> Tesla’s official response to the clip is that the auto high beam system still is classified as beta software, meaning it’s not fully developed
Things like this (and panel gap, and build quality, and over promising features) are why I have yet to buy a tesla, and likely won't any time soon. Even though they were the only viable luxury electric car (during my last car search), I decided to pass. I will probably seriously re-evaluate them for my next car purchase, but right now, no thanks!
BTW, I got my first demo of tesla FSD a few weeks ago. In san francisco, on market street, a place I'm sure is tested extensively. Within minutes, FSD had the car perform an illegal AND dangerous mid intersection merge followed by an extremely hard and jerky lane follow correction. lol.
I find it interesting when people list things like 'panel gap' as a reason they won't buy a car. I grew up in Detroit and honestly it's pretty embarrassing the stuff that goes on with most Big 4 cars. Case in point - my sister worked from one for ~30 years and still gets a 'green sheet discount' so she loyally buys their brand still. it's comical how every time she gets a new car something pretty significant goes wrong within 6 months. This time, her 3 week old Jeep bricked itself because the key fob wouldn't communicate with the car and refused to start. The emergency break also suffered a software failure that refuses to unlock the break. The panel gaps though were spot on. Whew.
Panel gap matters because it is a clear indicator of manufacturing precision, and it also affects weather proofing. It is a very strong signal (but not the only one!)
Most people buy cars based on brands, and also a very substantial amount of car purchases are impulse purchases.
Anecdotally, I rented a jeep a few months ago (compass) and it was truly abysmal.
Are they though the only one available though? Both Mercedes and Volvo (Via the Polestar 2) seem to have pretty decent options that match most of Tesla's features and seem more polished and better built. They won't drive themselves but Tesla's don't seem to do a great job of that either.
Telsa used to be the cool kid because their cars were fast-charging, neck-snapping fast, had autopilot, and were very efficient per-mile.
It was enough to make up for the assembly/initial quality, material quality and NVH issues in the interior, poor reliability, and for some models, very dated design.
Now they're being beaten in almost every category. GM has one of the best fully automatic cruise control systems now; Mercedes has full self driving on the autobahn. Many cars are as or more efficient per-mile, out-accelerate all but the most expensive Teslas, have as fast or faster DC charging systems. The Audi e-tron and Porsche match or out-accelerate them while having top-of-the-market build quality and design; stepping into any modern Porsche is a like stepping into a swiss watch. Lucid's car out-accelerates and screams "FUTURE!"
There are a ton of good EVs out or coming out in the next year or two. All of those cars are assembled better, are more reliable, have nicer interiors, less dated designs, are backed by much more extensive dealer/service/parts networks, etc.
Ultimately Musk's problem is that he thought he could to better at building cars than companies which have been building cars for a century.
The problem wasn't that car companies couldn't do better than Tesla; it's that they didn't want to. Now that EVs are a Thing, they want to. They have more cash, more manufacturing capacity, more R&D, more dealers, better supply chains, and much better relations with the media and automotive press.
Tesla is fucked long-term. I suspect Musk's plan is to pivot to being an energy/charging company - that Tesla cars were just a vehicle to get him a worldwide charger network and accompanying infrastructure (it's the one thing that Tesla does just about flawlessly.)
> GM has one of the best fully automatic cruise control systems now; Mercedes has full self driving on the autobahn.
Compare either of these to Tesla's Autopilot (Which is NOT the same as FSD!). I'd be surprised if there were significant differences.
> Many cars are as or more efficient per-mile,
Are they, though? I dunno about other cars, but the Taycan is extremely inefficient. The Taycan 4S Performance Battery Plus has a 93 kwh battery, but only 227 mile EPA estimated range. Meanwhile, my Model 3 Performance is 75 kwh, but EPA rated 300 miles.
> out-accelerate all but the most expensive Teslas
And significantly more expensive.
> have as fast or faster DC charging systems
That are fewer and farther between, not to mention expensive, and also I'm not sure about the faster part. The latest gen superchargers are up to 250 kW. I can drive on any Interstate highway in the continental USA and not have to worry about where to charge. Even the I-90 corridor across over 1,500 miles of rural highway has enough Superchargers on it to make a road trip viable.
That all said...
In the long term, you may be right. But we're still 5+ years away from their downfall. The other manufacturers have REALLY been dragging their feet on making good EVs. Honda could be selling tens (or even hundreds) of thousands of fully electric Civics and CRVs, but don't because ????.
