CR Engineers Show a Tesla Will Drive with No One in the Driver's Seat | Hacker News Reader

CR Engineers Show a Tesla Will Drive with No One in the Driver's Seat

consumerreports.org

Why is it possible to purchase this item that is specifically designed to override the autopilot safety features?

https://www.amazon.com/QCKJ-Steering-Autopilot-Assisted-Auxi...

You can turn off all kinds of safety features in your car

I got a 2021 MY car with the various features that the "legacy" manufacturers are starting to put in even low end cars, and it's interesting to me that some of the features stay shut off, while others turn on every time you start the car.

It doesn't seem like they had to make exactly the choices they did. I'm glad the lane departure warnings don't turn back on automatically.

I can't stand auto stop-start on a non-hybrid car. I've looked at and test driven cars with that over practically a decade and every time decided it was a deal-breaker. I have a deep seated insecurity about my car stalling just as I make a turn into traffic, so if the engine is going to turn off, there better be a backup. (and the vehicle better be made by one of the companies that sets the gold standard for reliability, not BMW or GM or whatever)

They put a weight on the wheel to actively defeat the countermeasures. At best you can say that Tesla didn't do a good enough job by using multiple signals to determine driver occupancy and attention. Beyond that, this is just obvious.

As a member of the general public, I absolutely believe it's on Tesla to ensure that their cars can't be tricked into driving completely without driver supervision.

I don't understand why you or others think it's fine just because somebody found a trivial way to defeat their checks. This is a severe public safety issue, since it's not just putting the occupants at risk.

> I absolutely believe it's on Tesla to ensure that their cars can't be tricked into driving completely without driver supervision

Do you also believe it's up to publishers to control how their cryptography libraries are used, and firearms manufacturers to predetermine each and every end user that will lay a finger on their products?

Are you using logic you know HN doesn't encourage that

Any normal car can be made to drive by itself if you put a weight on the gas pedal... If you are actively trying to circumvent the foolproofing you will always find a way.

That car will crash every time though which discourages doing it, the autopilot will only crash occasionally since it attempts to auto steer, which it makes it far more likely to be relied on and abused.

If people are actively trying to defeat the sensors I don't think it's Tesla's fault.

I think probably there shouldn't be a car available to the public that both can't drive safely without close monitoring and also tempts drivers to try this.

Nothing between basic driver's assist and L4. You can argue whether Tesla is in that range but that's definitely what their messaging is.

Yup. Level 2 isn't a good idea and level 3 absolutely should not exist. L4 is even marginal--we are seeing problems with jetliners from this. Pilots are so used to flying the automation that when the L4 pukes (or worse, gets confused) they're out of practice and make mistakes flying the plane.

(However, I think the problem with planes *could be* handled by changing the currency definitions. Time spent in L4 shouldn't count towards being current--and I wouldn't count time spent simply cruising at altitude, either. Only time where the pilot actually has to fly, whether simulator or a real bird.)

The title feels a bit misleading. Sure there was no one sitting in the seat but there was weight applied to the steering wheel to simulate hands...

Should they have a sensor in the seat as well as one in the steering wheel? I think so. Is this article partially clickbait? I'd tend to say yes.

How does the title feel misleading? It says directly and exactly what was done, and it appears that you agree that this happened.

From what I understand, it seems that it's entirely possible for this scenario to be replicated "in real life" by a driver losing consciousness or being in a state of partial consciousness.

That would cause a crash in non self driving vehicles as well.

The non self driving vehicle wouldn’t fake it though by auto steering, anyone who does this in a standard vehicle knows they will crash, people in the Tesla think it’s ok as countless videos have shown.

"Driving with no one in the seat" is much wider than "driving with only hands on the steering wheel". The former implies that a kid on the backseat could turn on autopilot and drive away without anyone else in the car, while the reality is that that hypothetical child would actually need to hold the steering wheel.

> From what I understand, it seems that it's entirely possible for this scenario to be replicated "in real life" by a driver losing consciousness or being in a state of partial consciousness.

