The catastrophe of FSD and erosion of trust in Tesla

[Knightshade]

That article is....not great...


for example "the automaker plans to roll out beta testing of Autopilot in Europe"

AP has been widely available in Europe for many many years now of course. They mean the new city streets FSDBeta stuff which is vastly different.


And it doesn't really set any precedent at all, since it'll be appealed with a decent chance of being overturned. And it isn't a US court, so it would hold no sway here even if upheld.


Remember when some rando lower german court said Tesla couldn't advertise AP and FSD anymore because the court didn't understand it?

Guess what happened? Nothing. They still advertise it, using the same wording as then, years later- with the local court ruling having no effect on anything.

 

[2101Guy]

when do we get "NYC to LA with no human intervention".

Thats what the CEO promised. Thats what we simply want before end of this year.

Oh, and Robotaxi service where FSD enabled car can generate "many multiples of $100k in revenue to the owner".

If thats not in place by Dec 31, 2022? It failed.

Next.

 

[2101Guy]

as it should be because everyone who bought FSD knew exactly what they were getting, a promise for beta software.

No, some were expecting to get, what the CEO promised multiple times in writing....

Sometimes when a company and/or their leader promises something and doesnt deliver? Even if it wasnt stated as part of a "contract"? Guess what can happen...

And if we think CEO's words cant be legally binding in some instances:

"Tesla’s lawyers claimed the the car is expected to last as long as comparable ICE cars, according to them around 250oookm, but the court decided to calculate the usage fee with a vehicle life of 800000km, citing a public statement by Elon Musk, and greatly reducing the deduction".

 

[DukeofURL]

The money isn't really the point, they could probably get all the money back just by selling the car in this current market. It is the precedent being set to hold manufactures more accountable for the software and its advertised functionality. This is not Microsoft Windows, it's a vehicle and people expect *sugar* to work in a vehicle. I hope the aviation industry does more testing with their autopilot.

This post right here is the entire (recurring) FSD argument across all of TMC in a nutshell.

There are two camps of people on TMC. Those who expect *sugar* to work in a vehicle 100% as described all the time, as this is a reasonable expectation for such an expensive purchase (that is intended to last you 10 years of high reliability). In addition, the automotive industry has built expectations over the past century or so, and all manufacturers tend to uphold that expectation. That expectation is in stark contrast to Microsoft Windows which has a much shorter useful life and a far lower expectation of reliability, so much so that it's the butt of jokes about it.

The other camp here expects "beta software" to be just that; generally unreliable and ever evolving. Tesla is weird as they are a automotive manufacturer, and by all accounts should be held to the expectations of such, but they are also a Silicon Valley company and are more about "disruption" and "moving fast and breaking things". This leads to "beta culture" where new features tend to be prioritized over properly working all and every bug out to a polished final release. They ultimately get there, but for instance, how long was Gmail in beta before it was finally considered "ready"?

At the end of the day, I knew I was buying into something that's not even half-baked yet. Personally, I think it's fun to follow the progress. I don't expect it to be done tomorrow, but I do expect it to be done eventually, and I have to say I am greatly enjoying being able to actively engaged in this bleeding edge technology - but make no mistake, I am acutely aware and waiting for the car to try to kill me (I also have a few other potentially deadly hobbies too, lol). On the other side, I have a friend who likes to keep his feet more firmly planted on terra firma, and has very little tolerance for when anything deviates.

Different strokes for different folks. It's pointless to argue across these lines, because they're two different (and mutually exclusive) preferences.

 

[raptor5244]

This post right here is the entire (recurring) FSD argument across all of TMC in a nutshell.

There are two camps of people on TMC. Those who expect *sugar* to work in a vehicle 100% as described all the time, as this is a reasonable expectation for such an expensive purchase (that is intended to last you 10 years of high reliability). In addition, the automotive industry has built expectations over the past century or so, and all manufacturers tend to uphold that expectation. That expectation is in stark contrast to Microsoft Windows which has a much shorter useful life and a far lower expectation of reliability, so much so that it's the butt of jokes about it.

