Hacker News Reader

Comment by ineedasername | Hacker News Reader

I don't think any reasonable person could think "autopilot" meant this was okay to do. It seems more like thrill seeking, maybe?

Yes, I think Tesla dangerously over-sells the autopilot capabilities, but I have a hard time laying the blame for this accident at those feet. If anything, it's unfortunate that such a ridiculous thing happened, because Tesla will be able to use it to paint all autopilot failures with the same brush of irrational behavior, when many are a rational response to the the over-sold marketing from Tesla.

View on HN

I guess we need the definition of "reasonable". Something like 20% of Americans believe that they can buy a self driving car today. That's not an accident, that's what Tesla is telling people.

Don't believe me? This video from the FSD beta opens up with the text "The person in the driver's seat is only there for legal reasons. He is not doing anything. The car is driving itself"

https://vimeo.com/188105076

At some point, we need to acknowledge that Elon Musk is deceiving his customers, his investors and all the hand-wavey legal bullshit around that is just that...bullshit.

Wow, you're right: I knew FSD was supposedly in beta & talked about, but now they're marketing it in those terms on their front page for the Model S. I wasn't aware they'd proceeded from calling it autopilot to FSD, and autopilot was bad enough.

Yes, it's therefore at least somewhat reasonable for someone to think they didn't need to be in the drivers seat. Although in this case it doesn't look like FSD was installed or activated on the car, if they're marketing it as a feature then sometimes even reasonable people will be confused and believe they have that option.

I mentioned this repeatedly in the thread yesterday, but I don't think that autopilot was actually in use here. The environment was a tiny cul de sac, and this was a high speed collision. Autopilot simply won't accelerate like that, it won't reach the speeds observed.

More likely they were playing with launch mode somehow and I guess thought that autopilot would magically engage and keep them safe? But IIRC those aren't compatible features.

I think it's correct to say that no one was driving the car. I just don't believe the CAR was attempting to drive itself either.

Edit: And per Musk, the card didn't have the FSD beta software (nor had even purchased it) and autopilot indeed wasn't enabled: https://twitter.com/elonmusk/status/1384254194975010826

Newspaper: "You're self-driving car killed people!"

Company: "Nah bro, it was my car's feature X that killed them, not the self-driving feature"

Newspaper: "But self-driving cars are supposed to be safe!"

Company: "They are safe. It's just feature X cars that aren't safe. And we give people a choice."

Newspaper: "But self-driving cars are supposed to be safe!"

Which is to say, no one paid to attention to the exact breakdown because most people aren't going to dive into the options. If Tesla wants self-driving to seem safe, they sure as better disable all unsafe self-driving features.

But... "Feature X" in this case is the accelerator pedal.

Yes, if Tesla marketing leads people to believe their cars safe just launching with no driver, Tesla technology should probably prevent drivers from just pressing the accelerator peddle from the passenger seat.

I mean, just automatic transmission imposes a few restrictions on the driver that standard doesn't. "Maybe self-driving" is going to create driver behaviors that definitely-not-self-driving doesn't. Cars have been elaborately idiot-proofed for a long time 'cause a lot of idiots drive.

Agreed. Even if they somehow had autopilot enabled, they defeated multiple safety measure to do it like the steering wheel weight and sitting on the buckled seat belt.

Right, but I think it's important to note the conflation of arguments here. If autopilot had been successfully engaged, and the only thing these drivers did wrong was to sit in the wrong seats while the car made all the decisions, then I think it would be very reasonable to view that as an autopilot failure and something that needs to be fixed. The car clearly would be guilty of trying to drive them into a tree at high speed.

That said, what seems much more likely here is that they were engaged in some kind of manual work to get the car to accelerate ludicrously. And that's not a situation against which the autopilot should be reasonably expected to protect against.

At a minimum, to drive with AP from the back seat, you would need to

a) Plug in a seat belt defeater, or plug in the seat belt after you hop out.

b) Hang a weight off the steering wheel so it feels torque.

