“Translation: all the times Tesla has vowed that all of its vehicles would soon be capable of fully driving themselves may have been a convenient act of salesmanship that ultimately turned out not to be true.”
Another way to say that, is Tesla scammed all of their customers, since you know, everyone saw this coming…
Still no LiDAR.
“Trust me, you just need to buy more compute for your car. We’ll figure out reliable driving by sight someday.”
I’m happily using all the sensors my car offers, even if I’m pretty ok with driving by eyesight. Better sensors that can easily see through fog/rain/snow/whatever? Hell yeah, give them to me.
LiDAR in particular actually kinda sucks at those conditions (basically any form of precipitation). It’s really only good in clear environments.
Elon Musks make engineering orders of magnitude more difficult. Those poor Tesla neoslaves
They were on the path of self driving cars till Musk pulled the plug on the LiDAR and opted for cameras (cost less). He is directly responsible for why autopilot isn’t so auto.
https://www.theverge.com/2024/5/7/24151497/tesla-lidar-luminar-elon-musk-sensor-autonomous
Tesla bought over $2 million worth of lidar sensors from Luminar this year
I personally don’t think it’s a matter of more sensory input. Whilst Lidar wouldn’t be a bad thing, autonomous cars are just a problem current technology can’t solve.
The vehicles with a higher automated driving rating than Tesla use a more diverse range of sensory inputs. While it may not make fully autonomous driving, it very clearly would have made Tesla closer to it based on the fact that cars that use things like lidar in addition to cameras surpassed Tesla’s rating many years ago.
I think there needs be car 2 car ( or even car 2 human f.e. in smartphones and wearables ) introduced before self driving cars is capable to exist. Of course until real AI is introduced. We’re nowhere near the human capable ai
It is. The machine learning algorithm has maxed out its parameters because Elon decided to get rid of redundancy. The machine learning algorithm had to invent new algorithms to do what redundancy would have easily done in far fewer lines of code. They are out of compute power BECAUSE they decided to cheap out and removed redundancy.
It is to some degree. Lots of other new cars have lane keeping assist and automatic braking, BLIS, adaptive cruise control etc, and so on with more capable sensors and can for the most part drive without input from the driver better than the Tesla models with ultrasonic sensors or simply cameras. In fact the ones that rely solely on cameras absolutely do reportedly perform worse in testing. Musk was insistent that they could cheap out on the types of sensors used in order to make more profit and it shows. I don’t think it’s that tech cannot handle self driving currently. I think that it’s a numbers game where the firms attempting it want to do it as cheaply as possible while promising the moon and stars which they can’t deliver on a cheap budget. Vehicles like Ford’s (Blue Cruise) use all kinds of sensors including radar and GPS to allow for handsfree (not self driving) and it does work. The proofs of concept are out there in the world, but the costs to go from something like that to full self driving just doesn’t make it feasible for the average car manufacturer.
Lots of other new cars have lane keeping assist and automatic braking, BLIS, adaptive cruise control etc, and so on with more capable sensors and can for the most part drive without input from the driver better than the Tesla
Self driving is a huge step on from these things though. Hyper cruise control is easy in comparison and that’s what those features give.
Self driving means capable in all situations and all conditions. Not just highways.
Translation: all the times Tesla has vowed that all of its vehicles would soon be capable of fully driving themselves may have been a convenient act of salesmanship that ultimately turned out not to be true."
There’s a word for that already. Lied. They/He lied.
No need for 30 words when 2 will do.
They Lied.
Maybe one extra world: They lied maliciously.
Also, they did so repeatedly, over a very long time and while it must have been fully apparent with insider knowledge that this setup cannot work.
First, let me clarify I bought my Tesla used, before Musk went full fascist, and autopilot came free. The car was updated to the newest hardware for free, since the original FSD equipment couldn’t do it either.
That out of the way, FSD sucks, and it’s getting worse, not better. When if first come out of beta it was okay. I remember describing it as driving with a teenager, they got the general idea, but would make bad decisions so you had to watch them. Years of updates later and it’s practically unusable to me. It tries to go way under or over the speed limit, it hesitates or slams on the brakes for green lights. It slams on the brakes for cars that pull out with plenty of gap but doesn’t even notice the risky merges. It can not seem to navigate intersections anymore, damn near stopping in the middle of a turn. It actually just updated yesterday and I tried it again, it took me less than 5 miles to disable it again. It is, in my opinion, a hazard to use. I talked to my partner about it and we both agree it didn’t used to be this bad.
Anyway, the stupidest part of all this, is they changed it so it’s either full self driving all the time or not. You want cruise while you’re in traffic because you know it’ll try to cut in front of someone? Silly idiot, no you don’t. So you now have to have a second profile* for cruise control and lane keep without FSD. And the odd thing is that lane keep and cruise are fine. They function like FSD used to. They can drive the highway with no problem and trust me, I do not have much faith in the car so I’m watching it close. It can’t navigate city streets, but neither can FSD…
TLDR, my car was a better deal for me than Tesla. After years of FSD access, it’s bad and getting worse, not better. I can’t believe people pay 5 figures for it and maybe that’s why they feel the need to clip perfect drives or defend it.