The other camp here expects "beta software" to be just that; generally unreliable and ever evolving. Tesla is weird as they are a automotive manufacturer, and by all accounts should be held to the expectations of such, but they are also a Silicon Valley company and are more about "disruption" and "moving fast and breaking things". This leads to "beta culture" where new features tend to be prioritized over properly working all and every bug out to a polished final release. They ultimately get there, but for instance, how long was Gmail in beta before it was finally considered "ready"?

At the end of the day, I knew I was buying into something that's not even half-baked yet. Personally, I think it's fun to follow the progress. I don't expect it to be done tomorrow, but I do expect it to be done eventually, and I have to say I am greatly enjoying being able to actively engaged in this bleeding edge technology - but make no mistake, I am acutely aware and waiting for the car to try to kill me (I also have a few other potentially deadly hobbies too, lol). On the other side, I have a friend who likes to keep his feet more firmly planted on terra firma, and has very little tolerance for when anything deviates.

Different strokes for different folks. It's pointless to argue across these lines, because they're two different (and mutually exclusive) preferences.

Well said, it is about setting proper expectations. I think a lot of folks that are new to Tesla get a quick lesson of what you described. You pay $5k, $7k, $10k and now $12k for this software with the "expectation" that the car will FSD in a "reasonable" amount of time. IMO, Tesla should be more forthcoming and set more realistic expectations since we are dealing with motor vehicle and an expensive piece of software, not a Happy Meal or Diaper Genie. I think they could get away with this when they are relatively small but once you start cranking out over a million vehicles a year then you are going to get some backlash over this kind of thing.

 

[Goose66]

"There are two camps of people on TMC. Those who expect *sugar* to work in a vehicle 100% as described all the time... [and those that expect] "beta software" to be just that; generally unreliable and ever evolving."

I think that this dramatically oversimplifies the issue. I understood when I purchased FSD Beta in 2018 that it was something that wouldn't be "feature complete" for some time, things would evolve, and "beta" testing (not a beta test) would occur. I also haven't ever believed it would be L5 - L4 at best, but likely L3 by design. However, I didn't think it would be "generally unreliable and ever evolving" forever! I thought we would eventually get a workable FSD product that could, under ideal conditions, drive you from your driveway to work, drop you off at your front door, and then go park itself. And I didn't think this because I was stupid, naive, didn't understand the "software development process", or any of that mess that's frequently thrown at critics of FSD progress. I thought this because of the messaging I was getting from Tesla's website and Tesla's CEO both before and after my purchase. Remember these:

• Nvidia Conference 2015 - "It's not something I think is very difficult. To do autonomous driving that is to a degree much safer than a person, is much easier than people think... I almost view it like a solved problem," estimating "complete autonomy" by 2018.
• October 2016 - "all new vehicles come with the necessary sensors and computing hardware for future full self-driving"
• October 2016 - Elon says "by the end of 2017, a Tesla will be able to drive from New York City to Los Angeles without the driver having to do anything."
• February 2018 - "The upcoming autonomous coast-to-coast drive will showcase a major leap forward for our self-driving technology”
• February 2019 - Elon says "Tesla's Full Self Driving capability will be "feature complete" by the end of 2019. "Meaning the car will be able to find you in a parking lot, pick you up and take you all the way to your destination without an intervention. This year. I would say I am of certain of that, that is not a question mark." Sometime after that, Tesla's order page for "Full Self-Driving Capability" stated "Coming later this year"
• Autonomy Day 2019 - Musk says Full Level 5 Autonomy by end of 2019 and estimated that by the middle of 2020, Tesla’s autonomous system will have improved to the point where drivers will not have to pay attention to the road. “We will have more than one million robotaxis on the road,” Musk said. “A year from now, we’ll have over a million cars with full self-driving, software... everything... These cars will be Level 5 autonomy with no geofence, which is a fancy way of saying they will be capable of driving themselves anywhere on the planet, under all possible conditions, with no limitations."
• etc., etc., etc.

