Thirty Tesla crashes linked to assisted driving system under investigation in US | Hacker News Reader

Thirty Tesla crashes linked to assisted driving system under investigation in US

theguardian.com

95 pointsmindracer82 comments

I wish HN would apply the same rules to discussions about Tesla as they recently have for Bitcoin. It's always the same predictable comments about how Elon's a liar, AP is a scam and Tesla is overvalued and overhyped. Gets so boring.

What are the new rules around Bitcoin discussions?

Of course the same rules apply: https://news.ycombinator.com/newsguidelines.html.

And yes, the repetition is exceedingly tedious and pretty much dominates all threads on these topics—kind of like one big ball of mud.

I think there’s enough Teslas on the road now that I’m willing to accept that each could be an independent and reasonable incident, unless routine investigation shows otherwise.

I'm curious mostly how Tesla "self driving" crashes compare to other carmakers "driver assist" crashes.

Because those two types of systems have mostly the same capabilities and same legal position, neither are self-driving in any way. But Tesla markets them more aggressively which could lead to a different attitude and behavior by some owners that think they have "self driving" and pay much less attention.

Tesla's driver assist tech is much better than most other cars. People are excited about that, making the media report a lot about mostly bad incidents. Nobody cares about your Peugeot 208 "Lane keep assist" crashing you into a cliff, because nobody cares about your Peugeot 208 in the first place.

Just for your reference none of the major auto companies do mid-range vehicle ADAS in-house (Higher-end vehicles and Halo product vehicles maybe sometimes).

There is no Peugeot LDW, there is just a bunch of OEM suppliers (Bosch, TRW, Mobileye, Continental, etc.) who do that for auto companies.

So raining on Peugeot because it is somehow second tier is misleading and misinformed because the electronics you find in a Peugeot ADAS sys, you will equally find in all other mid-range vehicles in the market.

So one of those failing would be a pretty big deal because it would apply to many models across many auto brands.

Tesla had something like 31 Billion in revenue in 2020 and as an example Bosch had about 71 Billion in revenue for the same period. If there was a case to sue lawyers would be jumping on it.

Are you confirming what I said or? It's pretty well known car brands integrate and work on various preexisting parts platforms

Just sort of a thought experiment...

I wonder if non-crashes - situations where the car took evasive action - may have been logged and could also be investigated?

On the other hand, evasive actions may be higher because drivers cede more control to the car.

The overall stats like "miles driven per accident" are probably the closest you'll get to this. Obviously Tesla bangs the drum every time figures come out in its favour. There is a level of incompleteness in the data and also some gray areas around the methodology that Tesla uses. But the overall figures do tend to indicate that Teslas using Autopilot and/or active safety features are safer than the average for cars (based on NHTSA and Tesla's reporting on Autopilot and/or active safety feature use).

If Tesla released more detailed data and its methodology it would be more helpful to see how significant this analysis is.

The most literal possible interpretation of "move fast and break stuff."

Are there stats on incidents per kilometer/mile?

Thirty sounds a lot but then again, it might be a very tiny number compared to the mileage.

Although it also needs to look at when it decides to hand back control. Its all well and good saying its 'safer' because it makes less mistakes than a human driver would doing the same amount of miles but if it only is activatable in locations/conditions that are considerably easier to control then its somewhat incomparable.

This is often quoted that tesla has a lower accident rate than other vehicles.

Comparatively, how often do those vehicles cause deaths while directly taking control of the vehicle.

> This is often quoted that tesla has a lower accident rate than other vehicles.

In general Teslas have higher accident rates than comparable cars, partly because EVs have higher accidents rates in general, but even among EVs Teslas have the highest. This goes back to data from an insurance company in Norway if I remember correctly. It's very hard to get reliable data though, as Tesla has no incentive to share this kind of data. Obviously you can't draw any conclusion from this to Autopilot-related accidents.

Source, please! Would be very interested in the original data. Thx.

I don't think the original data is available.

Re Norwegian data: https://www.elbil24.no/nyheter/elbiler-bulker-mest/71498210

https://cleantechnica.com/2018/06/04/electric-car-insurance-...

Critical article re autopilot claims: https://jalopnik.com/feds-tesla-autosteer-safety-investigati...

