Polaris OBark

From ArsTechnica today: Huge Tesla leak reveals thousands of safety concerns, privacy problems

I-Like-To-Bike

Verge also reported on this subject:
Tesla leak reportedly shows thousands of Full Self-Driving safety complaints

The most revealing news is the reporting of Tesla' internal guidance on responding to the customer complaints of safety issues and problems.
Extract:
Presumably this must be fake news since I didn't personally experience these incidents.

PeteHski

Reports of sudden unintended acceleration are not unusual for car manufacturers and are almost universally driver error. But when lawyers get involved it becomes farcical, like with the infamous Toyota recall. I’ve never experienced this issue in any Tesla I’ve driven or any other car.

Phantom braking however is a real issue for Tesla FSD, although it often gets blown out of all proportion. I’ve experienced this a few times and it is annoying when it happens, but not a big deal. The current version of AP I use is rarely affected and it doesn’t brake hard when it happens. It’s like lifting off the accelerator for a few seconds rather than slamming on the brakes. It usually happens when you are about to overtake a large truck on a sweeping motorway bend. You just over-ride the accelerator and only lose maybe 5 mph in the process. It doesn’t happen often enough to really worry about, but some owners seem spooked by it. It’s still a work in progress for Tesla to fully solve.

FSD will inevitably generate loads of complaints. It’s a new experience for most owners and there are loads of edge case issues to resolve. Expectations are high and it doesn’t fully deliver. But that doesn’t make it a bad system. There is a lot to like about FSD/AP and I find it really useful in daily driving. But my wife doesn’t use AP at all as she just can’t get her head around it, although it’s still her favourite car by a long way.

Polaris OBark

The gas light burns strong with this one.

livedarklions

If Tesla really wants us to believe that the number of complaints is distorting the issue, attempting to keep the data under wraps is not exactly reassuring. Quite the opposite, it makes it look very much like they don't want people to know about the extent of the problems.

I honestly don't know what the Toyota recall has to do with this and what lawyer misconduct it's supposed to represent.

Whether the acceleration is due to a software glitch or just reflects a normal rate of driver error is exactly the kind of question the data would be useful in answering. Assuming it's just driver error so we don't need to worry about the data is just begging the question. The fact that you haven't personally encountered the problem tells us absolutely nothing about it. Comparing reports when it does occur, however, might indicate commonalities that lead to the discovery of the source of the problem. One can't help but notice that Tesla's actions are seemingly designed to prevent people from making those comparisons.

Sorry, but Tesla's reputation for dishonesty and hype is absolutely horrible, and this smells rotten.

jon c. 

Had that happen in a rental car (not a Tesla). Someone in front of me slowed and turned right. I slowed but didn't brake as it wasn't necessary. The driver had already cleared my path when the warning lights and buzzers sounded. The car seemed to brake a bit, but advised me to do the same. Which I was doing because initially I had no idea why the car was yelling at me.

I turned the feature off for the rest of the rental. The science is good, but the science is young. It will get better but in the interim I can do without it.

Polaris OBark

The idea is that claims like the following were simply fabricated, and that there has never been a real problem with Toyota and sudden acceleration:

https://www.latimes.com/archives/la-...r03-story.html

The CHP cop and his family who were killed is simply a product of mass-hysteria (and I suppose perhaps the equally plausible idea that CHP officers have little experience driving.)

PeteHski

The idea was that almost all such claims are driver error and that all manufacturers routinely face such claims. It’s not some unique Tesla issue. Most of them are highly dubious car park shunts caused by driver brain farts. A tiny number may well be genuine issues, but the tech involved here is very mature and proven over billions of miles.

Also the idea that it would be caused by a software glitch is far less likely than a physical failure e.g. a floor mat or stone getting stuck under the pedal or a sticking throttle on a conventional engine.

The case you cited is almost identical to a UK case involving a different manufacturer (I think BMW from memory, but definitely not Lexus). The guy was driving on the motorway late at night and the accelerator apparently stuck fully open and he apparently had no brakes either. He drove for many miles before inevitably crashing and had called the emergency services for help. I don’t believe they found any fault with either the accelerator or the brakes in that case.

These simultaneous accelerator and brake failures are incredibly rare and hard to prove when human operation is involved. But they make sensational click bait media hype. Have you got any more for us to get scared about?

GhostRider62

Statie pulled me over doing 128 mph and explained the throttle was open w/o my foot on the pedal.

He looks at the registration and says, nice try. This is a Mercedes, not an Audi.

livedarklions

The Yoyota thing has nothing to do with the thread topic. I have no idea why you brought it up. Tesla is deliberately hiding its data. Are you defending that?

If the data is showing driver error, why would they hide it?

Since this is actually a cycling forum and notva driving forum, do you know of any data that the varying degrees of automation already employed are actually improving pedestrian/cyclist risk? Serious question, I have no idea.

