My parents used to have an Audi A5 - sometimes, when you've hit a decline just right, it would shut off completely, no steering, nothing, all you could do was to let the car roll and reboot it.

This is a known software problem: http://www.audiworld.com/forums/audi-a5-s5-rs5-coupe-cabrio-...

I.e., the future is already here and it sucks

A family friend's Mercedes E500 brakes stopped working on 280, and he had to shift the gears down and pull over to the side (this is back in the early 2000s). He brought it to the dealer, and something along the lines of the dealer claiming they couldn't see a record of electronic failure in the logs, then refusing to give the car back to our friend.

He only got the car back when he engaged a lawyer to sue the dealer. The lawyer was happy to work on a contingency since "we have no chance of losing this one in court".

That's scary, especially considering the controversy over a similar problem that also lead to unexpected shutoffs: https://en.wikipedia.org/wiki/General_Motors_ignition_switch...

But in the link you provided there's several comments that it happened shortly after refueling, which suggests to me a much simpler hypothesis of the root cause --- bubbles in the fuel line.

Honestly, I wouldn't classify 8 forum posts as a "known software problem".

I think if this was as wide spread as you make it seem, there would have already been a recall made by VW as this is clearly a huge safety hazard.

Take the number of vehicles in the field, A, multiply by the probable rate of failure, B, multiply by the average out-of-court settlement, C. A times B times C equals X. If X is less than the cost of a recall, we don't do one. (Fight Club)

It's actually a paraphrasing of the Ivey (GM) and Grush-Saunby reports (Ford) from the 1970s. This was several years after Ralph Nader's famous book, Unsafe at Any Speed (1965)

1. http://lawprofessors.typepad.com/tortsprof/files/FordMemo.pd...

2. https://en.m.wikipedia.org/wiki/Unsafe_at_Any_Speed

Something like this would happen in my Civic too, not a decline but a small bump on the road like where two concrete blocks meet would set it off. It turned out to be a MAP sensor issue. This is something in the exhaust that feeds back to the controller, it was giving wrong values which caused the software to decide, fuck it let's shut it down. We all know the Toyota acceleration story. So the point of my story is, dont trust even reliable cars' software.

This was already happening 15 years ago.

I drove a Fiat Stilo on the highway when it suddenly decided to shift to 2nd gear because a gearbox failure occurred. It was a known software error.

Or alternatively, you're trying to get out and neither the doors nor windows will open. To add more excitement, you really have to go to the washroom.

This is why I believe there should always be a mechanical override for safety-critical features; locks, brakes, steering. I trust hardware far more than software.

I don't disagree, but it is worth remembering that purely hardware systems aren't foolproof either. I had a fuel pump fail on my old car (would be 15+ years old if I still had it), and it took out the brakes and steering at the same time (virtually standing on the brakes to slow down at all, and I ended up running the car off the road as I couldn't physically brake or turn the wheel enough to stay on the road).

I don't know that I trust hardware more, but hardware is roughly equally involved in any case...

I meant hardware in the "old-school" sense of physical linkages and hydraulics, not computer hardware whose behaviour is controlled by some massively complex blob of software.

For example, if I pull the door handle of my car, it directly activates the latch, and the only way that could fail is if something physically breaks or jams. It's not just a switch sending a signal to some "brain" in between that could just decide, deliberately or otherwise, to not honour my "I want the door to open" request.

People are often surprised to learn that, despite me making a living off software development, my daily driver is one that has exactly zero lines of code controlling it.

I meant hardware in the same sense as you. My point was that even if we assume the brain (borrowing your metaphor) makes no mistakes, introducing it does not reduce the amount of physical breakage and jammage that is likely to occur.

You're not choosing hardware over software, here - in which case I might choose differently - you're choosing to eliminate an entire link in the chain. That's always going to be a win for reliability.

