Live data from Hacker News

SF police stumped after trying to ticket driverless car for illegal U-turn

theguardian.com

11–17 of 17 posts

Re: SF police stumped after trying to ticket driverless car for illegal U-turn

#11
post #7

Earlier quoted context omitted.

> the illegality of the U turn is implicit, the software learned that the path was valid from other drivers One would hope that the traffic laws of an area are "hard coded" and not dependent on what other drivers do.

I don’t think you can really bake that in because the traffic rules can change day-to-day. For example, there’s a street in my neighborhood that’s normally open for two-way traffic, but one of the buildings that fronts it is being renovated so the street was changed to one-way for about a month, and as of a couple of days ago it’s still one-way but in the other direction. Imagine trying to get a car to work that out…

That's exactly a situation where signage is mandatory (and, I think, legally required) to say what the current rule is. The hard-coded traffic law that would apply here is "obey the traffic signs".

The point I was making is that these cars should absolutely not be "learning" what rules to follow by observing what other drivers actually do. When there is no clear signage, there are well-defined laws about what drivers need to do and self-driving cars need to follow those even if human drivers don't.

Re: SF police stumped after trying to ticket driverless car for illegal U-turn

#12
post #7

>During a DUI enforcement operation, officers in San Bruno pulled over a car without anyone behind the wheel after the autonomous vehicle made an illegal U-turn at a light. I'd bet a lot of money that the intersection isn't signed or is poorly signed (because if there's one thing these vehicles do pretty well it's obey obvious signage) or not signed and the illegality of the U turn is implicit, the software learned t…

> the illegality of the U turn is implicit, the software learned that the path was valid from other drivers One would hope that the traffic laws of an area are "hard coded" and not dependent on what other drivers do.

That depends on what other drivers, I guess

The Waymos I ride have developed a habit of swinging right into bus turnouts.

A bus turnout looks like a dedicated right-turn lane, except it ends and you must merge before turning right!

So these Waymos are illegally entering these lanes for no good reason and seem completely unaware of their actual designated use or signage. I cannot figure out why.

Re: SF police stumped after trying to ticket driverless car for illegal U-turn

#13

In most jurisdictions, if a speed camera catches a car breaking the speed limit, doesn’t the ticket go to the registered owner of the car, and it's their responsibility to claim someone else was driving or pay the ticket?

What if the car is stolen?

Re: SF police stumped after trying to ticket driverless car for illegal U-turn

#14
post #8

If it is a problem to give a ticket for such minor infraction, how is police going to find a person who is responsible for accident or death? Police will just shrug? Until it is exactly specified who is responsible for what in written law, these cars should not be on public roads.

When people are actually put in danger, the historical response is that the entire fleet gets suspended / pulled off the roads (on top of the making headlines and triggering investigation stuff). 2018 Uber, 2023 Cruise, 2025 Zoox cases.

That will not be a viable response for very long. Will it affect the entire fleet in one market or the entire fleet worldwide? How long will it affect the fleet? When robotaxis are an integral part of a local economy, then the incentives diminish for taking them entirely offline. The three instances you mention are fleets under testing circumstances and not commercial service, so there's a distinction!

Sooner or later we'll be making so many omelettes that the occasional broken egg will be, reluctantly, accepted, and the penalties will amount to the usual corporate wrist-slapping, and that will be that.

Re: SF police stumped after trying to ticket driverless car for illegal U-turn

#15

In most jurisdictions, if a speed camera catches a car breaking the speed limit, doesn’t the ticket go to the registered owner of the car, and it's their responsibility to claim someone else was driving or pay the ticket?

There has to be a positive identification of the driver.

Here, let's ask Gemini: https://g.co/gemini/share/13b5989fca49

So it turns out that some infractions in some jurisdictions are "owner liability" and so all the authority needs to do is cite the registered owner, because they're always responsible for anyone who is permitted to drive their car.

If the jurisdiction is not "owner liability" then it may be necessary to ID the driver instead, but the ticket is still sent off to the registered owner for handling, because the ticket and the car and the driver all go together.

Regarding a sibling query if the car is stolen, I'd say the owner has bigger fish to fry, and therefore would have reported the incident already, and so untangling a speeding offense would necessarily involve documenting the report and the incident of theft prior to those infractions. It may get a little sus if the owner said "hey my car was stolen and then returned to me just around the hour this photo was taken! it weren't me bruh!"

Let's not forget all the disparate jurisdictions and patchwork of laws which make up this sort of scenario. We can never make blanket statements that cover every situation.

Re: SF police stumped after trying to ticket driverless car for illegal U-turn

#16
post #7

Earlier quoted context omitted.

> the illegality of the U turn is implicit, the software learned that the path was valid from other drivers One would hope that the traffic laws of an area are "hard coded" and not dependent on what other drivers do.

I don’t think you can really bake that in because the traffic rules can change day-to-day. For example, there’s a street in my neighborhood that’s normally open for two-way traffic, but one of the buildings that fronts it is being renovated so the street was changed to one-way for about a month, and as of a couple of days ago it’s still one-way but in the other direction. Imagine trying to get a car to work that out…

> Imagine trying to get a car to work that out on its own.

Wdym imagine? I was told self driving cars are better at literally following rules than humans?

Re: SF police stumped after trying to ticket driverless car for illegal U-turn

#17
post #8

Earlier quoted context omitted.

When people are actually put in danger, the historical response is that the entire fleet gets suspended / pulled off the roads (on top of the making headlines and triggering investigation stuff). 2018 Uber, 2023 Cruise, 2025 Zoox cases.

That will not be a viable response for very long. Will it affect the entire fleet in one market or the entire fleet worldwide? How long will it affect the fleet? When robotaxis are an integral part of a local economy, then the incentives diminish for taking them entirely offline. The three instances you mention are fleets under testing circumstances and not commercial service, so there's a distinction! Sooner or late…

That’s a good point - I’d naively thought of it as “self driving gets way better than humans -> self driving ‘goes to prod’”, but there is a real possibility that regulatory scrutiny falls off too early, or worse, the lack of incentives results in backsliding on safety after regulatory scrutiny is off.

Rollouts seem conservative enough for the former, but the latter…

Post reply on HN