a normal driver would notice that he should get into the car again and back up a couple of feet. I'm not against Waymo.
- Mageek 3y ago
  My mom was hit by a driver when she was biking and fell with her arm in front of the wheel. The driver then decided to pull forward and drove over her arm. So humans don’t really solve that problem.
smcl 3y ago
Getting big "this is actually good for Bitcoin!" vibes from this. I don't think we can fairly assess who was at fault, given that none of us were there. But it is insane to just assume a manned vehicle would have fared worse, and declare it a victory for driverless cars.

gnicholas 3y ago

I wonder if a human driver would have (for better or worse) swerved to avoid hitting the cyclist. That is, the Waymo might be programmed to only execute certain types of evasive maneuvers, whereas a human would instinctively do other things as well. Of course, if there were someone in the vehicle's blind spot, it's possible this would have resulted in a worse accident, since there would have been no time for a human driver to double check.

petre 3y ago
In a recent accident that happened in my country, a human driver swerved to avoid a cyclist that was quickly switching lanes without signaling in order to cross from one side to another. The swerving car went over the sidewalk and killed a pedestrian. You should draw your own conclusions.
- viraptor 3y ago
  This is my pet peeve with people being so sure online about their abilities of avoiding things on the road. Yeah, maybe you could avoid it with your super human reflexes, but there physically no way you could check that side of your car, register what's there, decide whether you can move that way and then did it. The kneejerk avoidance reaction may be extremely dangerous on its own. (Also in Oz, rolling your car when you try to swerve to avoid a kangaroo is so common that you're normally taught to just break instead)
  - panopticon 3y ago
    I swerved into another lane to avoid another car cutting me off during rush hour once. It all happened so fast, and I will always remember how incredibly lucky I was that no one was in the lane I moved into.
  - cgriswald 3y ago
    Do people really not mentally track the vehicles around them? I remember in driver’s ed even before we were behind the wheel to have “exit strategies” for situations exactly like this. Cars don’t just appear magically next to you.
    - viraptor 3y ago
      > Cars don’t just appear magically next to you. They don't, but: 1. Your perception is not perfect. 2. A truck passing you means there was a blind spot and you have no idea what was there. 3. There are areas with many lanes merging / splitting, where it takes most of your attention to know where you're going; little left to track everything around. Tracking your surrounding is great. Accepting that you're going to make mistakes, adjusting for that and also allowing assistive tech is even better.
    - gnicholas 3y ago
      People try to, but are imperfect at doing so. For example, most people do not closely track what is happening 2 lanes away. But someone can move from your blind spot 2 lanes away into your blind spot right next to you. Also, if something potentially dangerous/important is happening to your left, you will focus your attention there and not be aware of things happening to your right, even if you normally would have seen those things had you been looking straight ahead. I try to know what is going on around me, for the reason you describe (exit strategies), but I'm aware that sometimes I simply do not know what is to my left — and would not want to swerve there. Having seen friends driving Teslas, I can see how the surround-view could be useful for situations like this. However, I have also seen that cars appear/vanish for no reason, and it is clearly not reliable at this point.
shadowgovt 3y ago
Broadly speaking (and this is a very broad brush, details may differ), autonomous vehicles will resolve collision avoidance by doing something legal. Swerving is not legal and it's not legal because you might hit something you can't see. On the other hand, an autonomous vehicle maintains perpetual 360° situational awareness so it already knows whether it's safe to swerve.
- lozenge 3y ago
  Can you provide a source that swerving to prevent a collision is not legal?
- gnicholas 3y ago
  > Swerving is not legal and it's not legal because you might hit something you can't see. If I'm driving down the road and someone pushes a kid in front of my car, I'm very likely to swerve, even if I'm not 100% sure if there is someone next to me. If I hit a car next to me and cause damage/injure people, there is a good chance that a jury would find in my favor because a "reasonable person" (the relevant legal standard in a civil case) would seek to avoid the certain death of a faultless child, even if it means a chance of injury to passengers in a vehicle that might be nearby. I wouldn't swerve if I thought a bicyclist were there, but I would likely swerve into a vehicle, which affords occupants significant protection. And I certainly would not be found guilty of a crime ("beyond a reasonable doubt", the standard in a criminal trial) because it is not criminally negligent to swerve when faced with the certainty of killing a child. Under the "defense of self/others" doctrine, you can do many things that would normally be illegal, if you are doing them to prevent the imminent death of yourself or someone else. (These analyses are not super detailed, but IAAL.)
  - shadowgovt 3y ago
    AVs aren't doing near enough calculus to make a decision like that; they have obstacles and they avoid them. Their avoidance can allow for lane departure (that 360 situational awareness is handy), but they aren't going to solve a trolley problem in realtime if boxed in. The thing is... Neither are people. A kid darts out on front of you and you won't have time to make a conscious decision. The maneuver you make will be too fast for rational thought, and your brain will explain to itself why it did that later. ... Incidentally, this is one of the things that scares me about human drivers. Put a human on the stand to testify and most will say they never saw that kid, no matter what physical process happened to their eyes and neurons. I think most even actually believe it. In any case, I'm glad that in this incident they'll be able to do a full NHTSA workup, pull logs, and reprogram the machine to avoid an accident like this in the future. Can't do that to a human driver.

zx8080 3y ago

Thare's a, probably not very obvious, situation when human could have no such accident in the same scenario. At all. Preventing an accident does not leave any trace. No one knows about the scenario if there's no accident. How many similar (and others) accidents are being prevented by human drivers daily, and no one knows about them, at all? We just don't know. That's why I take any stats of "human vs AI" driving with a grain of salt.

CityOfThrowaway 3y ago
We can compare this statistically by looking at accidents per 1,000,000 miles driven of autonomous cars vs humans. Over long mileage, the rate of prevented accidents shows up in the data as a lower overall accident rate. I don't have the stats at hand but Waymo does great on this stat. Now, that's not to say that Waymo would avoid any specific accident better than a human. But from a public health standpoint, that's not really the right way to think about it anyways.
- talldatethrow 3y ago
  IMO these are a slightly fake stat. You're comparing to overall stats, that include the drunk, the high, the sleep deprived, and the crazy wild aggressive wreckless. If you compare to my mother driving, the injuries per mile are higher in autonomous. You can argue that I'm making an unfair distinction since those people exist, but I'd say the people driving under those conditions are commiting a crime and shouldn't be counted the same.
  - resolutebat 3y ago
    You're making a classic is-ought fallacy here. Drunk, high, sleep deprived or "crazy wild aggressive reckless" human drivers ought not to exist, but they do, and thus it's absolutely fair to compared autonomous drivers against them.
    - eurleif 3y ago
      If GP's mother switches from driving for herself to having Waymo drive for her, her risk of injury will increase. Therefore, if her goal is to avoid injury, it would be counterproductive for her to make that switch. That conclusion does not require any is-ought fallacy.

qwertox 3y ago

Yet it makes me wonder how the car would react if it came to a stop on top of the leg of the cyclist or in some situation where a normal driver would notice that he should get into the car again and back up a couple of feet. I'm not against Waymo.

Mageek 3y ago
My mom was hit by a driver when she was biking and fell with her arm in front of the wheel. The driver then decided to pull forward and drove over her arm. So humans don’t really solve that problem.

smcl 3y ago

Getting big "this is actually good for Bitcoin!" vibes from this. I don't think we can fairly assess who was at fault, given that none of us were there. But it is insane to just assume a manned vehicle would have fared worse, and declare it a victory for driverless cars.