Perhaps "trustworthiness" is not the best word for the concept. I'm not sure mechanical hardware is more trustworthy (mechanical stuff also fail often and can be unreliable), but their behavior is somewhat more malleable by the user (i.e. via extra force), as opposed to complex blackbox electronics/software.

I think we need a good term to describe this specific kind of trustworthiness.

nice... what do you drive? muscle car?

cycle ?

Yep, that's what scares me about the "drive by wire" systems standard on most cars today. Oh, and to add even more excitement, the car is on fire and the doors/windows won't open.

One should always keep a tool in the car in reach to use to break the windows in case of emergency.

I am not sure mechanical hardware is less likely to fail than software. For example, ford recalled door latches for a mechanical problem - https://media.ford.com/content/fordmedia/fna/us/en/news/2015... what is worse is my friend can't get the door latch replaced on her ford because "the part's on backorder."

You can smash the window by driving the pillars of the headrest down into the door to the edge of the glass. Twist down to start a crack at the weak edge and the whole tempered sheet will shatter.

https://youtu.be/tZTa8Nh0VlE?t=25

In normal cars, when you're on the freeway, it doesn't really matter if the door opens, because the air pressure will keep it tightly shut. It takes some considerable effort to push open a car door at 80 km/h.

Gull wing doors are different though, not to mention rear-hinged doors (aka suicide doors).

That's a very good point about rear-hinged doors. Never thought of that. I've seen Honda Element and some other cars have that rear-hinged doors and the setup didn't feel quite right but never able to figure out why.

The Element, and all the other modern, pillar-less 'suicide rear door' cars I've seen (Landcruiser FJ, Mazda something-something), have the rear door latched by the front one, so you can't open the rear without first opening the front, 'normally hinged' door - it's a pretty good safeguard against latches failing and doors being sucked open while driving.

Yep. Same on the electric BMW i3 btw.

In Element and similar pickup-SUVs the rear door is still quite small so it doesn't feel so scary. In others, such as current Opel Meriva B, the door about as big as the front door. And just look at Rolls Royce Wraith.

My dad's first car (which he bought in 1960) also had suicide doors in the rear; that was a Moskvitch 400, which in turn was practically the same design as Opel Kadett model 1938 - the Soviets took the design, toolings and machinery from occupied Germany after WWII.

Right, and in older cars you can end up hitting a pedestrian that would totally be preventable with this technology.

For some reason I feel like this will save a lot more lives than it will take away.

I don't think that any of the Tesla owners really thought their cutting-edge cars would be worry free. They have the same expectations as someone who uses beta software, and in exchange they get to experience an amazing car. Also don't forget that we have had problems like this with mechanical devices in cars, too. Slowly, we started using regulation and other tools to mandate additional safety checks (like the latch on your hood that prevents it from blowing back and smashing your windshield) to prevent those from happening. We're nowhere near anything like this for chips, really.

On the other hand, this is what the future is really like: in a few days or weeks, Tesla will release an update to fix this, and will probably deliver it wirelessly to every affected vehicle, and it won't be a problem any more. That's magic.

> They have the same expectations as someone who uses beta software,

Really? After paying $140,000? I think for that price many of them were expecting (quite reasonably) that their cars would be worry free.

Alternatively many companies and governments spend licensing fees magnitudes higher than that for software that works terribly if at all. It's nothing new in the technology sector just a bit new in a consumer device.

But ultimately I agree with you

It may happen, but it is not acceptable.

> Alternatively many companies and governments spend licensing fees magnitudes higher than that for software that works terribly if at all.

That's right, because the person making the purchase order doesn't really feel like they are spending their own money.

I think Tesla owners bought the car expecting a fully functioning vehicle. Running a long running public "beta" a la Gmail isn't an option when your customers have paid you tens of thousands of $.

Unfortunately hardware updates can't be delivered wirelessly.

Stock manipulating is a federal offence. Funny how unproven claims get picked up by the press and various outlets.

How hard is it to get 30 ppl to complain about model X?