How would a driver losing consciousness suddenly have no weight in the driving seat? This is just saying that there should be a pressure sensor in the seat combined with the sensors in the steering wheel. In both case nothing prevents people from falling asleep here.

I think a sensor in the seat is pretty standard thing nowadays in almost every car. It is used to alert you when you or your passenger has forgotten to fasten a seat belt.

I agree completely, this seems like an oversight for sure, but the title seems to imply that you don't need anyone in the drivers seat to turn on autopilot. This is false since you obviously need someone or something to "hold" the steering wheel.

It's simply not as big of a deal safety-wise if you still needs hands on the steering wheel.

I think they were standard even back in the day. My 2004 Mercedes had one since when I was moving and had boxes on the front passenger seat, it actually deactivated the airbag since I guess the weight was just in the "there's something but it's not a person -> possibly child seat" range.

Yeah. Even my small lawn tractor has a seat sensor...

I don’t think there is a feature to avoid warning you if the seatbelt is not fastened in case there is no driver.

They already do have a sensor in the seat. It's one of the ways the car knows when to turn on.

Well then I suppose we will see a software update soon to address this.

I'm gonna bet they're intentionally going lightweight on the driver checks because all the ticktocks are free advertisement for their 8k$ "autopilot" option

I believe they're working on using the rear-view mirror camera to watch for driver attentiveness, which would solve both this issue and the issue of their current driver attentiveness system sucking so, so bad.

I feel like we would enter in a problematic part of driver monitoring with that feature though. Tesla has shown that they were willing to use their data to protect themselves when a customer would crash their car so they might have access to a constant video feed that could be used by the police.

I am probably reading too much into this, but I just don't like possibly live cameras recording me...

On the positive side, though, if you get accused of being distracted and causing an accident when you were paying attention, the camera inside the car would prove that for you.

From my understanding, if someone were to do this IRL, and the autopilot disengaged in the middle of the ride, then the occupants are fucked because there is no way to force the autopilot to engage. I wonder if this has caused crashes already.

I'm thinking that might be what happened here.

Tesla is saying the autopilot wasn't engaged and couldn't be due to a lack of painted lines.

Ok, they defeat the sensors, take it out for a drive, they hit an area with no lane markings and the autopilot disengages and they can't recover.

This is an interesting conundrum. If it's currently on autopilot (for whatever reason) and the requirements aren't met anymore it should probably handle it gracefully and wind it down progressively until a human driver takes over, rather than shutdown assuming someone is at the wheel. Auto pilot in adverse conditions is better than no pilot.

In some multicopters when you loose signal it doesn't fall off the sky, instead it hovers gently down until it touches down on the ground, (or tries to fly back home which can cause a host of other problems)

In that case it has a safe failure mode. In the Tesla case it doesn't--it's relying on those lines for guidance. The car suddenly has no idea where the road is. The choices are basically turn off, or turn off and engage the brakes.

With a human able to take control the former is the better choice.

This is ridiculous. Yes, you can cut your hand if you put the knife on your finger than the steak. I wonder if they tested putting a mannequin or a robot with human like head with other cars.

You realize many saw blades have sensors that will turn off if a human touches them? Engineers have a responsibility to prevent as much misuse of their products as possible. It’s why antifreeze isn’t sweet any more and why we put artificial odors in gas.

They have some obligation, but certainly not to maximize. Nothing would ever get done.

tl;dr: "Funkhouser sat in the rear seat, and Fisher sat in the driver seat on top of a buckled seat belt. <... > Fisher engaged Autopilot while the car was in motion on the track, then set the speed dial (on the right spoke of the steering wheel) to 0, which brought the car to a complete stop. Fisher next placed a small, weighted chain on the steering wheel, to simulate the weight of a driver’s hand, and slid over into the front passenger seat. Using the same steering wheel dial, Fisher reached over and was able to accelerate the vehicle from a full stop. He stopped the vehicle by dialing the speed back down to zero."

Reply on news.ycombinator.com