Tesla's self-driving tech ditched by 98 percent of customers that tried it
the-express.com
the-express.com
I thoroughly enjoyed FSD.
The best way I have to describe it is: Picture a timid new driver, then give them 10 cups of coffee and make them really nervous. Now make them adhere slavishly to a random assortment of rules, and not at all to others.
One of my biggest pet peeves is that my car will try really hard to stay on the right side of a narrow residential road, coming inches from parked cars when there is no oncoming traffic. I always takeover, because I just know somebody is going to open a door into me, or an animal or child will dart out from between cars.
It will often hesitate to merge or turn left, even when given ample opportunity. And then it will suddenly dart across traffic with no warning.
I bought it in 2017 with the car, and drove like a grandma for a month to be eligible for the initial FSD beta a year or two ago. It was a complete waste of time and money. My hope when buying it was that I could have my car drop off my then 11yo son at school and save me a commute. My son is now 18 and drives far better than the car ever will.
Instead, I bought a used Model S with AP1.
I have never regretted that decision, it's been a great car. I do feel bad about everyone who bought into those promises.
I've been in software dev way too long to buy that nonsense. As far as my expectations have been it's actually ahead of schedule. That kind of pessimism keeps me constantly happy with the rate of improvement. Lovely psychological trick I've played on myself.
I completely forgot that stuff like this was supposed to be a huge new thing years ago. Sending your car to be a taxi while you're at work to bring in more income, picking people up..
Like, wtf? How is making these claims as an Officer of a public company remotely legal?
That's on-going! Just keep in-touch with news.
https://www.reuters.com/business/autos-transportation/tesla-...
* Matt Levine ™
It seems one of the potential stumbling blocks is mandatory arbitration unless you thought to opt-out very early.
It's actually so wild that the judge upheld arbitration[1]. I guess he's technically correct that it doesn't "waive a plaintiff's right to seek public injunctive relief" but it definitely makes it way more annoying.
[1] https://arstechnica.com/tech-policy/2023/10/judge-upholds-te...
There's a bill before Congress that would make arbitration agreements unenforceable in class-action, employment, civil rights, and antitrust disputes, but it hasn't gone anywhere despite being introduced in the last 4 Congresses.
I'm not a lawyer, but there seems to definitely be some discretion allowed. In fact, the McGill rule is often cited as being a litmus test[1]. Honestly, in the Tesla case, and why I said it's wild, is because Tesla is still selling cars touting full self driving[2]. The injunction here would presumably bar Tesla from selling (or at least advertising) FSD in the interest of public good. In other words, I think the judge got it wrong.
Again, I'm not a lawyer, so I might be misunderstanding something, but that's my take.
[1] https://www.foley.com/insights/publications/2022/02/a-new-er...
> May 8 (Reuters) - U.S. prosecutors are examining whether Tesla (TSLA.O), opens new tab committed securities or wire fraud by misleading investors and consumers about its electric vehicles’ self-driving capabilities, three people familiar with the matter told Reuters.
While it was enabled, I'll say it was terrible at things like driving in the city, but it was actually pretty good at interstate driving.
So hasn't Tesla stolen from you? You should demand your money back.
This guy did: https://teslamotorsclub.com/tmc/threads/my-experience-taking...
He got his money back with interest.
But look, FSD is smoke and mirrors. v12 is marginally better, yes, but it still acts horribly. It can't merge well. It's jerky. Last night it tried to drive in a breakdown lane. It often misses on-ramps for highways around here.
Man, the auto wipers don't even work. That makes me the angriest - what would they pay for a rain sensor? Jeez.
Vision only may work with better cameras but seriously Elon put Ali Express $0.20 cameras in this thing.
I’m always slightly amazed that people bought this. Like, everyone who worked in the space was saying “yeah, no” at the time, as far as I remember; it was really just the techno-optimists claiming this stuff.
The last couple months have been much better. It's STILL not perfect (and it's pretty much imperative to simply ignore anything that Musk optimistically tosses out regarding capability) but it's a remarkable improvement.
It'll drive me from my home to my co-working space with the only intervention by me being to click a parking spot. The automated parallel parking actually works well now too.
My guess is we are 5-10 years out from it being a truly automatic experience in every way, but it's pretty good now. It's also nice that my car keeps improving (quicker lately) rather than being stuck with whatever nonsense it was deployed with.
