Consumers want autonomous self-driving cars. So Tesla creates marketing videos and then calls their glorified lane assistance and distance keeping assistant an "autopilot".
And now, Tesla buyers are frustrated that the software isn't ready yet, or dying due to bugs in said software.
But Tesla gets to dodge legal responsibility by writing into the fine print of their EULA that the so-called "autopilot" isn't actually meant to be used like what autopilot means to most people.
If anything, Tesla has amazing marketing.
This is necessary for human progress. Gutenberg with this idea that will make the investors rich. Colombus stands up and says: I am going to the Indies across the Atlantic, it is going to be easy! Or Magallanes crossing the world, or going to the moon, Knuth creating Latex or the AoCP as a "semester" project, Artificial intelligence would be solved in 10 years...
In the end Gutenberg almost bankrupted his inventors and died penniless, Colombus faced riots and almost starved, most of the Magallanes' crew perished, going to the moon took so much more money, Knuth project took his entire life...
But the courage paid off for society as a whole because they got involved and took responsibility and risk on their lives.
It has always been easier to criticize and risk zero, while cynically putting all progress down.
Tesla or Elon is not perfect, but no human being is, if you wait for something new to be perfect before is ready, then you will never get it.
Progress is not automatic, needs people that believe on it, probably even delusional people.
I am a customer not stunt double.
Tesla is compared to the industry average that is so bad that it kills a 737 load of passengers every hour. Tesla is better than average in that industry.
It is fair to debate why and whether or not Tesla could be even more above average, but that is the baseline.
Both autopilots get sometimes confused and drive into solid wall.
Boeing did recall, fix the problem and is now trying to restore trust.
Tesla just does whoopsie, we are startup, need to fix bugs. Or blames customer: "please do not use our $20k autopilot as an autopilot"
If Tesla takes safety seriously, it should at least rename its software.
I don't know where you buy your autopilot, but even with the recent price increase, autopilot is just $7k.
I doubt the self driving feature drives many sales. It does look flashy and even in its current form it can be very good. Do I think they can deliver self driving cars? No, but I also don't think anyone else can. What they have delivered is a car that can relieve the stress of driving and counter being distracted while driving long distances
As far as being frustrated by the software I am more annoyed how long it has taken for quality of life features to get put in, namely my number one complaint is full blue tooth integration with smart phones. While they finally added the ability to navigate a playlist on screen you still cannot select the playlist you want; or any other selection outside the list like artist and album. I dislike their fixation on games and other features only useful while parked.
Everyone who has ridden in my car had two comments, first being about how clean the interior looked and second being how it felt when accelerating. Every once in a while I will let the car drive and I explain it as letting your highschool kid drive but you have immediate ability to take control. About the only negative comments I get are the door handles which some say are cool but take use getting to but the top one is always pricing.
When I had a Model S loaner, I was blown away how much they've simplified just from S to 3: wide open interior, reduction in the [already spartan] amount of physical controls. (I also briefly had a Kia loaner for that same visit, and I felt like a monkey trying to operate an Apollo spacecraft, unable to find the switches and dials for everything.)
Tesla, whether for cost cutting or genius, simply said "do you really need all of these manual/physical controls, or can it be A) done automatically via software, or B) put onto a single screen?", and it, in my opinion, has worked quite well. Yes the "AI" wipers took time to come around, and you hit the occasional UI bug, but it's far and away a very acceptable trade off for a magnitude more space in the cabin and (if you appreciate such things) clean, simple, lines throughout. Even the silly flip down LED visor mirrors delight people the first time they use them in my car.
AP/FSD may never get where they claim, but I've taken many-hundred mile trips with AP on (on highways), and even my drive to work is immensely better with AP in long stretches, so I consider it well worth the [AP] cost. As ridiculous as it sounds, all those micro-movements on the road do add up and take a toll on you during drives.
I actually drive fairly conservatively (turning into my "slow lane is fast enough" father), but for demos and when I do really want to pass someone, the acceleration and instant power still marvels. :)
The benefits of the touchscreen controls (nav, settings, media player) are that they suck much less than most manufacturers, and with over-the-air updates can suck even less over time.
Driving the S never felt weird to me until I owned the 3 and went 'back' to the S. I was simply overwhelmed at how many buttons there are, everywhere, for everything. There are something like 4 or 5 stalks coming out of the steering wheel, somehow. So many buttons ON the wheel (reminds me of my old Golf; it's great when it's your only car, but when you switch cars it's impossible to remember which buttons are for cruise, which are for the stereo, which are for calls, etc). Which is pretty shocking when I remember the first experience in an S surprised me at how SIMPLE the interior was.