You should look up the new Mercedes system. If you haven’t heard about it it only functions up to like 37 mph, but it’s level three self driving. The kind where the driver doesn’t have to monitor it. So where it can work (map freeways) it can completely take over in slow traffic. Pretty impressive.
I agree with you about the 5+ years thing. Ford released an EV that they basically never advertised and got swamped with orders. They announced the next one and got even MORE swamped because it was the F150. They’ve had to delay the introduction of at least two more EV models just to try and keep up with demand for the first two.
My understanding is VW has been doing extremely well with the ID.4. I know there is heavy interest in the Kia Niro and the Cadillac Lyriq and of course the new Hummer.
Other newcomers like Rivian and Lucid are pulling a lot of eyes/orders.
Tesla owned the EV segment because outside the Volt/Bolt/Leaf there was nothing else. Certainly those aren’t equivalent to the S/X/Y or even 3.
But the question is, can Tesla improve faster than the rest of the entire automotive industry can move to EVs. I doubt it.
In some ways this feels like a double disruption. Tesla came along and disrupted car manufacturing I proving that EVs could work and were even popular/desirable. They also proved online car buying could be quite successful.
But I think they will now be disrupted by the companies they tried to replace. I just don’t see how they can hold off every single major car company in the world. They may have the most efficient cars (from my understanding) and they’ve obviously got the whole over the air update thing figured out. But they don’t make enough models to cover everyone’s taste and they still have some real manufacturing/customer service issues that they could have grown a out of by now if they had wanted to. I feel like they may have spread themselves much too thin, and lost their chance at king of the hill because of it.
In my mind they’ll never be a VW or Toyota sized car manufacturer in the long term. But will they be as big as Audi? Or just a Jaguar? Or gone like Pontiac?
Don’t know. Bankruptcy? Merger? Acquisition? Exit the consumer market? Will be interesting to watch.
> You should look up the new Mercedes system. If you haven’t heard about it it only functions up to like 37 mph, but it’s level three self driving. The kind where the driver doesn’t have to monitor it. So where it can work (map freeways) it can completely take over in slow traffic. Pretty impressive.
Tesla's Autopilot works perfectly on the highways, in my experience. I've done a few thousand miles on the highway, and I think the only times I've had to manually take control are when my one lane is becoming two lanes, and it gets confused.
I do agree that eventually, the other manufacturers will surpass Tesla, but it's going to be a long timeframe (I think ~10 years), and it will hinge on being compatible with Tesla chargers, since the other networks are either too slow (Seriously, whoever decided that 7.2 kW J-1772 chargers should exist did a major disservice to EV adoption), too expensive, and too spread out.
I’m not arguing against autopilot. But it’s level 2. You still have to monitor it. I find the Mercedes system interesting because it’s level 3, even if only in limited circumstances.
Also level 2 charging? Preach. Public charging (except at like hotels) seems like it should be minimum 50kw.
You only "have" to monitor Tesla's AP for liability purposes.
If the only difference between Level 2 and Level 3 is whether or not you have to monitor it, then I think its just market buzz, and Mercedes will kill someone because they'll claim that their system didn't need to be monitored, so someone won't, and it will make a mistake and someone will die.
Limited to 37 mph with the claim you don't need to monitor it means someone is going to be in heavy traffic, activate it, then fall asleep. When traffic clears up and they're still sleeping, they're going to get rear-ended when the car is putting along at 37 mph while everyone else is going 60+.
There's a non-zero chance that one of those OTA updates will end up killing someone (most likely a pedestrian). I absolutely would not trust a company with Tesla's track record on inability to deliver reliable, mature software to never ship an update that'd cause FSD to suddenly decide a pedestrian was empty air.
>Tesla is fucked long-term. I suspect Musk's plan is to pivot to being an energy/charging company - that Tesla cars were just a vehicle to get him a worldwide charger network and accompanying infrastructure (it's the one thing that Tesla does just about flawlessly.)
This was always my take as well. He wants to make cars (they're a halo product), but the real value of tesla is in skateboards, batteries, charging. I suspect that things like the roadster/cybertruck are tesla preparing to transition to a business model where car sales are not the largest part of their business.
Sorry! To be clear, at the time I purchased my car, there weren't any that ticked off the boxes I was looking for.
Teslas were the only ones that did, that were also electric.
If I were to buy an electric car today, i would probably consider a polestar, taycan, audi, tesla. I would give them a fair shake! I'd probably end up with the taycan though, tbh. I would rather wait a year or two though, there's some really interesting things coming down the pipeline.
I notice that you said not that "my tesla is great" but that "they are great". Do you own one?