Right now, it works just about flawlessly on the freeway. If you turn it on when you are on the on-ramp, and turn it off when it delivers you to that first traffic light after you exit, it actually is maybe better than actually driving. It passes slower cars, it is way safer than I am at checking my blind spots. It uses the HOV lane at appropriate times. I enjoy letting it perform the chore of getting across 5 lanes of traffic in time for an exit much more than doing that myself.
For local driving, it is still not perfect enough. It WILL get you home safe... but it is more stressful having it drive than driving myself. It pauses awkwardly sometimes... it makes an occasional wrong turn. It handles traffic lights and stop signs reasonably well but overall it's like driving with someone who is just learning to drive. It's more stressful than I prefer my transportation to be.
That being said, I think in a couple of years it will stand as an incredible achievement. The driving behavior it exhibits is incredibly lifelike at times and it is really so close... like all engineering projects the last few percent is sometimes the hardest, but I can see it getting there.
Edge cases.
It does have adaptive cruise control and the ability to stay inside the lines, but it absolutely will not automatically change lanes or exit off the freeway.
The average person isn’t going to close their eyes and take a nap, but the average person will start spacing out, will start seeking their phone out to kill boredom, etc.
Being on your phone is completely different thing tho. I'm sure nags for distracted driving will stay for a while.
The lying is bad enough to turn me away regardless of the performance. If you call something FULL. SELF. DRIVING. then explain that it is not in fact fully driving itself, I'm not interested in your product. FSD means I should be able to talk to my watch and have the car arrive to pick me up like I'm Michael Knight.
However, you can't save everyone because then people would get suspicious. So "They" calculate social scores for everyone and people with a sufficiently low score don't get saved when they mess up.
The real punchline here, though, is exactly what you say:
> If you can't stop paying attention, what's the point?
Why would anyone actually want a full self driving car that you have to pay attention to all the time outside of a shadowy cabal from a half-baked conspiracy theory? I'm really not sure.
If you don't see the advantage until it is actual FSD, then simply wait to purchase/use it until it is out of beta.
Paying attention != nags. You can be staring into traffic and holding wheel and still not paying attention (which nags do help with). Nags when you are paying attention is super annoying tho.
Being on your phone, etc is completely different category. That is distracted driving.
This is not the selling point you may think it is, and not what was promised
I often drive with my elbow on the top of the door, hand on the side of the wheel. The weight of my arm is enough that I frequently go many minutes without needing to nudge.
Otherwise, I agree. The nudging is so annoying that it's "more fun" to just drive.
When using autopilot on a single-lane road with a 60 km/h speed limit, it performed a phantom emergency brake because there was an oncoming car (which was driving normally in its own lane) on a curve. Maybe I shouldn't have used autopilot in that situation?
But what's even worse is the cruise control. It's a basic feature, yet Tesla's cruise control suddenly changed the speed limit to 30 km/h when passing through a tunnel with no signals and performed an emergency brake in an 80 km/h zone. This is incomprehensible.
Not to mention the issues with noisy brakes, distorted sound effects.
However, I don't want to switch to another car. I've driven Polestar 2, Ioniq 5, Kia EV6, and Nio ET5T, and none surpass Tesla. The only cars that make me consider switching are Lucid and Taycan, but they are in a different price range.
With Norway soon stopping the sale of all gasoline cars, Tesla is almost the only choice. I just hope they can improve on these basic issues.
i3 famously and its successor iX are clean sheet expressly EV designs.
i4 is also available as M430 ICE. i5 and i7 also are dual platforms but they follow original 2013 i3 clean sheet by about a decade.
I would never use Autopilot with oncoming cars. FSD is fine with oncoming cars, though.
What makes it better?
https://www.tesla.com/blog/all-tesla-cars-being-produced-now...
It's from 2016 - almost a decade already. But fear not, FSD v12.5 is coming to the Cybertruck in ~2 weeks probably !
Tesla and Musk, spent the past 10 years saying that many things they work on are <1 year away. For any sane person, that timeframe implies the product is ready, already tested and it's just a matter of figuring out the details.
I've recently spoke with somebody who absolutely believed Tesla is going to automate all their US manufacturing and fire all the workers, thanks to Optimus, beginning next year !