At some point, it's hard to ban people being stupid. You can buy a woodchipper and take the guard off and shove your arm in, do we need to ban woodchippers? Or if we build something with layers of guards and safeties.. at some point you the user are responsible for what happens when you bypass them?

> Autopilot simply won't accelerate like that, it won't reach the speeds observed.

They said the same thing about the Toyota acceleration problem.

I'm not saying it's impossible, just that it seems poorly justified. Tesla autopilot and FSD footage is EVERYWHERE on youtube. If there was really some kind of anomalous acceleration problem surely someone would see it. There are all sorts of other persistent gripes about the system, but AFAIK rapid acceleration and overspeed simply hasn't been observed.

Which was never concluded to be a Toyota software issue.

Then there's the pesky fact that a car needs like 1000HP to overpower the brakes on something like a modern Camry. So if people were actually pressing a brake pedal, which is physically connected to the hydraulic brake system, the car will stop.

Yeah, they were successfully sued over it, but all that means is that they were probably responsible based on the information the jury was allowed to hear. And I believe much of the culpability came down to the fact that the floormats did actually move into the pedal box for that era Toyota (mine did it all the time, even post-recall).

If I remember correctly, they found many bugs in the Toyota software which could cause unintended acceleration and also could cause braking to be ignored. This report has some info: https://users.ece.cmu.edu/~koopman/pubs/koopman14_toyota_ua_...

PS: I had a Prius during this time. Even after selling it, without experiencing any acceleration issues, Toyota wrote me a $2k check to agree to never sue them.

Didn't this turn out to be floor mats getting stuck under the accelerator?

That was a story Toyota used to evade investigation, which came later anyway.

The car should not drive without the weight of a person in the drivers seat and the steering wheel sensing 2 hands on it. Autopilot and FSD should immediately disengage if those two criteria are not met. Easy fix.

It does have a steering wheel weight sensor. People learned you can hang a weight on a rope to "Trick" it. It should not require an arms race to outsmart people trying to defeat safeties.

“But they are so close to FSD.”

What are people going to think? People will take what they think is a small risk.

“Major improvements are being made to the vision stack every week. Beta button hopefully next month.

This is a “march of 9’s” trying to get probability of no injury above 99.999999% of miles for city driving. Production Autopilot is already above that for highway driving. 6:17 PM · Apr 14, 2021”

UPDATE

To be clear, I obviously understand they are not close to FSD. I’m trying to get people to understand how many in the general public will view this.

> But they are so close to FSD.

No, they are not. I own a Tesla with the FSD package and I've seen many of the videos with FSD beta. FSD is definitely a remarkable technology but they are nowhere near close to FSD.

Lets say that somehow the only thing these people knew about a Tesla was that tweet and they bought a Tesla based off of that Tweet and someone else enabled autosteer (beta), which is off by default on a new Tesla. And some how they buckled in the driver's seatbelt and put a weight on the seat accidently, not because they were trying to circumvent Tesla safety features.

They would still be in a situation where they knew that there was a chance that the could be injured by relying only on autopilot. No one should expect autopilot to be perfect and if they value their safety they should pay attention at all times just like the Tesla UI tells you when you enable autopilot and while you are driving.

We’re told Coronavirus vaccines are something like 94% effective

We’re also told FSD is 99.999% effective and getting better fast

There are those in the general public who will take that to mean that it’s safer than they should

Tesla's FSD is a complete fraud. Musk has been saying it's less than a year away since 2018

https://youtu.be/antLneVlxcs?t=639

https://youtu.be/antLneVlxcs?t=725

https://youtu.be/antLneVlxcs?t=489

https://youtu.be/antLneVlxcs?t=582

> since 2018

At least 2016, actually.

I think your definition of reasonable may exclude large numbers of people then, possibly over half or more. What maybe self-evident to you as a technical person may not be to folks who think “aol is the internet” equivalent of today

Reply on news.ycombinator.com