I think car automation peaked at adaptive cruise control. It’s a simple tractable problem that’s generally well confined and improves the drivers ability to concentrate on other road risks.
I agree with that. Adaptive cruise and lane keep do reduce road trip fatigue in my experience. Tesla-bros bought the idea that this would be a fully autonomous car and it’s not. Rather than learning their lesson and using it as a tool, they put their faith in it anyway, weighting the wheel or whatever to get what they paid for regardless of what the car can reliably do.
Though they can induce another type of driver fatigue - it makes driving boring as heck as you don’t need to do anything. I can’t use line keep myself as it just makes me really tired and I’ll risk falling asleep.
I agree. VWs’ drive assists are absolutely stellar. It’s just line assist, speed limit recognition with cruise control and active distance assist, that’s essentially it. It’s not FSD but on the highway it almost feels like it. I was very skeptical and distrusted the sensors at first because my previous car had none of that, but after a while I got very comfortable with them.
I can even safely get something out of my bag on the passenger seat without worrying that the car is going to fly of the road if I take my eyes of it for a second.
The only thing that kind of annoys me, but that goes for all line assists, is that they don’t seem to follow a center line between the road markings, rather they bounce around inside a “zone” with margins left and right.
So if you are on the inside of your “zone” and approach a sharp turn, the car enters the outside margin at a fairly steep angle and often skims the outside road markings before bouncing back. It just feels like the assist is on a constant rubber band, so I don’t really trust it with high speed turns.
I’m a big fan of assists where I am still actively driving. They are there if I make an error (e.g. drift to the edge of a lane) rather than doing the driving for me.
Lane-keeping is actually on by default in my vehicle, and I find it to be a nice feature. Lane-centering feels too weird for me, so I’ve tried it out but am uncomfortable using it.
I absolutely love my adaptive cruise control, I use it all the time. I have a hybrid and it does a much better job of keeping the engine from kicking in than I do. Thankfully with Honda I can use it everywhere not just highways. It’s been my absolute favorite “new” thing to have in a car!
My wife’s hybrid Rav4 has it and loves it. I wish my Prius had it. I’m glad Toyota apparently knows how to do it right.
That out of the way, FSD sucks, and it’s getting worse, not better.
It’s almost like they bet on the AI to teach the AI, rather than continuing to pay for skilled engineers.
Buckle up folks, we’re going to see a lot more of this, across every industry, before the lawsuits go into high gear and anything gets better.
I don’t believe that you can use traditional algorithms to teach the car street driving, because there are to many different variations of intersections, traffic signs, special conditions like accidents, heavy Rain or fog, road closures or construction sites to get it right every time. Even if your autopilot is 99% correct and you drive 20000km a year, you still drive wrong 200km of it.
This doesn’t mean that AI will be better, because then you don’t even have a source code to track down where it went wrong to correct it in future updates.
I don’t believe that you can use traditional algorithms to teach the car street driving, because there are to many different variations… Even if your autopilot is 99% correct and you drive 20000km a year, you still drive wrong 200km of it.
Exactly!
And this is why, if the problem is solveable, it must be solved by learning models shepherded by expert engineers. The LLMs can take care of the long boring stretches, freeing skilled engineer time to fine-tune an LLM algorithm hybrid for the tricky bits.
I’m inclined to believe the problem is solveable, but since I’m not selling anything, I’m allowed to say “if”. Heh.
Since the first time I heard about FSD I’ve been wondering why Tesla (or others) doesn’t set up a system where drivers opt-in (no opt-in by default) to sending anonymized driving data to help train the model. The vast majority of the time, it’s probably modeling OK driving. At least no accidents. But the shitty driving and accidents are also useful as data about what to avoid.
Maybe they’re already doing this? But then I wonder why their FSD is getting shittier rather than improving. One would think with more driving data, good and bad examples, would only help.
I would be shocked to discover they’re not already doing this.
I’ve been wondering why Tesla (or others) doesn’t set up a system where drivers opt-in (no opt-in by default) to sending anonymized driving data to help train the model.
That’s exactly how they train the model, but every Tesla is opted in with, to my knowledge, no option to opt out.
Not enough paid humans sorting between which data is examples of good behaviour and which data is examples of bad behaviour. Not saying that is what is happening as we don’t even know if there is data, but that would be the weakness in that plan when run the way it would be run if instituted by elon.
Sounds like it’d be nice if you had real control over the car’s software, and you could roll it back.
This… also makes me a little more weary driving around Teslas in traffic.
Some of us Tesla drivers refuse to use any of their bullshit auto-driving software (I don’t even use lane assist anymore) because of bad experiences so hopefully most of them are just driving normally. Which I do admit may not spark much confidence given how terrible some drivers are.
