Waymo illegal overtake into wrong-way driving
old.reddit.com
old.reddit.com
If the car understood the state of the world around this, this is kind of a huge mistake in behaviour. And if it didn't it is concerning as well as it would mean it can end up in the opposite lane without knowing it. I would suspect maybe a bit of both, that it misunderstood the world when it started to overtake, and then was unable to abort the process safely. In that case I probably would have expected it to stop driving forwards and put on hazard warning lights as long as it is in the wrong lane.
Seems like a failure of their path planning in this case, the same way if someone were to try to overtake on a two lane road, one would have to abort the overtake if the vehicle one is trying to pass keeps on accelerating such that overtaking isn't safe anymore.
FWIW: The driving test (at least in Blighty) teaches you to look as far down the road as you can, never to just do what the bloke in front is doing.
It stays in the wrong lane; is it because it can't rejoin because the lane is occupied by the unicyclist (looking at the video it does not appear to make efforts to go back) ? or because it has already forgotten the context and is now driving happily in a lane ...
It only stopped when its path is blocked ...
“After starting that maneuver, out of an abundance of caution around these vulnerable road users, and to avoid getting too close or cutting them off, the Waymo remained in the oncoming lane for longer than necessary before returning to its original lane of travel,” the company said.
> before returning to its original lane of travel
When it was only able to do so when someone started to block its path to force it to stop ...
These kinds of mistakes will very likely happen with self-driving cars, the important part is to address them and improve the cars. I don't expect the cars to be flawless from the start (I do expect some fundamental safety mechanisms to work reliably from the start like the automatic braking before obstacles). Downplaying the issue here is a bad look for Waymo.
"Waymo you're going the wrong way!" and the human uni-cyclist steers in front of Waymo, knowing that it will brake, in an attempt to force it back into the correct lane.
That's a level of optimism I don't have....
...on the same level as a bloke I saw advertising a food blender on the shopping channel, sticking his fingers into the thing to prove the safety mechanism would stop the blades.
I'd put this inherent danger at somewhere around 0.1% of the time. I'd say I'm forced to drive dangerously about once every 10 miles of driving, but the fact that people must drive dangerously occasionally to use the system means that a machine can't ever learn to use it safely.
I fully admit to having done things in the car far dumber and more dangerous than this when I was young. I'm not excusing the Waymo, rather I'm putting it in perspective.
But the goal is not exactly to have autonomous driving at their level ...
I would not expect a human to drive two blocks while failing to go back to the correct lane; I would expect him to realize after 50m that there are too many and slow down and go back behind the uni cycle. In the same way I do not expect a human driver to stop after hitting someone then start again and drag the body with it.
The problem is that it's easy to pump the statistic while driving miles without issue then it utterly fails when the situation is not well-defined anymore.
Machines don't make "mistakes". A mistake require intent. In case of autonomous driving, any problem is a defect in the system that the whole fleet share.
> I'd also expect these mistakes to become less and less common
So the people should just tolerate it in the mean time "for the greater good"? oh please, it's an enterprise.
I don't think it's unfair to point out that humans, quite often, also suck quite badly at driving.
Many of those reasons are illegal (alcohol, fiddling with your phone) so it's not a question of fairness an automated program was supposed to be better than that.
I don't see anything unfair in requiring that automated driving which is supposed to not be afflicted from our human failing to not perform illegal and dangerous driving ...
Especially since autonomous driving might bring more car not less (similar to Uber and Lift impact https://www.nature.com/articles/s41893-020-00678-z#Sec8 )
I could point to a double pedestrian crossing near my place where cars routinely speed. Various people have told me that the drivers cannot see the pedestrians, for a variety of reasons (buildings, cars, and especially a large tree near one corner), yet the human drivers are confident to go 10-20 kph over the speed limit through the area, despite this lack of information. I can give many similar examples - people really do suck at driving, they just tend to get away with it most of the time because rare events are, well, rare.
AI sucks for a variety of other reasons. The only real question is whether on average, over all scenarios and ratios of humans to AIs populating the roads, one sucks more than the other. It's something I don't think we have the answer to.
what is the threshold for permanently terminating Waymo's license?
- an example like this of dangerous driving?
- 1 death by dangerous driving?
- 10 deaths by dangerous driving? (say it veered right in this video into the other road users)
- 100 deaths by dangerous driving? (say it veers into a sidewalk full of pedestrians)(Possibly also not using the indicator which is another point, but I don't think we can be 100% sure it wasn't used before the change started - California seems to require it only before not during lane change if I read it correctly)
doing it for a second or two would likely be classed as "driving without due care and attention", resulting in a fine if convicted
however doing it for as long as in this video I suspect would be classed as "dangerous driving", with disqualification being the likely result
This is a variation on the trolley problem. Would you rather have roads that are 10 times safer, with 10 times fewer accidents and 10 times fewer deaths, but where the accidents are completely unexplainable and where nobody can be held guilty, or would you rather have the world stay as it is?