I will not purchase a car that has a panel gap I can fit my pinky in. I will not purchase a car that is at the bottom of reliability ratings. I will not purchase a car without radar sensors for the adaptive cruise control.
Regardless, I ended up buying a faf bmw and I'm loving it. So I did okay, anyways.
Yes, 2020 M3 - I have had zero problems in 30K miles. I think there is a misconception that there are a lot of Tesla owners that are not satisfied with their purchases - It seems to me they are a tiny (albeit vocal) minority.
The various software issues my Tesla M3 has are really grating at this point. I don't even have FSD, its just the regular AP software. I get:
- Random decisions about turning windshield wipers on and off. Frequently it just keeps wiping a dry window, or never starts.
- PHANTOM BRAKING, I've had it slam the brakes multiples times. Usually a bunch of times a week. Just for funsies.
- Garbage night AP performance. Car coming in the opposing lane with its lights on? Better slam on the brakes.
- Garbage bad weather performance. Just a bit of rain or snow and even TACC can't go a reasonable speed.
A whole lot of this probably comes down to the lack of radar or other sensors. And the lack of ability for them to get machine learning with only cameras to perform.
Maybe they should just send the engineering team up far north during the winter and force them to dog food their software to get around. Really it's not the individual engineers fault, it's always crap management.
> Just a bit of rain or snow and even TACC can't go a reasonable speed.
As you note, unlike most manufacturers' TACC, Tesla has made the (IMO terrible) decision to use a purely computer vision based system instead of radar. That's why it works so poorly in adverse weather conditions, and that's also why it will never work well in adverse weather conditions.
Tesla’s “don’t blame us it’s beta software” schtick is getting real old. No other automaker deflects all blame on faulty safety related system that way.
It's the Silicon Valley way, and it has no place in safety critical systems like cars. Ironically, Tesla is the worlds highest valued car company specifically because they pretend to be a software company. In a way they've made their testing an external cost that customers and other members of the public pay for them.
Governments (or EU) should just ban all beta software from public roads. You want to ship a feature to customers? Fine, but you'll have to take responsibility for it first. Make sure it's tested, has well understood failure modes, who's to blame if it malfunctions, etc.
>Fine, but you'll have to take responsibility for it first.
This is kind of where I draw the line. If the manufacturer isn't willing to accept liability, then I'm not going to believe them when they say that the car can drive itself, especially if I'm paying for said feature.
Also the argument that "the driver can opt-out" is so facile. I don't drive a Tesla, I can't opt out of the beta but I still need to live with the Teslas in my town.
Tesla is aggressively and non-consensually consuming part of my daily road-safety budget to subsidize development of beta software in pursuit of long-term profits.
Now if they open-sourced their FSD sensor data to help train other self-driving systems I think my perspective would be different.
How is beta software controlling the car, but with a licensed adult who can override it, any more dangerous than someone on a learner's permit driving a car, with a licensed adult in the passenger seat?
Because the kid with the learner's permit can ask the adult for help if they are confused about what to do, rather than having to simply pick and option and go with it.
It's a feature of many cars, and the majority of them buy a standard well working chip for it (usually from Bosch). Same with the automatic wipers.
Tesla is the odd one where they decided to implement those kinds of features themselves, using their own machine learning algorithms for the cameras. Which results in random weird behavior that just doesn't happen with other brands.
It's a fundamental problem with Tesla, they think they know better but the result is shit. Apart from the wipers and now high beam, autopilot is another example. While Tesla was a pioneer in the area, it now performs worse than most other brand's adaptive cruise control. They've shot themselves in the foot with computer vision, with many phantom braking events as a result.
It kind of looks as though the moose reflects enough light back to Tesla's cameras to make it think another vehicle is oncoming so it switches off the beams. Then turns them back on when the "vehicle" (aka moose) disappears.
The only obvious explaination is that it detected a "car" in the opposing lane.
Given what went on, it's in the direction of what would generally be the opposing lane.
It seems that the large mass of light haired Moose reflected enough of the Telsa's own light back at it to make it think it was a car approaching from far away.
It wasn't a car far away with lots of light, it was a moose, close, with little light.
Both appear to the Tesla as the same thing.
This happens ALL the time with UFO's. It's a problem with the human eye, light and distance are indeterminate. It could be either way around, close and dim, far and bright.
Seems better integration of light vs radar is needed.
Ha yup this is my suspicion. I posted the same exact time the same theory lol. Seems like it behaves normally. I would expect Tesla to not want to look for cars first then turn the brights off, but error on the side of anything resembling oncoming traffic triggering low-beams.