I think there is definitely at least one more camp of people that understood the software would initially be "beta" but would eventually culminate with delivery of an at least L3 system in some reasonable timeframe. That was the expectation set by Tesla and Elon, and that is is what we haven't seen.

 

[S4WRXTTCS]

At the end of the day, I knew I was buying into something that's not even half-baked yet. Personally, I think it's fun to follow the progress. I don't expect it to be done tomorrow, but I do expect it to be done eventually, and I have to say I am greatly enjoying being able to actively engaged in this bleeding edge technology - but make no mistake, I am acutely aware and waiting for the car to try to kill me (I also have a few other potentially deadly hobbies too, lol). On the other side, I have a friend who likes to keep his feet more firmly planted on terra firma, and has very little tolerance for when anything deviates.

Much of this is the same reason I bought into it.

It was only $3K which isn't bad for a once in a lifetime kind of opportunity.

When it really comes down to it I have two main disappointments.

EAP features -> Simply hasn't worked anywhere close to 100%, and I think even Tesla knows most of the features that were under the EAP umbrella kinda sucks. That they needed the HW/SW that was under the FSD Banner so that's why they moved them over in 2019.

FSD -> The main gripe is the lack of enjoyment. The enjoyment is hampered because of lack of customer engagement from Tesla. I didn't expect to talk to Engineers at Tesla, but I did expect to be able to report things and to get things fixed. In a lot of ways people of my nature (to test and test again) are hugely beneficial to a manufacture, but we're not going to go through with that if it feels like were sending things into the ether. I can't think of any other company who I've done business with having less interest in my feedback than Tesla.

 

[drtimhill]

The problem with so many people is they try to make a blanket "it works awesome" or "it sucks and tried to kill me" statements. Its way to complex for that. In many scenarios, it works perfectly and is much safer than humans. In others it fails miserably and will crash unless the driver intervenes. I've been testing since 10.2. The number of fails has gotten less but it still has a ways to go. That said, it provides value to me RIGHT NOW for the many scenarios where it works great.

In my experience, it basically drives perfectly when it can stay on the same road. It can wind through turns, go through stop lights and stop signs and merge and pass other cars just fine. It can do this for miles in all types of traffic, day or night and usually in the rain. It can do this on highways or back country roads or 2 lane surface streets. When it has to exit from that road to the right or left, it starts to struggle - usually because it is WAY to cautious and slows so much it will annoy people. If its a simple exit with a gradual curve to the next road with it's own dedicated merge lane, great. FSD will nail it. If its an unprotected left turn through a median across 3 lanes of traffic at rush hour, it doesn't stand a chance. Obviously, there is an entire spectrum of scenarios between these two extremes and I've learned where I think it will be successful and where I should take over. 99% of my disengagements are because of this. If no one is around, I'll let it do its thing and it will usually get it eventually. Only about 1% of the time do I disengage because I think it may crash or do something really bad and even in those cases, I've only had 1 time when I think it would have caused serious damage. Usually, it's because I'm afraid it will hit a curb and mess up my expensive rims .

I don't have 0 disengagement drives but its not because the car can't move me from point A to B safely. Its because it won't do it the way I would do it yet. With each new version, I'm seeing progress in the right direction.

Well said. Most of my disengagements fall into the “ohh that was close” bucket when I suspect the car was doing ok but was taking it finer than I was comfy with and I didn’t want to risk a ding on the car. The others are mostly about the car being so cautious it annoys other drivers behind me, such as waiting on every possible pedestrian even when they don’t have the right of way, or being very cautious on an unprotected left.

 

[drtimhill]

FSD -> The main gripe is the lack of enjoyment. The enjoyment is hampered because of lack of customer engagement from Tesla. I didn't expect to talk to Engineers at Tesla, but I did expect to be able to report things and to get things fixed. In a lot of ways people of my nature (to test and test again) are hugely beneficial to a manufacture, but we're not going to go through with that if it feels like were sending things into the ether. I can't think of any other company who I've done business with having less interest in my feedback than Tesla.