I think with 100 deaths the company would get liquidated. This incident alone probably caused more damage than a cost of one single driver's license suspension.
Just yesterday I got yelled at by an Italian police officer who was directing traffic at a busy junction (I was in a rental car). [Full disclosure: I stopped, wound my window down, we had a polite conversation, everything turned out fine...]
The key difference is that if I screw up badly enough when driving, I risk a monetary fine and/or losing my licence and/or getting sent to prison. For most human drivers these represent clear incentives not to screw up when driving.
Questions:
Have we agreed on suitable penalties when an autonomous vehicle screws up, given there isn't a single human driver to sanction?
Could the fleet get grounded?
Could the company get fined?
Could relevant executives go to prison?
Cruise's permit was suspended after their accident
https://youtu.be/zdKCQKBvH-A?t=742
These companies should be fined hefty amounts and barred from beta testing out in public and putting people in danger. Governments should carry out comprehensive tests on a self-driving car's capabilities, same as cars without proven passenger safety (Euro NCAP) aren't allowed to be on roads carrying passengers.
And now this great AI company is also providing AI services to pariah states like Apartheid Israel to be used on people. That's scary.
Thats how the issues with automobiles vs other road users was resolved in the 1920's. The impetus is there, corporations want more and more robots for everything.
> a former Baidu executive, Drive.AI board member, and one of the industry’s most prominent boosters — argues the problem is less about building a perfect driving system than training bystanders to anticipate self-driving behavior. In other words, we can make roads safe for the cars instead of the other way around. As an example of an unpredictable case, I asked him whether he thought modern systems could handle a pedestrian on a pogo stick, even if they had never seen one before. “I think many AV teams could handle a pogo stick user in pedestrian crosswalk,” he told me. “Having said that, bouncing on a pogo stick in the middle of a highway would be really dangerous.”
> “Rather than building AI to solve the pogo stick problem, we should partner with the government to ask people to be lawful and considerate,” he said. “Safety isn’t just about the quality of the AI technology.”
https://www.theverge.com/2018/7/3/17530232/self-driving-ai-w...
This law that sets the default of it being illegal to be in the road doesn't exist in much of the world. It's really up to the self driving cars to respect other road users at least as well as a human driver would, not expect to be able to mould the laws around the world to fit their limited capabilities.
Now, this does not say anything about the Waymo incident. In that case, the car performed an illegal maneuver, which cannot be blamed on other street users.
If we need to design roads for Artificial Stupidity then we should design this roads from scratch and gain some extreme speeds and extreme safety from the work.
For real. Not /s.
Not in downtown Melbourne, but a definite possibility in suburban streets.
https://www.bbc.com/news/uk-england-london-68887800
They smashed up a bus and a regular taxi en route to Freedom!
Maybe it thought they were on a bike lane? Where I'm from it's illegal to ride like that in a group outside of police-sanctioned events - you're supposed to ride one after the other and keep right.
I’m in Massachusetts where lane splitting is legal in some cases (single lane each direction, posted speed limit below some amount, probably something else). Some other states have similar rules, but in most it’s flatly illegal.
I believe California is an unusual outlier in that lane splitting is allowed with no additional restrictions.
> you're supposed to ride one after the other and keep right.
In many states bicycles are entitled to use the full lane. I often feel safer riding in the center of the lane for a few reasons:
1. The center of the lane is less likely to have small ride-ruining debris like glass, sticks, and nails
2. Cars surprisingly give me more space when passing if I’m not riding off to the side
3. It seems that potholes occur more often to the side of the rode than in the center, or at least get fixed more often when they’re toward the middle of the road
Note: I’m talking about the middle of the lane vs the side, not the middle of the lane vs the shoulder. The shoulder isn’t something you should travel in anyway, even on a bike.
Ok. So what are the consequences for this human operator and/or Waymo?
Human drivers make mistakes all the time, it’s true. But there are consequences for those mistakes. If a human drives into the opposite lane that’s a big problem, especially if they end up causing property damage or bodily injury. Why are the same standards not being applied here?
OTOH, I doubt the operator still has a job, and Waymo is going to face significant consequences. Likely formal consequences, but if not, the punishment by the court of public opinion is going to be massive.
If you're Waymo apparently you can tell the car to run a red light, cause a moped to crash and not even need to report it to the state.
https://www.forbes.com/sites/bradtempleton/2024/03/26/waymo-...