Safety Tests Reveal Tesla FSD Will Hit a Child Mannequin in a Stroller | Hacker News Reader

Safety Tests Reveal Tesla FSD Will Hit a Child Mannequin in a Stroller

dawnproject.com

35 pointsVeserv24 comments

Well this isn’t anything new.

When the auto industry faced problems of pedestrian deaths in the 20’s it was pedestrians because they were being shifty jaywalkers causing accidents so naturally pedestrians were criminalized. Then cars took their rightful place as the paramount concern in city planning.

How is this any different? These kids need blamed and shamed for impeding progress in 2020 just like 1920. Cars are the number one priority, not people.

By the way this is sarcasm. Hopefully history does not repeat.

It’s different because this product is billed as “full self driving” and it is programmed to run over strollers.

Love they still label it “self-driving” yet the fine print is anything other. Unfortunately, there is a cult following on musk/Tesla where facts sometimes are ignored.

I believe Musk is starting to loose its shine unless gets another rabbit out the hat. The Tesla Bot may be a such attempt but it's a hard sell because it's not really something new(i.e Boston Dynamics have been trying it for decades)

Ford is at least a year ahead of Tesla in the pickup market, and with a product that traditional pickup buyers are excited about.

So it sure seems like Tesla is still stuck on the premise that obstacles won't need to be avoided if the computer can't tell what they are from a single-view camera feed. It's hard to interpret all the failures differently.

I'd love for someone inside Tesla's self-driving division to show up in these threads to explain that they know their fundamental model is bad but are forbidden from even using stereopsis.

This is what I don't understand. Why is it ok (from Tesla's point of view) for the car to hit __anything__? Even hitting an empty stroller would be a bad thing.

I think it's surely safety-critical that the car hit _some_ objects. For instance, you don't want to be going 70mph down the highway only for the car to attempt to come to a screeching halt when a paper bag or some other litter blows into your lane. Behavior like that is liable to itself cause a pretty serious accident, since it's so unlike the reaction any of the drivers around you will be anticipating.

Likewise, unpleasant though it is to admit, it's probably (in many circumstances, especially inclement weather) safer to hit a squirrel or a raccoon that dashes across the road than it is to try to swerve to avoid them. On the other hand, there's going to be a good case from a driver safety perspective to avoid hitting a deer, and it is _critical_ that you attempt to avoid hitting a moose. Obviously, things here get much more complex when you're talking about a child or adult human being.

Point being, I think it's dismissive to just go "the car shouldn't hit anything it detects" - not all objects are equal, even assuming perfect detection (i.e. not mistaking a pothole for a giant rock).

What if it’s only going 10mph over a speed bump, like in this example?

You miss the point. If there is an object the size of a baby carriage in front of my car, I am going to avoid hitting it if at all possible. It doesn't matter what I think it is. It might damage my car if I hit or, or make me lose control (if the front of the car rides up on top of it), or possibly even deploy the airbags. But I am not going to even think of any of these things while driving -- I am just going to avoid hitting it. Tesla is unable to do that and I don't understand why that is ok with anyone.

As mikequinlan said, you're so close and yet thinking about it completely backward. Not knowing enough about what the object is to know whether it's an obscured human (or a boulder that will total your car and leave you in the hospital) is all the reason you need to not hit it. You decide whether to hit a squirrel after you determine that hitting it won't ruin your day, not before. Not hitting something needs to be the default unless determined that it's safe to do otherwise.

The Dawn Project's mission is interesting; addressing the threat of critical software written in unsafe ways. Their first target is less so: "Our first campaign is targeting Tesla full self-driving cars." Why? Just for the press coverage? Because it's an easy target?

Seems like plenty of other much higher impact places to try to get leverage on the problems of unsafe software, like encouraging and improving safe systems programming languages or retrofitting / improving safety systems for legacy code. Compared to the safety discussions at the C++23 meeting (https://news.ycombinator.com/item?id=33762154), this seems like pure theater.

Tesla/Musk are very convenient targets for political reasons. The most glaring one is Tesla's refusal to be a part of the UAW. Musk also is problematic because he personally has done more to help climate change than democrats, and he did it with 1/10,000th amount of capital.

It has nothing to do with the UAW. The guy running the political ads owns a competitor driver assistance technology company.

Tesla's self-driving claims were a joke and blasted by people in the field long before Elon showed his politics

> Seems like plenty of other much higher impact places to try to get leverage on the problems of unsafe software, like encouraging and improving safe systems programming languages or retrofitting / improving safety systems for legacy code

Tesla seems like a perfect example of "critical software written in unsafe ways" if you ask me.

I don't think you're saying that it's c++'s fault that Teslas are programmed to run over babies.

So can you provide a better real world example of where unsafe software is directly putting people's lives at risk?

Windows PCs at your doctor's office running Epic.

O'Dowd uploaded a video a few months ago that also showed FSD hitting a child mannequin, which was shortly thereafter proven to be completely faked. Why should we take him at his word this time?

Because the video was not faked. There are statements under oath that the video was not faked. The underlying footage shows that it was not faked. Subsequent videos which address complaints of transparency and clarity show they are committed to proving the videos are not faked.

The only official Tesla statements on the matter are a cease-and-desist claiming that the videos were faked threatening immediate legal action for defamation if the videos were not taken down that they delivered over three months ago [1]. In the mean time even more videos were released, more ads were run, more public statements were made, and direct challenges to Tesla were made, but Tesla has not followed up with further cease-and-desist letters or followed through on their baseless legal threats. Given their traditionally litigious nature, there is no reason for them not to follow through even if they could not prove all prongs of a defamation case since with discovery it would be trivial to prove publicly in a court of law that the videos were faked if they were, in fact, faked. Their reluctance to do so is strong evidence that even Elon and his lawyers think a lawsuit would reveal truths they would much prefer silenced and buried.

[1] https://dawnproject.com/wp-content/uploads/2022/08/2022.08.1...

Duh. It's only beta. It will hit and kill people every now and then. Get over it. Musk is amazing.

That's not how the argument goes. If you don't let Tesla run down children today, the blood is on your hands for preventing the development of L5 which would have saved infinite lives in the future.

"Training data"

Full self driving means fully self driving.

Beta product, sold to consumers, should not mean, it's kills people some of the time

I mean, I assume this is /s, but with the worshippers at the Church of Saint Car, really, who knows?

Reply on news.ycombinator.com