PeteHski

I brought up the issue of unexpected acceleration simply because it was cited in the link about Tesla data. I’m not defending Tesla in hiding data, although I can imagine various reasons why they might not want their data to be freely available. Given the history, I am not particularly surprised they are guarded.

Your second question is far more interesting for me. Cyclists and pedestrians are actively tracked in FSD/AP, displayed on the dash and the car does actively brake to avoid collision. Simple scenario is when you are driving along with AP active and there is a cyclist in front. AP detects the cyclist and gently brakes to follow behind. This works very reliably, but the system has to be active. A lot of Tesla drivers simply don’t drive on AP/FSD and in that case it is like any other car, but you will get visible and audio warnings and emergency collision braking if you ignore them. For me that’s a lot better than having nothing at all. The usual counter-argument is that it may make you more complacent, but I don’t buy into that. Enough cyclists get plowed into by distracted drivers to make these systems invaluable. The times I’ve found it most useful are in low visibility situations, especially low dawn/dusk sun, which we all know can be deadly for cyclists.

It would be interesting to hear from any other Tesla drivers who think these systems do NOT improve their road safety to some extent. I’m a fan of AP, but I don’t think FSD adds much to it in terms of safety and fully autonomous driving while watching Netflix is still a long way from reality. Just too many complex edge cases, inconsistent road regulations and unpredictable human drivers. But that doesn’t mean we should ignore the safety features of semi-autonomous systems like AP/EAP/FSD and equivalents from other manufacturers. They are improving all the time and can only get better.

livedarklions

I'm going to say that I'm not optimistic about FSD actually improving pedestrian/cyclist safety if the designers think that the OP video is an example of how they want the AI to function. Taking illegal actions just because you know you'll likely get away with it is not the human behavior we want AI to emulate. Even if there is no chance that the vehicle was going to hit the pedestrian, it increases the chances that the car behind the AI will hiit the pedestrian. That's one of the major reasons why the law is stop and wait at the crosswalk, not race through. Drivers take cues from each other's behavior, especially when the position of that car is directly between you and the crosswalk. If Tesla engineers want to go to the lowest common denominator of risk assessment on determining whether to take an illegal action, that doesn't bode well for the direction of future developments. This looks very much like an over-correction to deal with the "frozen by indecision" problem.

PeteHski

This is why I don't believe AI-controlled cars can successfully mix with human-controlled cars. But there is no doubt in my mind that FSD drives more conservatively than the average human driver. There is no road rage, no running red lights, no aggressive overtaking to gain a car length, no swerving across lanes when distracted and no hatred toward cyclists! But it does still need a human to manage all the edge cases. It is an advanced driver aid, not true FSD as the name implies.

I've seen enough cars run straight through pedestrian crossings with kids waiting to cross with their mums to have little respect for human drivers as a whole. As a pedestrian I certainly don't rely on them to stop when I'm at a crossing. Unfortunately, we humans are a pretty low benchmark in that respect.

79pmooney

I posted in Just the Headline about a Tesla not stopping for an off-road tree.

Hypno Toad

I'm always a little shocked (but not really) about people's fear of automation/technology. People driving cars is NOT SAFE! People look at phones, get tired, get mad, get drunk, speed, fail to stop/yield, .... These stats make it clear, people driving cars is not working great:

In 2020, 6,516 pedestrians died in traffic crashes, and 938 cyclists died. Those figures were about a 4% increase in pedestrian and cyclist deaths from the previous year. This increase occurred despite a 12% decrease in vehicle miles traveled in 2020. [source: [url]https://usafacts.org/articles/how-many-pedestrians-and-cyclists-are-killed-by-cars-in-america/]

I'm not in love with Tesla and I'm not all about autonomous vehicles. Our personal car (Volvo) has many of the features in a Tesla and I don't have interest in using the auto-driving function. OTOH the collision avoidancesystem stopped me for hitting a deer in the earlier morning light, I was too late seeing the deer to react, but the car stopped quick enough to avoid a collision and saving me from some frontend damage (I'm sure Bambi is happier without a broken leg too).

Another counterpoint, try a car without any automation, I've got my grandfather's 1937 Chrysler Royal. Driving that vehicle takes 100% of your attention, with manual steering and a manual transition, you can even think about checking your phone while driving. Maybe we need to legally require all vehicles go back to manual transition, manual steering, etc,

livedarklions

I think this allegedly irrational fear of automation is a bit overstated on your part. I'm not seeing anyone arguing for the removal of collision avoidance or anything like that, but there really is no solid evidence that self-driving cars at the level of AI we are at are going to be as safe or safer than human drivers. The OP video basically shows an AI vehicle essentially simulating the driving habits of a distracted driver--actually choosing to ignore the required stop for a pedestrian in the crosswalk and that being celebrated because that's how urban human drivers "really drive". Sorry, but if the AI is incorporating a strategy of illegally risking close calls with pedestrians (and by implication cyclists), I don't think that skepticism about the likelihood it will actually make us as safe or safer is anything like calling for the removal of power steering. The argument against attended self-driving is that it invites the human driver to become completely inattentive as the car will still keep going along the route. None of the nature of collision avoidance, automatic transmission or power steering invites anything of the kind. Quite the contrary, I think it could be argued that manual transmission is actually distracting the driver from knowing what's in front of them.