The problem is even 99.99% good is not 100% good and for something that has your life in its hands, it has to be 100%. Arguably a human driver is not 100% either, but at least they understand the world around them on a fundamental level, plus they have a pretty dramatic forcing-function ('do not die') that instills confidence. An AI has nothing to lose and this makes a 99% safe AI system less trustworthy than a 80% safe human driver.
Given its performance I wouldn't dare trust it with any of my daily driving even if it was free.
Basically Tesla gave out a free FSD trial to every Tesla owner, and per this report, 2% of them actually signed up for the feature and are paying Tesla either the one-time or monthly fee.
I'm not sure about other "free-to-paid conversion" marketing campaigns, but 2% actually seems like a good conversion rate.
At Tesla's market cap, as in, "worth" more than the rest of the auto industry, any conversion rate less than about 90% is a huge failure. They have been pushing self driving for a decade now, with billions of dollars worth of investment and some of the best engineers for this in the field, and all they have to show for it is a 2% conversion rate on people who have already self identified as being willing to pay more for a new, unpolished product.
Consider how much other companies have advanced in self driving in the same time, or how much better solar panels got, or the massive improvements and developments Space X had in the same time frame. Meanwhile, FSD has barely increased in value in the same time frame, and is no longer a unique selling point, but a competitor in a busy market.
Tesla needed to be revolutionary. They needed world changing improvements, huge step changes in functionality. They needed something that every single human being desperately needed, even taking it's faults into account. Instead they've only really made incremental improvements over ten years and lost any tech lead they might have previously had. They've failed.
It literally doesn't even matter if Tesla cars are better than any other car on the market. To justify their valuation, they needed something magical.
I'm not sure how or what Yipitdata measured (and I can't find a link to their data), but if was the older releases it makes perfect sense that very few people would buy it.
So you’ve got Tesla, which sells a stupidly priced ADAS system with or without a subscription that drives like a teenager.
Comma, great price, but insists they’re solving self-driving cars and their engineering team scoffs at object detection and the unit basically can’t turn at intersections anymore.
And Waymo. Great passenger experience, highly tuned longitudinal comfort, industry leading perception and sensor tech. But you can only use them as a taxi.
But unless I can take my hand off the wheel or do something else it's just not worth it.
Regular autopilot is good enough for highways.
If they can make it level 3-5, I think it will EXPLODE.
I understand why you're asking, but it's terrifying that there are people who would gamble other people's lives with the answer to this question.
Even though FSD was glitchy, (entered a road without stopping, nearly got into an accident trying to take an exit,) I really enjoyed FSD. Specifically, I really enjoyed it on the freeway. I just don't see it as something that's worth an eye-watering amount of money, especially on a car that's underwater, or a car that I'll probably replace in 3-4 years.
I really wish Tesla offered a deep discount on FSD as goodwill for those of us who bought the Model Y right before they dropped the price. It's really hard to justify spending that kind of money after the car's value dropped so quickly.
$8k might be what Tesla needs to charge to break even, but as anyone involved in business knows, what something is worth is what someone is willing to pay, not what it costs to make.
* Tesla absorbs all legal consequences of any accident while using their self-driving technology
* The driver doesn't have to sit in the driver seat continually monitoring the situation, i.e. driving - but in a really weird way
Which is to say, Tesla does not have self-driving tech and they're not going to have self-driving tech at any time in the near future. If I have to continue sitting in the driver seat and continue driving/not-driving and be responsible for anything the car does, then the hell with it. That's not self-driving and I wouldn't pay for that either. Especially the insane prices they're charging. Nope. Not gonna happen.
https://www.motortrend.com/news/nhtsa-tesla-autopilot-invest...
Waymos aren't running with a human driver ready to take over at all times, so that's not correct.
You can disagree on the value and still recognize that $25/week is not pathological.
It's wrong too often to be convenient. I'm not paying $25/week for more work as a driver.
This is imo the biggest problems that forces users from not buying the tech. I used the beta, and not only did I have to basically pay attention, but at times, when I was already paying attention, Tesla would ask me to apply pressure on the steering to the point where self driving disengaged and I had to take quick actions. On top of that, just driving around my neighborhood, it failed miserably. It couldn't park itself when the entire curb was empty and I stopped in a non ideal spot. It would often incorrectly identify parked cars on curved streets and brake. A couple of times it almost ran over pedestrians at a stop sign, mostly because it accelerates too much on turns at a stop sign. It absolutely refuses to drive faster than the speed limits, going 25 on two lane main streets where everyone expects you to go 40.