Fully agree. The sort of good news for driving around them is that most of my frustrations come from it being overly cautious and almost getting rear-ended because it decided to stop for a green light or some other odd decision. It’s rare to have it interact poorly with someone that is driving predictably. Like, cut it off without a signal and you have introduced something has not already accounted for. Driving alongside it on the highway, it sees you and knows where you are. But people are unpredictable and it only takes one mistake.
That out of the way, FSD sucks, and it’s getting worse, not better. When if first come out of beta it was okay. I remember describing it as driving with a teenager, they got the general idea, but would make bad decisions so you had to watch them. Years of updates later and it’s practically unusable to me. It tries to go way under or over the speed limit, it hesitates or slams on the brakes for green lights. It slams on the brakes for cars that pull out with plenty of gap but doesn’t even notice the risky merges. It can not seem to navigate intersections anymore, damn near stopping in the middle of a turn. It actually just updated yesterday and I tried it again, it took me less than 5 miles to disable it again. It is, in my opinion, a hazard to use. I talked to my partner about it and we both agree it didn’t used to be this bad.
Sounds like it still drives like a teenager!
Which of course is terrible since it should be improving over time.
I remember describing it as driving with a teenager, they got the general idea, but would make bad decisions so you had to watch them.
This is worse than just driving yourself. I either need to be engaged in actively driving, or it really needs to be able to handle the task by itself.
It’s why I find the lane-keeping feature in my vehicle to be useful, but lane-centering is just too weird for me to use.
I went to Texas for the eclipse. Made a big family vacation out of it…landed in Houston, rented a Mustang Mach-E, stayed there for a few days, drove to Austin for a few days, drove to Dallas for a few days (and for the eclipse, was at the Perot), then back to Houston for a few more days.
I say this because this was a lot of highway driving. More than I would usually do. And I absolutely loved one-pedal driving in the city, and the adaptive cruise control and lane keeping on the highway. I trusted it much, much more than in our 2019 Odyssey.
Anything more than that, I don’t think the tech is really ready for. I wish it were. I know theoretically a computer could be a much, much better driver than humans…but it takes a non-trivial amount of intelligence to drive. We take it for granted, because a lot of it is practically instinctual to us, and almost entirely subconscious. It’s an incredible amount of identification and complex decision making that goes into it if you actually break down the number of inputs you observe and variables you “know” the values of (such as stopping distance for various surface and weather conditions).
I guess he’s talking about other upgrades like radar sensors and all the stuff that people told him from the get go.
This article specifically mentioned only changes to the computer hardware, not the cameras.
they changed it so it’s either full self driving all the time or not
No they didn’t…you can still activate regular adaptive cruise control without any of the FSD nonsense.
Yes, I said that…
So you now have to have a second profile* for cruise control and lane keep without FSD.
No you don’t need a second profile for that, you can just pull down once on the right stalk to enable regular adaptive CC with lane keep instead of twice for FSD. This second profile requirement is complete nonsense.
The “pull once for FSD” that removes the regular adaptive CC is a voluntary option in the settings, you can just disable that. If you have that enabled and don’t like it, that’s your fault not theirs.
I can’t help but think he’s saying this now as an attempt to distract from the stories of "Musk has been talking to Putin since the spring when they were both faced with problems: Musk being forced to buy Xitter and Putin unable to steal Ukraine. Odd how Musk has been becoming more rabidly pro-Russian-interests, isn’t it?
I rather think he’s trying to tank the Tesla stock price lately. No idea why though.
He and Trump deserve each other.
I think his intense commitment to getting Trump elected makes more sense when you consider this article.
His enormous wealth is largely stored in the form of Tesla stock, and that stock has been valued based on the belief that it isn’t a car company, it’s a robotaxi service currently selling the hardware to finance the software development. The value – and his wealth – can persist indefinitely as long as investors continue to accept that premise, no matter how long delayed. But if something tangibly undermines that premise, Musk could conceivably lose the majority of his wealth overnight.
The National Highway Traffic Safety Agency is probably the greatest threat to his wealth. He doesn’t worry about competitors or protestors or Twitter users or advertisers. They’re all just petty nuisances. But the federal regulator over roads… that is his proverbial killer snail. And I think fully capturing the entire federal regulatory state is his strategy to permanently confine that snail.
More than anything else, I think that’s what is motivating his radical embrace of fascism.
Well any political choice will increase the wealth of the billionaire class in general. Any candidate that is a threat to that will be smeared to dust before they get anywhere near power.
But Musk is going for the more direct approach, and tbh I think the second Trump gets power again he’ll have no need for Musk and will treat him like the parasite he is. He’s certainly done that before.
Sometimes I’m reminded that there’s always a chance that they go submarine diving or some such with another overconfident crony who thinks their skills got them where they are today.
I don’t think he would go submarine diving, it would be more in-character for him to try and fail a moon vacation.