What I've noticed, from Tesla groups, is that virtually nobody is disappointed with autopilot. People who like it, LOVE it, and wouldn't buy the car without it. Those of us who don't think it's all that great, either don't buy it, or bought it cheap (I bought it during the 1 week it cost $2000, as a "why not"). IMO, it's too aggressive to center the steering, too late to brake (waiting until the car directly in front of you begins slowing), often closes the gap too aggressively, etc. Don't get me wrong, I am an aggressive driver, but the AP is too.. "abrupt" for me. When I'm in traffic I like to ride the wave a bit and let gaps enlarge and slowly shrink.
The car is fantastic without autopilot, and it really changes the driving/car ownership experience, and in an almost entirely positive way.
The automatic wipers used to suck; they're great now. The addition of Spotify has made me SO happy. Seat memory based on which driver's phone is on the driver's side when the car unlocks is magic, I'm glad they added it. In general, the phone-as-key is the best feature ever for a 2-driver car in particular.
Things that I wish were buttons/switches:
* Mirror controls.
* Glovebox
* Wipers, probably, though it's mostly fixed now.
And to be fair, the mirrors are only because I've swapped mirror glass 3x now (factory -> suma wide angle reproduction mirrors -> tesla OEM european mirrors) and you never get your adjustment right on the first try.. and we have to do it for 2 drivers.. clicking through the screens while driving is a tad unsafe. But then, once they're set, you pretty much forget them, so I can forgive this. Would be hard for someone borrowing the car to deal with though.
And I just don't see that as much of an everyday advantage or selling point, just like you also pointed out.
Now how are my puts doing...
I recently gave a "state of the art" talk about optical flow and I found it extremely disappointing that current technology still struggles with recognizing shiny surfaces and fences, despite 16 years of hardware progress.
But most AI researchers appear to be more interested in re-inventing the wheel on a GPU instead of doing actual mathematical algorithm research.
Tesla is doing marginally better than the average due to their custom hardware, which grants them more than the 10 GB that people using NVIDIA GPUs are limited to.
But still, Tesla, Skydio, and NVIDIA appear to be using optical flow AI architectures that are fundamentally incapable of performing well, due to being based on the (mathematically provably wrong) assumption that a scale pyramid works for optical flow detection.
Here's the Sintel ranking: http://sintel.is.tue.mpg.de/results
And if you look at the AIs, they all repurpose this paper: https://arxiv.org/abs/1709.0237
However, you cannot store one optical flow per pixel, because the solution is not unique. For example, that's why Tesla and Skydio have problems with reflective surfaces. You see both the surface and its reflection move, but the AI can only store 1 result out of those 2 movements.
In my opinion, the solution is to use Fourier methods, which were shown to work well in a 2000 dagm paper called minimum values of the structure tensor. But nobody has even attempted to port that to GPU yet or use it as part of AI.
[1]: https://electrek.co/2020/03/04/tesla-car-find-larking-spot-r...
So it's only misleading due to a common misunderstanding.
That's the buyer expectation for the word "autopilot" that I was referring to.
EDIT: Apparently, the sensor package is still called "self-driving hardware", which I'm sure creates a certain expectation if you purchase the "self-driving hardware & autopilot software" bundle.
But the disclaimer on the bottom is pretty noncommittal:
> "The activation and use of these features are dependent on achieving reliability far in excess of human drivers as demonstrated by billions of miles of experience".
On one hand "are dependent on" actively avoids stating that they're there but the "billions of miles of experience" is clearly aimed at suggesting to buyers that they are. Which they are not since the car cannot handle many/most situations that a human driver is expected to handle to be called "a driver".
If you set an autopilot to hold a heading, it will hold a heading. You don't have to supervise it, you can concentrate on other stuff.
If you set a tesla autopilot to keep in a lane, you have to supervise it because it might fail to do so.
Granted, the latter problem is a harder one. But the point of an autopilot is that it manages to do want it promises to do on its own without constant supervision.
The important bit, which is apparently not understood by the public is that it's only used as an aid in controlling the vehicle. You still have to monitor it and be prepared to take over if it doesn't behave properly. I once flew an airplane, set the autopilot and the airplane decided to go into a rapid climb unexpectedly. That's exactly equivalent to the lane assist failing to keep the car in the lane.
That said, it's clear that because the public doesn't understand what it means to be an autopilot it's a bad choice of name. Or at least Tesla should be spending a _lot_ more time educating people on what it means before they use it.
Not _strictly_ true; faults can occur and the system always needs monitoring. Strong turbulence will also disconnect the autopilot (with appropriately attention-grabbing aural and visual alerts).