I get this, and agree it would be nice to feel "noticed", though trust me Microsoft beta testing was no better in that regard. However, I think it's important to realize that FSD is a different class of project. With (say) Windows, you would notice a bug, report it, and hopefully notice the bug had gone away in an update (or not!). It's a nice feeling to see a bug that you found get fixed.

But with FSD, things are more .. well .. "spongy". When testers find a problem, that scenario can (and probably is) incorporated into the large corpus of training for the NN, but it can take time (and many more reports) for any change/improvement to become apparent. Ultimately, the whole process is more organic .. the car improves gradually over time, but its tricky to see much change on a release-to-release basis. I have a private video of my car on 10.4, and its interesting to watch compared to 10.11 .. steering wheel jerking all over the place even when stopped, mid-course corrections like crazy when making left turns, curious lane positions on right turns, un-natural speedups and slow-downs when dealing with red lights etc. It's come a long way.

My son is mildly autistic, and initially everyone was in a panic when he was diagnosed. But I looked at him, and watched, and realized he was learning and improving at his own pace. Sure, it wasn't the pace of other kids, but he never plateaud, and is now a great guy with a great job and life. While I'm not calling the car autistic (!), there is an analogy here: FSD continues to improve all the time at its own pace. Is this the pace people want? Nope, we all want stuff now. Is it as fast as Elon promised or wanted? Of course not! Should Elon do something about early adopters who feel shortchanged? Probably, but I would not hold my breath on that.

 

[t3sl4drvr]

1) people who paid for FSD and got software that did not match their expectations set by Elon’s vision

2) people who paid for FSD and expected beta that they can be part of. Excited to see how it develops over time.

Add

3) people who paid for FSD, but got nothing [the early buyers who did not even get the beta]

 

[macmee]

No, some were expecting to get, what the CEO promised multiple times in writing....

Sometimes when a company and/or their leader promises something and doesnt deliver? Even if it wasnt stated as part of a "contract"? Guess what can happen...

And if we think CEO's words cant be legally binding in some instances:

"Tesla’s lawyers claimed the the car is expected to last as long as comparable ICE cars, according to them around 250oookm, but the court decided to calculate the usage fee with a vehicle life of 800000km, citing a public statement by Elon Musk, and greatly reducing the deduction".

Tesla literally did deliver though on FSD, they just couldn't make the car drive itself in 1 year, they needed 5. Given the scope of the problem and the amount of disposable money we're able to spend on it as a hobby, I honestly don't see how anyone can be that mad about it.

 

[Ruffles]

1) people who paid for FSD and got software that did not match their expectations set by Elon’s vision

2) people who paid for FSD and expected beta that they can be part of. Excited to see how it develops over time.

Add

3) people who paid for FSD, but got nothing [the early buyers who did not even get the beta]

#3 does not exist. The FSD package has always included other little things like summon or lane change etc. that customers got. You can argue if those features alone are worth the price paid but it is not correct to say the spent money and got absolutely nothing for it.

 

[_Redshift_]

Tesla features graduate down, and not up.

Anything that works correctly is examined for ways to break it.

People are happy with autowipers, and the sensor is a few dollars? Obviously we can kill that by doing it ourselves in our Deep Rain Sense Neural Network.

The UI is too useful you say? Well how about Hide the button V11?

Preconditioning 30 minutes before Supercharging? That makes too much sense, lets have it precondition during the entire drive. Customers will love that.

Rolling Stops? Oh, geesh that's just too useful. Lets tells everyone we do that so the NHTSA forces us to turn it off.

Zigackly. These are some of the reasons is why my next car will not be a Tesla.

I have been in Hardware and Software development all my life, and I have over 26 years of experience.

FOREVER BETA is not something to be lauded, like this guy Raurele seems to misguidedly suggest.

It is something that is mocked in Software and Hardware circles. It means, you are just not good enough.

It is okay to be Beta for a while, (no hard and fast rule exactly how many years, it depends on the software complexity and criticality) but to be Beta for this long, for software that is considered a safety issue is mad.

Especially when people report regressions that can lead them into head-on collisions.

