Hacker News Reader

Comment by meritt | Hacker News Reader

https://www.nbcbayarea.com/news/local/2-Pedestrians-Suffer-M...

Pedestrian death from a Tesla just happened this afternoon. If this is found to be AP fault instead of driver fault, guessing we'll see AP banned by Wednesday.

View on HN

Also about 10 pedestrian deaths happened this afternoon from non Teslas, driven by humans. And 20 yesterday. So I wouldn't jump to conclusions about whether "AP" will be banned anytime soon.

Source: https://www.cdc.gov/motorvehiclesafety/pedestrian_safety/ind...

What is this statistic supposed to prove? I would expect there to be more deaths from non-Teslas because non-Teslas make up a majority of the market and furthermore not every person with a Tesla uses autopilot and even those who do use autopilot do not use it all the time, so the chance the autopilot makes a mistake is minimized.

We do not know if autopilot reduces the number of deaths, so let's not overhype this technology. I would wager it increases the number of deaths because

1. Autopilot technology is not good enough to be used with 0% human intervention (i.e -- there are many edge cases where human intervention is required that autopilot cannot handle).

2. Even though statement 1 is true, drivers naturally treat autopilot as if it is capable of driving with 0% intervention, for example they might check their phone while the car is moving.

You seem to be under the misapprehension that 1 pedestrian death equals 1 pedestrian death per day.

Just because one rando on the internet makes an "if autopilot" argument doesn't make it so. Even if that were the case, you are under the misapprehension that any blame should ever go to augmentation technologies for the actions of a licensed driver—whether that action be turning a wheel or pressing a button.

Sure, but this is one of the predictable downsides of Tesla's high-visibility, high-hype strategies. We shouldn't jump to conclusions, but we shouldn't be surprised that an attention-seeking company gets a lot of attention.

That article is very light on content and details, I couldn't find any other sourced either.

It would be interesting to know the circumstances around it running the red and "losing control".

I mean it just happened 2 hours ago, so pending a formal police report, I imagine details will indeed be light. Here's aftermath footage and bystander details: https://twitter.com/JennyGShao/status/1153066059802402816

These sort of unfortunate accidents happen every day but any self-driving vehicle will be scrutinized. If the driver was in control and ran a red light, they're obviously at fault. If the AP went haywire and caused this, who is legally culpable?

> If the driver was in control

> If the AP went haywire

It's disingenuous to imply that these are the only two possibilities, or that they are even mutually exclusive. The driver is always responsible for the operation of the car. Even if AP was engaged, AP never takes control away from the driver.

Remember, AP isn't a licensed driver.

AP does actually "take control away" in that it performs some emergency safety maneuvers automatically. For instance it can swerve to avoid an imminent collision with another vehicle, but would it swerve into a pedestrian?

AP never override's the driver's steering wheel input and it never stops the driver from pushing the brake pedal. AP never takes control away.

While I am 99.9% certain you are correct, there is a chance it malfunctioned. What you are saying is exactly what people were saying when people first started claiming they couldn't stop their Toyotas from accelerating years ago, and then it turned out there was some merit to their claims.

Sure—and when there's evidence of that, we can discuss it. Meanwhile, Occam's Razor says a guy behind the wheel of one of the fastest accelerating cars on the road was driving dangerously.

The new Emergency Lane Departure Avoidance actually briefly overrides my steering. It is somewhat annoying on a very narrow road, so I turn it off.

It's not accurate to say that it overrides your steering; it combines its decisions with yours—like on an aircraft where the co-pilot has access to the same steering controls as the pilot.

Personally I think lane keeping systems are mostly junk. Unless an automation (e.g. transmissions, adaptive cruise, auto headlights, auto wipers, lane keeping) does the not-wrong thing 99.9%* of the time, it doesn't lower cognitive load on the driver and is therefore pointless. If I still have to concentrate on staying in my lane, I'd much rather be 100% responsible for it instead of 1% responsible.

* Not sure how many nines, but probably more than that.

I am not sure I follow. Maybe we use different terminology. The car decides I am leaving my lane in an unsafe way, it does two things: it beeps loudly at me and it steers to avoid the side of the road. I can feel how it overrides my steering input.

Tesla’s lane keeping system is on 95% of the time during long trips. I rarely have to intervene and it is always obvious that I have to. I would rather not be without. But the good thing is one can chose not to use it.

AP is driver assist, so if the pedestrian isn't at fault, then the driver is at fault.

Edit: Looks horrible. I hope the cameras caught the action and can apportion blame appropriately.

Also, current AP doesn’t support stopping at lights. The best it does is blare alarms if it thinks you are going to blow one.

So unless the driver has some beta software, I can’t imagine AP being blamed.

If the driver knowingly had beta software and was testing it at high speeds on city streets, and they weren't paying attention to stop the car careening through a red light, then the driver is even more culpable because the manslaughter was premeditated.

Reply on news.ycombinator.com