The reason, at least in the Norwegian article is "big fast car that Norwegians are not used to" and not auto pilot at least.

As a comparison my former car, a Toyota Avensis, was 175 cm wide. My model X is 220 wide or something. That's almost half a meter more. Have yet to crash it though :)

EVs in general have higher crash rate, most likely due to the instant torque.

You have to control for the demographics of typical Tesla owners (richer and older - less prone to accident if they weren't using Autopilot), as well as the type of driving that Autopilot is typically used for (tilt towards highway/freeway) when you make that comparison.

https://www.tesladeaths.com/miles.html

Is this looking accurate?

Why? Is it okay for the automated system to make objectively bad decisions until it reaches the magic threshold of however many incidents the non-automated system had? In a rewrite, do we not look at bugs until they exceed the number of the original system?

I don't get this radical zero-covid-like mentality. At unbounded cost we can diminish all risk. How do people get to this radical/dystopian point of view that no trade-off should exist?

The point is that you are deliberately misinterpreting the original article and then trying to turn that stupidity into silencing the problem entirely. The original article doesn't relate this to other statistics at all, that was a fabrication of the grandparent.

Of course NHTSA needs to investigate each and every one of these crashes. These are automated systems failing and so the issue could be systemic. It's the same reason every rail or airplane crash is investigated.

100 people are killed on the road in the US every single day of the year.

30 sounds tiny to me.

So what? This is the NHTSA job.

In the future every single car accident will be investigated, like it is done with airplanes, that is the advantage of autonomous cars, they log everything before the accident.

That is what made airplanes so safe that the last accidents have been basically suicidal pilots:

You investigate what creates accidents, you learn and take actions so accidents do not happen again. Rinse and repeat.

Lots of the "bugs" they are going to find are not in the cars, but in the roads, like perpendicular road crossings (too common in the US), instead of roundabouts, or concrete medians with sudden changes and no visual clues.

It’s pretty far from true that all airplane accidents are investigated. Airliner accidents are, but plenty of minor light aircraft accidents aren’t investigated any more deeply than minor traffic accidents are.

Not if there's no obvious fault anyway. Small planes go down often because they are usually flown VFR, by less trained pilots, or as training aircraft.

For an airliner to be able to go down at all is a problem. For a small propeller-driven plane it's pretty expected, they're manual as hell, less stable, and less massive.

The big choice NHTSA has to make is when to investigate and what counts as a fault of system design failure/operation failure and when that will result in a safe/unsafe recall situation. Case one car encounters situation it should be able to respond to and doesn't would need an in depth investigation whereas case 2 car encounters a new situation cant handle would need a more theoretical analysis on should car be able to handle that or is it safe to continue . i.e baby in road car should handle but they train model to ignore things like plastic cups/ plastic bags or any other "safe" road debris. but should it magically handle baby inside a plastic bag on the road. NHTSA will have to at some point make those calls on it will never be perfect but where is the line of good enough

It's not surprising. Tesla marketing is basically a scam. The tech is great, but Tesla always overpromise, underdeliver, and lie.

The worst part is that few people seems to care or understand that it's BS. Take the latest launch as an example. The car is stated to do 0-60 in 1.99 seconds. Does it do that? No, not even close. It's also supposed to know if you want to go forward or back, and people are surprised when it doesn't do that.

And then comes the most dangerous part: Autopilot and full self driving. Tesla cars has over multiple years promised that you can rent out your car as a robotaxi soon if you get the self driving package. Has that happened? No. Will it any time soon? No.

Not only is people surprised when self driving doesn't work, but it also has fatal consequences.

Heck, even I who have no experience or insights into Tesla can much more accurately than Elon tell people what Tesla will and won't be able to do. And I do of course think Elon knows more, but he chooses to trick people to buy Tesla.

This needs to stop...

This is what's so strange about US. There are laws with very hefty fines about false advertising, especially when human life and health is involved, this is true for EU and most other locations. Try to remember how much VW paid for their diesel emission scam. It was almost 3 billion dollars. And Tesla...

Also remember that GM paid a fraction of that for the ignition switch scandal [0] where GM management intentionally hid the issue which directly lead to at least 120 deaths, and that was just the official count.