At some point, its more stressful for the driver on self driving than driving yourself.
This is not an issue with self driving but rather law enforcement.
I’m not sure what a self driving car is supposed to do here. Hell, as a foreigner, I myself wouldn’t know what to do.
But the again, you raise a fair point. I don't know what course of action an autonomous vehicle is supposed to take here. If anything, in HCI, you want consistent actions to be taken, for users to reward you with a higher level of trust.
https://www.ticketcrusherslaw.com/traffic-ticket/speeding-ti....
So yeah, I still don't believe you that I'd be fined for driving 25mph in an area, where 25mph is the speed limit.
If somehow you got a ticket for this, you’d just say it was necessary for safe operation, after all, why else would there be a speed limit in the first place?
Elon claims the next version (12.4 / 12.5) will be milestone improvements in terms of miles per disengagement, but only time will tell.
Allow me to stop you right there,
When navigation tells you to "keep left", FSD happily changed lane to the right.
It makes more sense to just use FSD as a Lane Keep Assistant System on highway, with occasional surprises like switching to a faster lane mostly automated.
That being said, these are existing customers and I think the pricing for FSD is just bad.
I suspect they will eventually lower it.
Tesla won't win a price dumping competition on the mechanical engineering and manufacturing quality for their cars.
They are very good value cars in their category though.
Right now most major car corporations have something competing with them. To be honest if I was buying an electric car tomorrow, for the price of a Tesla, I wouldn't be buying a Tesla.
It’s a serious question. I own a few years old Tesla but I wouldn’t buy another one until Musk is out, but everything else seems worse or a lot more expensive.
I think what you should do is try to figure out what kind of car you want and what you want to do with it. The likelihood that you won't find anything that fits seems quite small, since most of the world's largest automakers are trying to make cars exactly for people like you.
BMW is completely in another price range once you put back the equipment that comes standard from the cheapest Tesla.
The people I know who didn't buy a Tesla but something similar, before the CEO turned out to be a nazi sympathiser, they mostly bought other brands due to their brand loyalty or aversion to change.
But, if I would look for an electric car in Tesla segment I would look into Hyundai Ionic 5/6 or Kia EV6 - they are in general great cars, which have proper service networks and will be around for years, what sets them apart from a chinese tesla-killer du jour.
Bought in 2018 and v12 has been absolutely amazing. Next version 12.4 will also remove the steering wheel nag.
Just going from sitting in the driver's seat to sitting in the back seat would require that, by now, FSD would have to be reliably comfortable and the only issue with not using it would be price.
There are safer-than-Uber, but still not flawless, robotaxis in limited areas. We know what it takes to get there. We also know it has to be done with great care and a huge amount of preparation. It isn't cheap or fast.
Narrative violation
No problem. They already paid 10,000 bucks for it. Now you can sell version 2 in three years
"We are almost there"
But, after trying the FSD for 30, I was pretty impressed. You do have to pay attention but it’s usually easy to see where it will get confused and you may need to intervene.
The reason I didn’t buy it is not because it’s not good, but it’s because of the cost. $8k is a lot of money.
So yeah, the title feels a bit dishonest to me.
I'd imagine their going to do something like FSD with Full Autonomy requires Tesla Insurance to optimize their overall cost in the event of accidents that will happen (just like with even the best of human drivers)
Before you retort with "Musk lies", this is a very specific claim and not a prediction, and would land him in legal trouble as said by an CEO of a public company.
As always anything touching Tesla in the media is extremely biased, like this headline.
Before you retort with "Musk lies", this is a very specific claim and not a prediction, and would land him in legal trouble as said by an CEO of a public company.
Musk has spent years making very specific claims about full self-driving that turned out to be lies: https://motherfrunker.ca/fsd/
In 2016 Tesla claimed that "as of today, all Tesla vehicles produced in our factory – including Model 3 – will have the hardware needed for full self-driving capability at a safety level substantially greater than that of a human driver":
https://www.tesla.com/blog/all-tesla-cars-being-produced-now...
That was a lie.