 

[S4WRXTTCS]

FOREVER BETA is not something to be lauded, like this guy Raurele seems to misguidedly suggest.

I think its a double edged sword.

On the one hand forever beta means things are constantly in flux, and can improve.

To me it simply means a feature is not static.

For the most part I wouldn't call any L2 feature from any manufacture anything other than beta. Sure it might be a more proven out beta, but the very nature of L2 is that the driver has to be ready for the car to do something dumb. It took Subaru multiple iterations of Eyesight to get it to where it is today. They never called it beta, but I imagine Eyesight 1.0 is terrible compared to the latest version. I don't believe Subaru owners can upgrade either. The adaptive cruise control in my Jeep doesn't phantom brake but has other ways of annoying me, but that's static. It just is what it is.

The double edge aspect of Tesla SW means it can also regress, and that's really what I find disappointing with Tesla.

It's so nice when things work, but they can't leave things alone. They have to monkey with things for who knows what reason, and then they fail to test the changes or whether owners even wanted them.

Take for example the AutoLights which used to turn the lights on when the wipers were going during the day. They even had a description of that behavior in the manual, but then in a follow up SW revision they removed the functionality and they changed the manual.

Why was that change made? Who decided to make all 3/Y owners look like idiots who can't use their lights in the rain during the day?

Who's piloting the ship?

 

[Knightshade]

Gmail was in beta for 5 years.

An email client. That launched decades after the invention of email.

But yeah, self driving car software still being in beta after 5 years is embarrassing. Or something.

 

[_Redshift_]

I think its a double edged sword.

On the one hand forever beta means things are constantly in flux, and can improve.

To me it simply means a feature is not static.

For the most part I wouldn't call any L2 feature from any manufacture anything other than beta. Sure it might be a more proven out beta, but the very nature of L2 is that the driver has to be ready for the car to do something dumb. It took Subaru multiple iterations of Eyesight to get it to where it is today. They never called it beta, but I imagine Eyesight 1.0 is terrible compared to the latest version. I don't believe Subaru owners can upgrade either. The adaptive cruise control in my Jeep doesn't phantom brake but has other ways of annoying me, but that's static. It just is what it is.

The double edge aspect of Tesla SW means it can also regress, and that's really what I find disappointing with Tesla.

It's so nice when things work, but they can't leave things alone. They have to monkey with things for who knows what reason, and then they fail to test the changes or whether owners even wanted them.

Take for example the AutoLights which used to turn the lights on when the wipers were going during the day. They even had a description of that behavior in the manual, but then in a follow up SW revision they removed the functionality and they changed the manual.

Why was that change made? Who decided to make all 3/Y owners look like idiots who can't use their lights in the rain during the day?

Who's piloting the ship?

You can have development happen all the time. You cannot have regression back to an older problem that was fixed, in a newer release. That should be the golden rule.

With Tesla, newer releases typically seem to be a mixed bag.

The reason is very simple. Not enough testing in-house.

Also, the problem itself is devilishly complicated. When Elon started promising all the things 5 years ago, I was silently wondering if he will EVER achieve those things.

I feel this problem will be solved, but it will take another 10 years. I’m talking L4 here.

To all those who ponied up $12K for a feature that will never be complete before your warranties are up, I will say this : today you are willing to fight anything and everything that is being critical of Tesla.

Many of you won’t be doing that, in a few years.

Of course, I’d be happy if I am proven wrong. If Elon pulls a rabbit out of his hat and does the fully autonomous drive from NY to LA, great! I have serious doubts, though.

 

[Knightshade]

You can have development happen all the time. You cannot have regression back to an older problem that was fixed, in a newer release. That should be the golden rule.

Gonna go out on a limb and suggest none of your software experience is with neural networks or machine learning based on that remark.

To all those who ponied up $12K for a feature that will never be complete before your warranties are up

To be clear, the product sold for 12k only promises 7, specific, things.

6 of them you get immediately upon paying for FSD.