Hypno Toad

Nothing is overstated, it is just my opinion. My post clearly states both sides of the issue. I did not say that AI is better than human drivers, there is far too little data available; but I'm saying people driving cars have a proven-record of being inattentive and dangerous. In my humble opinion, autonomous vehicles would have to be terribly designed and tested to do worse that human drivers.

Also, if driving a manual transmission vehicle it taking your attention away from what's in front of you, you are doing it wrong, you do not need to look at the clutch or shifter. So it appears that "it could be argued" is just you creating a silly argument for no reason other than to argue.

I am very sorry, I don't know what I did to you, but seriously ... WTAF?!?

livedarklions

Hey, I just posted in the other thread that you can handle reasonable disagreement, but maybe I need to retract that....

I was reacting to the first sentence of your post, you seem to be saying that you are shocked (your word) that people have a fear of automation. Then you equate full self driving with a bunch of other forms of automation that are very different in kind and degree. Sorry, but you can't claim in one post that manual transmission requires your attention but then go on to the next post and claim that that attention won't come at the cost of attending to some other task (i.e., paying attention to what' in front of you). One thing everyone who has studied human attention knows is that it is a finite resource, operating a clutch and stick may not require your eyes off the road, but it very well may interfere with your ability to process the information. .Seriously, if it's a matter of actually making people safer drivers, why is manual transmission practically extinct in modern cars on the road?

I don't think I was wrong to point out that full self driving is something that could actually make road safety worse. You want to consider fear of that "shocking", fine, but I think that's way over the top to compare that concern with an antipathy towards power steering.

No idea why you think that's a personal attack warranting a WTAF, but WTAF back at you on that accord.

Polaris OBark

Your reactions to stuff lately have been way over the top. Maybe dial it down a little. This is bike forums, not divorce court.

79pmooney

You are aware the Texas and other states have banned (or will soon ban) all transitions, manual or otherwise.

livedarklions

Still lying about having me on ignore, huh?

I don't take tone advice from you, cheap-shot guy.

79pmooney

You are mixing two completely separate things in the bolded type. 1) People are lazy. Once automatic transmissions became good enough to feel like a quality drive, wasn't hard at all for manufacturers to push it and (almost) everybody to buy into it. That is all about bottom line and marketing. Not needing to make two versions of the same car. 2) Safety - every study I've ever seen shows manual transmission drivers being more alert and better drivers.

Hypno Toad

LOL!

There's a good reason I'm not a copy editor ... I've long said: "if there isn't a typo/misspelling or edit on my post, somebody hacked my account"

Hypno Toad

We're back-and-forth on different threads, and there's spillover that's outta sequence. I'm sorry for creating any confusion. Let's just move on, thanks.

PeteHski

Is that simply because drivers of manual transmission cars are far more likely to be keen motoring enthusiasts rather than anything to do with the actual transmission? How did those studies account for different driver skill/experience levels? I drive both manual and automatic cars and I don't feel any difference in terms of safety or attention to the road.

Although automatics do allow drivers to "multi-task" a bit easier i.e. mess with their mobiles etc if they are being stupid - so I guess there is that. You could argue that to be the case with semi-autonomous driving too, but the driver requires less manual input. Then it becomes a question of whether or not the semi-autonomous system makes the driver even more inattentive vs how much the system compensates for any lack of attention. In my experience of Tesla AP/FSD you end up paying at least the same (certainly more initially!) attention to what is happening around the car and less on physically driving the car i.e. steering, acceleration/braking.

As these systems approach true self-driving (with no driver input or attention required) then the game changes dramatically. This is why true FSD is still relatively far away. The current systems are simply not foolproof enough for the myriad of complex and inconsistent road scenarios, especially when mixing it up with erratic, unpredictable human drivers! They can't even avoid large pot-holes in the road at this point.

This thread topic about the way FSD deals with pedestrian crossings is just one specific example of why there still needs to be a human making the calls. If you are required by local law to stop then you just brake and stop like any other car has to. None of the manual cars will stop either unless the driver chooses to stop. There are loads of other edge cases in which a driver may need to over-ride FSD. But that doesn't necessarily make it more dangerous than a car without any form of driver aid. Cars routinely run through pedestrian crossings and actually mow people down on occasion. Would FSD or similar systems prevent that happening in more cases than it actually caused an accident? I think so.