The US has little interest in hitting home companies in any way. Preventing the false advertising would impact the value of Tesla because the FSD claims were an integral part of Tesla's branding and success. Some countries in Europa, like Germany, blocked Tesla from using misleading claims in their advertising like suggesting anything about self driving [1]. Of course this is usually dismissed as Germany protecting its auto industry. But the reality is that the claims Tesla made consistently misleading and made to suggest the cars is able to self-drive and it's only regulations that prevent it from doing this.

[0] https://en.wikipedia.org/wiki/General_Motors_ignition_switch...

[1] https://www.nytimes.com/2020/07/14/business/tesla-autopilot-...

I feel that there's almost 'Protect Tesla at any cost' order, even though there clearly isn't one. I have this heavy feeling that should Tesla crash and burn, this is going to be a major hit not only for Nasdaq or S&P, it can lead to major disillusionment with what tech sector really is and what it can and can not do for humanity in general and for US as a country specifically.

It's a very valuable company that's just taken off and for the first time in a long time gives the US an edge in the global auto industry. Shooting it in the wing now, even if it's the legal duty, gives everyone else an advantage. So blind eye it is.

Tesla has no reason to crash and burn but without a real or perceived edge over the competition it wouldn't thrive. And the self driving claim is a hefty part of that edge.

Yeah, but this is a very US-centric view. China is (arguably) far ahead of US both in EV technology in general and Tesla specifically. It's hard to separate personality cult of Elon Musk from Tesla, but for many outside US, there's absolutely nothing special about Tesla except for the valuation part.

I am completely oblivious as to how Tesla is seen in China. But plenty of the hype Tesla generated in Europe some years ago had FSD as a backbone (maybe a bit of ludicrous mode too), it was an easy way to cover generic EV drawbacks. I'm not saying that this is why people bought Teslas, rather that that's the reason people spoke about Teslas. If everywhere you look you see a discussion about how awesome one feature of a product is you're slowly and unconsciously going to start assuming the whole package must be awesome.

Tesla not only got "infinite" marketing and PR points by claiming "FSD", they even got money from people who are still waiting for that functionality. Those who expected driver assists are happy. Those who expected the claimed FSD are certainly not.

Today with competing companies like Volvo announcing L4 cars in 2022, the generic FSD claims are nowhere nearly as valuable as a few years ago.

> It's a very valuable company that's just taken off and for the first time in a long time gives the US an edge in the global auto industry. Shooting it in the wing now, even if it's the legal duty, gives everyone else an advantage. So blind eye it is.

Reminds me of Wirecard, which imploded one year ago. Everyone, from regulation agencies over banks to politics - including Chancellor Merkel and her advisors - wanted Wirecard to succeed at all costs. In the latest hours of Wirecard, after the "we're missing two billion euros" statement, there even was the idea to pump hundreds of millions of euros taxpayer money to prevent its collapse.

The principle of defending your superstars applies everywhere, albeit the US is doing it to a far greater degree compared to others simply because they're in a better position to do it.

But Tesla and Wirecard are in no way in the same league. Tesla is actually delivering cars, batteries, building factories, and developing tech. Wirecard on the other hand was a hot air balloon, at least after a point.

Having looked at a lot of the issues surrounding Tesla, I would strongly disagree with that statement. I think it's mostly informed by hindsight. While it remains speculative it is certainly possible that Tesla is in the same league as Wirecard, and maybe Enron.

People are easily thrown off by the fact that Tesla is successfully delivering a physical product and that the product is excellent in many aspects (driving UX, charger network, software integration, EV evangelism etc).

However once you look closer, a whole host of parallels emerge: 'Very aggressive accounting' (non-GAAP, 20% accounts receivables, junior CFO, warranty repairs as goodwill etc), business is not profitable (revenues stem from tax credits not cars), bail out/enrichment of Musk family (SolarCity), securities fraud (funding secured), retaliation against whistleblowers, silencing/threatening critics and journalists, legal record, occupational health violations (injured workers not being treated, Covid), workers at the factories not being allowed to speak to visitors, selling something that doesn't exist (FSD), environmental concerns at Fremont, ...