Only 1 of them, autosteer on city streets, an explicitly L2 feature, remains undelivered... (and is the thing being tested in the FSDBeta program currently)

 

[_Redshift_]

I subscribe to FSD and dont have the beta. ive opted in and am excited to get accepted, but I couldn't imagine not having the other features of FSD. Driving 150-200 miles a day, on multiple freeways and highways, having it change lanes, and take exits without any effort from me is extremely valuable. I only drive about 5% of the day manually.

You're in the Bay Area so you'll know my route. Santa Cruz to SF to Oakland, back down to San Jose, then Santa Cruz. It drives the 17 to 85, to 280, to 101, over the bay bridge, and all the way back around. I dont drive AT ALL except for through SF. it's flawless for me. there is more value than the beta. Hell, I even use summon because my parking spot is tight and its easier to have it pull out for me in the mornings.

people believing the 12k or subscription are just for the beta, are stupid, and missing the point.

As a frequent driver on 17, I am somewhat skeptical about your claim that you drive entirely on AP on 17.

For one, there are a few sections where the car says to take over immediately, as the feature becomes unavailable for brief periods.

For another, the 50 mph hilly section is plainly difficult with AP. Most drivers drive 65 there and you cannot use the fast lane. (AP won’t do more than 60 there, obviously)

So you are stuck in the slow lane. Even there, Your are being tailgated by some impatient driver on most drives. That’s because the AP slows down to 35 mph on some tight curves.

Also, the steering goes a bit too ‘ping pong’ for me in the right turns. Couple of times, it asks me to take over and disengages.

Most of the time, I drive myself on the 50 mph section.

Not once, in the 30 or so times I have driven in the past 3 years have I ever completed the drive without the AP disengaging. Not once.

May be you are incredibly lucky!

 

[DrDabbles]

But I don't consider myself unskilled or untrained

You literally are. And you clearly don't understand what the industry considers "trained", which makes your attempt to claim this a clear indication that you don;t understand what's going on here.

You dont find out anything using trained testers, because they are trained.

You actually do. Trained operators annotate and capture other extremely valuable data for training. You do none of those things. That's why Waymo, Cruise, Baidu, and a dozen other companies are "ahead" of Tesla here even in cases where they started working on self driving years after Tesla.

Oh, and remember Uber

Tell me you don't know what "trained" means without telling me you don't know what "trained" means. Uber is the exact case that demonstrates how Tesla is playing with fire, and you don't even know enough about the crash to have that conversation. You should really go read the details of what happened, because you seem to be under the impression that Uber was doing anything right, and they weren't.

Would you want them to stop working on these betas and instead go work on improving the park/summon stuff?

I've actually stated multiple times in the past what I wish Tesla did / would do, but it's been a while.

What I want Tesla to do is sell a software package to consumers, tell those consumers that as newer safety features are developed, tested, and certified to work they will be delivered automatically to your car. But they didn't do that. Instead, they "beta tested" smart summon, pretended it was working, got a bunch of influencers to make videos, released to EAP owners, promised an update within two weeks when peoples cars failed to work at all and many drove over walls and parking islands. That never happened, but Tesla recognized the revenue. They're absolutely going to do the same with FSD.

Tesla had a real opportunity here to say their end goal was automated driving, but in the interim they'd release convenience and safety updates. But they didn't do that, and now Elon has gone on record for 6 years saying the AP system was "superhuman". Which is a bald faced lie. Now the entire share price is tied directly to FSD actually working, which is why you see some people in here so vehemently defending it even with clear evidence to the contrary. Many of the people you see commenting haven't even received the beta, which should tell you everything you need to know.

 

[t3sl4drvr]

#3 does not exist. The FSD package has always included other little things like summon or lane change etc. that customers got. You can argue if those features alone are worth the price paid but it is not correct to say the spent money and got absolutely nothing for it.

Are are 99% wrong here.

Everything you mentioned is already included in EAP that customers already had before FSD upgrade.

(The 1% part is the "stop at red lights [beta]" that is not included in EAP. In this context, it really does not count: such feature did not exist nor was any way communicated at the time when the customer purchased FSD. Thus it would have been impossible for anyone to buy FSD to get "stop at red lights" feature.)