A perhaps-stupid question: Does the software involved attempt any real classification of objects in its field of vision? Or is it a neural network - where all sorts of "interesting" things tend to happen for inputs which are a poor match for its training set?
The softwareization of cars has so much potential, but is terrifyingly poorly implemented. It should arguably be developed using, and be subject to, the same extreme rigor as aircraft autopilot, but it seems like it's the wild west.
I recall reading somewhere that some animals get fixated on the lights before they get hit. Perhaps flickering the lights at high speed can get their brains to not zero in on the lights and get out of the way.
2021 Toyota Corolla owner here .. it works well enough, but random reflections off of street signs and such cause them to randomly flick on and off. I wouldn't call high beam assist a "safety feature", but more of a "hey, it's better than nothing". I tend to use my high beams more in scenarios where it would have been too annoying to turn them on and off, so it's net safer. If that makes sense. I wouldn't classify this article as news ... it's sort of a given that this technology is a bit iffy?
For that matter, the Toyota lane keeping is one of the better and safer ones that I've driven, but it flakes out and gives up way too easily to rely on (a small gap in the line and it disables, as opposed to trying to push you into curbs). Again, just a "better than nothing" experience that can be useful on the tail end of a long drive. In the case of Queensland -> Melbourne, the car was the better driver at the end.
The street sign reader, speed warning and rear cross assist is absolute black magic and I have no idea how it's so reliable!
> "hey, it's better than nothing". I tend to use my high beams more in scenarios where it would have been too annoying to turn them on and off, so it's net safer.
I wonder what are those scenarios? Shifting gears? Aren't drivers suppose to be more careful before shifting?
I'm just asking because at the moment I'm not sure If I can agree with "better than nothing". Putting assistant in to MS Office is one thing but cars?
Well it's a fully auto hybrid, but often times I won't use high beams on a semi busy rural road because it's too much bother to turn them on and off every minute or so. Especially if you're going around corners and things like that, and someone appears.. at least with the auto dimmer, you can just leave them on and yes, it's a bit over-sensitive, but it still correctly dims them immediately when it needs to. More importantly, I leave the high beams on rather than get irritated and just turn them off, hence it's net safer.
High-beam assist with matrix headlights is amazing - keeps the high-beam on but just changes the shape to avoid the exact area that would dazzle someone else.
I think matrix LED lights are not allowed in US? But yes - having matrix headlights is a huge safety benefit. I am driving on high beam most of the time and only small parts are dimmed if there is a car in that sector. Works really well and it was well worth the money (Audi A6).
I have it on my Honda. I'd say it's just way to conservative. Conditions have to be perfect for it to turn on the high beams. Any time they are on they are on correctly, but there are many times when they should be on but aren't.
Telsa's auto-lane keep, adaptive cruise control, auto brights and auto wipers aren't much better or worse than any other manufacturer's. Still they deserve the criticism for over-hyping and over-using AI for every automatic feature.
Well, I drive a 2019 ford fusion that has all these features (auto-lane keep, adaptive cruise control, auto brights and auto wipers) and I never faced problems using them. These features are old and stable at this point and most cars do a good job with them. Except Tesla.
The auto high beams in my Y drive me batshit. Not sensitive enough to oncoming traffic. Brights people from behind for no reason. Just generally obnoxious
This argument has never really held water; Volvo and Subaru both use vision only ADAS on various models, and did it long before Tesla too. Vision based ADAS only became a "problem" in the eyes of some when Tesla adopted it...
The main feature of the poles is to mark the position of the road when snow is covering the usual markers. You don't need the poles most times, but during the window between snowfall and snowplowing, especially if the snow is being blown into snowdrifts further obscuring the typical "road shape" you would have to look for it the poles weren't there, they're your only means of navigation, and can be used to judge the size of snow drifts.
The poles have a standard height, and a standard distance between poles.
In places with less snow it’s some times just a reflex on the ground rather than on a pole. I think those little matchbox ground reflexes are common in parts of the US at least?
Things like this (and panel gap, and build quality, and over promising features) are why I have yet to buy a tesla, and likely won't any time soon. Even though they were the only viable luxury electric car (during my last car search), I decided to pass. I will probably seriously re-evaluate them for my next car purchase, but right now, no thanks!
BTW, I got my first demo of tesla FSD a few weeks ago. In san francisco, on market street, a place I'm sure is tested extensively. Within minutes, FSD had the car perform an illegal AND dangerous mid intersection merge followed by an extremely hard and jerky lane follow correction. lol.