It's a different business than Wirecard for sure, and only time will tell, but the red flags are all there.

> retaliation against whistleblowers, silencing/threatening critics and journalists, legal record, occupational health violations (injured workers not being treated, Covid), workers at the factories not being allowed to speak to visitors, [...] environmental concerns at Fremont

Unfortunately, these issues - while perfectly legitimate to name them! - are extremely common throughout the entire US corporate and government culture. Exploitation of all kinds is pervasive.

> I think it's mostly informed by hindsight

Well... yes, simply because hindsight usually provides evidence while foresight provides guidance. So while I agree with your points, and as much as I dislike Musk (a criminal who's crimes are overlooked because of his status), your case pivots on the assumption that they're certainly faking it, just weren't caught yet. And you could be right. But so could Scientology.

The fact that Tesla is delivering what we can easily verify with our eyes at the very least puts them in a vastly different league from Wirecard. Tesla could be a unprofitable business like so many others. But Wirecard wasn't just a failing business, it was massive scale fraud and nothing it left behind still proves useful and can provide value now.

That is simply not true. The majority of Wirecard has been sold in the insolvency proceedings, e.g. the payments platform is still in use today! (Interestingly still using the domain wirecard.de!)

The liquidation price was nowhere near its market value and their technology was nowhere near as innovative - both points that are also true for Tesla.

Again, don't be led astray by the fact that Tesla is making physical products that you can see.

On a different note I don't think of corporate fraud as something that brings you from 0 to 1 - it's more akin to cheating in sports. It will give you a competitive edge, an unfair advantage.

If you are interested there is an interview with the CFO of Enron about legal fraud on Youtube - highly fascinating.

Tesla may be delivering cars and other physical properties but that doesn't make them immune to fraud, be it from customers or investors. This is a $500 billion company that has made some hefty promises toward both groups that it has repeatedly fallen short on.

The behavior of the investors in Tesla is extremely scary. Tesla bulls have organized some large intimidation campaigns against anyone whose tried to question why this company is worth $500 billion.

https://www.motortrend.com/cars/tesla/model-s/2021/2022-tesl...

https://www.topgear.com/car-news/big-reads/it-mad-bad-840bhp...

So Model S Plaid can do 0-60 in 1.98s provided sticky VHT substance on the track. Same stuff used for other drag racers, super- and hypercars to get the required road friction.

https://en.wikipedia.org/wiki/PJ1_TrackBite

Without VHT:

"Once ready and in its cheetah stance, the Model S Plaid's launch is drama-free, even without the added advantage of VHT. The electric car accelerates from 0 to 60 mph in just 2.07 seconds, more than 0.2 second quicker than our previous record holder."

Based on this, what NorwegianDude suggested that it doesn't come even near to that acceleration is just false.

Same article, first thing: "we can confirm the 2.0-second 0-60 barrier remains unbroken. Unless, that is, you write your own rules."

"Plaid accelerates from 0-60 mph in 2.28 seconds"

So which one is it? Defenitly not 0-60 mph in 1.98.

0.29 seconds off is a huge amount. It's exponentially harder the lower you go, so I'd say that it's not even close.

Roll-out is standard my dude. So is testing at drag strips. Next you're going to point out that a tesla accelerating on the beach gets to 60mph in 11 seconds and therefore Elon is lying.

https://www.caranddriver.com/features/a30085446/acceleration...

https://www.motortrend.com/news/motor-trend-testing/

Many magazines use rollout and different surfaces. But pretty much every car manufacturer displays their acceleration numbers without excluding rollout or using a prepped surface. The dodge deamon is the exception here because it is designed for drag racing.

Tesla is interesting because they selectively include or exclude the rollout in their acceleration numbers. For their slower (but still very fast) models they include the rollout, and for their faster ones they exclude it, artificially increasing the gap between their models. Not sure why.

Engineering Explained did a good video on this, specifically mentioning rollout:

> […] No, Tesla's Model S Plaid won't hit 60 mph in under 2 seconds. In this video we'll discuss rollout, Tesla's actual 0-60 time, 0-60 theory, and what today's theoretical limit on 0-60 mph is. Watch to learn more!

* https://www.youtube.com/watch?v=i7yigpPSu_o