Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Tesla allegedly in autopilot mode crashes into Texas house, woman killed

abcnews.com

Post reply on HN
Print view View original

61–70 of 125 posts

Re: Tesla allegedly in autopilot mode crashes into Texas house, woman killed

#61

Post by Reason077 »

gerdesj wrote: post #47
wnmurphy wrote: post #18

The telemetry will tell the story, not the clickbait news headline based on preliminary information. I seriously doubt FSD would drive into a house.

Are there any US federal or state laws stipulating some sort of black box style recording of data for accident investigation? If not then I suspect a Tesla will turn out to be quite surprisingly forgetful about what it was up to in a road traffic collision. RTC is a UK term that took over from RTA (road traffic accident) - it describes what happened rather than heading off into the weeds as to cause.

I don’t know about laws, but Teslas automatically record and save everything in the event of a collision. My brother was hit by another car while driving a Tesla a few years ago and it was very easy to retrieve video from the Tesla’s cameras and show who was at fault.

Re: Tesla allegedly in autopilot mode crashes into Texas house, woman killed

#62

Post by Reason077 »

Retric wrote: post #54
Reason077 wrote: post #48

Earlier quoted context omitted.

The garage door I can certainly believe. Its AI brain just won’t be trained to look for that sort of thing. I’m surprised about the red lights and animals, however: ours seems very cautious around any kind of live animal on the road, even braking and manoeuvring to avoid birds on the road. It’s not so good at avoiding the corpses of already dead ones, however (bump!).

> The garage door I can certainly believe. Its AI brain just won’t be trained to look for that sort of thing. Why not? That’s likely to come up at minimum thousands of times per day, and likely vastly more as the system improves.

I get the impression that it kind of comprehends the world as a horizontal plane. It’s focused on objects on the surface and isn’t looking for hazards coming from above. Could that be improved? Sure! Is it a priority for them? Maybe not…

Re: Tesla allegedly in autopilot mode crashes into Texas house, woman killed

#63

Post by jcgrillo »

aetherspawn wrote: post #58
erentz wrote: post #46

Earlier quoted context omitted.

You’re extremely confident about this just based on your experience of it not happening to you. It’s good that it hasn’t driven you into a house but that doesn’t mean it can’t fail in a way that does drive someone else into a house.

Well, I’ve seen it very confidently and correctly navigate rural unmarked roads, and drive around rocks and logs across the road and things like this. I find it hard to believe it would drive through a huge stationary object like a house, which I imagine is definitely in the training data. I guess I’m not saying it’s impossible, but I think at this stage it’s improbable.

The computer does not see things like you see things ;)

Re: Tesla allegedly in autopilot mode crashes into Texas house, woman killed

#64

Post by Retric »

Reason077 wrote: post #62
Retric wrote: post #54

Earlier quoted context omitted.

> The garage door I can certainly believe. Its AI brain just won’t be trained to look for that sort of thing. Why not? That’s likely to come up at minimum thousands of times per day, and likely vastly more as the system improves.

I get the impression that it kind of comprehends the world as a horizontal plane. It’s focused on objects on the surface and isn’t looking for hazards coming from above. Could that be improved? Sure! Is it a priority for them? Maybe not…

At minimum it needs to pay attention to railroad crossing bars and bridges.

They show a 2D representation to the driver because that’s good enough for drivers, but I wouldn’t assume that represents how the system operates internally at every stage. Even navigating requires the concept of bridges crossing roads without intersecting them.

Re: Tesla allegedly in autopilot mode crashes into Texas house, woman killed

#65

Post by jcgrillo »

ares623 wrote: post #36
jcgrillo wrote: post #29

Earlier quoted context omitted.

It's shocking. I can't understand why people repeatedly fall for it.

My nightmare has become reality. Software engineers now rule the world. Life and death policies being made by people who last week you were arguing about whether using Kubernetes is worth it for a throwaway project, or whether tabs or spaces are more appropriate.

I just wonder how bad it will have to get before they decide we need to be reigned in. Going by how the 20th Century went, if it's anything like that we're in for a rough time before we get there--like, to a place where software engineers have to act like, well, engineers.

Re: Tesla allegedly in autopilot mode crashes into Texas house, woman killed

#67

Post by JumpCrisscross »

Reason077 wrote: post #48
JumpCrisscross wrote: post #41

Earlier quoted context omitted.

> There are very few (if any?) known deaths caused by FSD accidents It’s tough to say given many data sources are aggregated. For what it’s worth, my parents’ car is a Tesla with FSD and I’ve stopped it from, off the top of my head, racing into a red-lit intersection, running over a small dog and running into a closing garage door. I still use it. It mostly works. But I’m vigilantly monitoring it in a way that isn’t…

The garage door I can certainly believe. Its AI brain just won’t be trained to look for that sort of thing. I’m surprised about the red lights and animals, however: ours seems very cautious around any kind of live animal on the road, even braking and manoeuvring to avoid birds on the road. It’s not so good at avoiding the corpses of already dead ones, however (bump!).

> I’m surprised about the red lights and animals

Both only happened once. But they were shocking when they did. (To be fair, my Subaru tried to push me into oncoming traffic because it was avoiding a “collision” from the guy in the other lane turning weirdly. Turned at collision-avoidance feature off.)

Re: Tesla allegedly in autopilot mode crashes into Texas house, woman killed

#68

Post by water-data-dude »

aetherspawn wrote: post #30

Ok, 3 things. 1) I’ve put enough kms on FSD - it’s taken me across Australia a few times, probably 10,000kms in total - to know that it isn’t going to drive into a house. 2) Even if FSD is enabled, there’s loads of things you can do to create an accident like press the brakes or accelerator pedals, which doesn’t necessarily disengage FSD right away, so let’s just wait for the telemetry to get released. 3) Regardless…

"I haven't encountered a rare edge case, therefor the rare edge case must not exist"

Re: Tesla allegedly in autopilot mode crashes into Texas house, woman killed

#69

Post by thephyber »

wnmurphy wrote: post #18

The telemetry will tell the story, not the clickbait news headline based on preliminary information. I seriously doubt FSD would drive into a house.

And _we_ will never see the telemetry.

Maybe NTSB will.

Certainly Tesla will. And that will inform how their PR team responds to this collision.

Maybe some plaintiffs will if they can manage to subpoena the data from Tesla in some hypothetical future court case.

The reason I will never buy a Tesla is because it is one of the most advanced surveillance systems against the driver, but there is no one empowered to inspect the car / company (comprehensively, not superficially).

> I seriously doubt FSD would drive into a house.

Strawman. You seem to be insinuating that FSD intended to aim for a house. Usually the chain of events would start with something like “FSD was engaged on the road and in the intended lane” then “FSD lost track of the boundary of the lane” or maybe “FSD identified an obstruction in the lane so it maneuvered out of lane”.

Re: Tesla allegedly in autopilot mode crashes into Texas house, woman killed

#70

Post by golem14 »

paulryanrogers wrote: post #50
iancarroll wrote: post #8

Earlier quoted context omitted.

The latest FSD does not attempt to check if your hands are on the wheel at all.

Does it do alternate checks for human awareness, like gaze detection?

Yes. But if you place screens strategically, the camera won't be able to see that you use them.
Post reply on HN