The cruise control is much more an autopilot than the lane assist, which is just an assist but wrongly advertised.
As long as that other stuff includes paying attention to where you're going. If not you're bound to collide with something at some point.
That's autopilot.
Although it can be argued that with technological progress the meaning of the word has also shifted over time together with the expectations.
I'm neither frustrated or dissatisfied with my purchase. It's the best car I have ever experienced and it just keeps getting better.
That cost is already sunk, Tesla can't recuperate that, and everytime electric cars become a little bit more common, the coolness factor of the electric engine decreases which means the value of that sunk cost goes down. There's two things that save Tesla, the first is plain consumer attitude. Some (many?) consumers will simply not buy an ICE car anymore if they can afford not to, which means the other manufacturers have to compete on Tesla's playing field.
The second is cheap innovations. Tesla can spend a couple tens of millions, and maybe $1000 in hardware, and get something that looks like an AI into their cars that adds maybe $10k of coolness value. Add fancy doors, a big LCD screen, flush doorhandles, and whatever other little thing you have to spend R&D on anyway because you're a new car manufacturer so everything costs more R&D for you than the establishment, and you close the gap.
It's about putting a $80K Tesla next to a $80K Jaguar or whatever, and being able to justify why they cost the same amount of money when the Jaguar is better made.
edit: Note this whole story is about the Model S, which was Tesla's bootstrap model. The goal of Tesla is to produce a car that doesn't compete with Jaguar, but with Volkswagen (or Ford).
ICE engines have larger complex elements that can fail for more reasons, but the most complex part is just this big block that they pour into a mold. After that's done you just have to mount all the little components onto it that the suppliers have refined the tolerances on for at least 20 years.
But the ICE and the electric engine are just peanuts compared to the material cost of producing these huge batteries.
That being said, the customer experience of owning and driving a Tesla far exceeds the traditional options. I fail to find anyone who is unhappy with their purchase. They love their cars and there are still hundreds of features that Tesla has where other manufacturers are barely catching up.
Add in the ecosystem of solar/powerwalls and superchargers along with supplying the wider energy market, and this company has a great future ahead.
Given that autopilot for cars doesn't and never has existed it is likely that it means different things to different people.
When I'm driving a car, my function is pretty much a glorified lane assist. If it can do that for me then I'm sold on it being an autopilot. I don't need my car to actually be smarter than I am.
"The nation’s top safety investigator slammed Tesla on Tuesday for failing to take adequate measures to prevent “foreseeable abuse” of its Autopilot driver-assistance technology" https://www.latimes.com/business/story/2020-02-25/tesla-auto...
And the evidence in that article is that the car is basically working on autopilot. So the brand name is quite reasonable. If you expect any transport system not to have accidents you are crazy - all modes of transport have accidents from time to time.
Also; a tired point; but Tesla's autopilot is going to learn from each fatal accident. That is a massive difference from humans; who stop learning after fatal accidents. Fatal accidents are not going to be a major long-term problem here which stands in sharp contrast to human drivers. It isn't a particularly bad sign as it means Tesla's autopilot will have learned a valuable lesson due to someone-who-is-not-me's misfortune. I'd happily pay money for that. Deaths from bad human driving are meaningless tragedies. Deaths in a new but engineer-able system are extremely meaningful tragedies and are almost good news for everyone who isn't immediately involved.
So that'll be a completely different hardware than what Tesla currently sells as their "self-driving hardware" package.
Somebody has already linked an article showing that people in these cars are taking their hands off the wheel and playing videogames while 'driving'. These cars are already self-driving by a reasonable standard, and is certainly fairly described as an autopilot system. Your standards aren't normal.
It's not that others have too high standards, it's that you have low standards and they make basically every mundane thing, that you use every, day a "self-whatever" thing.
That guy died in a car crash while doing that.
Don't be shocked if they[1] do write guidelines for driver attention monitoring that Tesla has to adhere to. Also, don't be shocked if they apply across the board, even to cars that do not have autopilot like features.
[1] By "they" in this case, I do not mean the writers of that report, as that was the NTSB. The NHTSA will likely investigate the concepts more holistically.
I don't mind it being called marketing. But I'd call it overall experience design. Or magic. You really feel it!
I notice the quality of the interior isn't great.
But I do notice that it's quiet and has a lot of torque obviously.
If it wasn't for the emissions I prefer being in a Merc any day.
I agree that's boring.
But it only is boring because BMW has such a crushing market share that the occasional non-BMW car feels like a breath of fresh air. Tesla feeling nice due to their low market share is, however, the opposite of what this article argues for.
My only disappointment was that the giant LCD panel in the middle didn't refresh at high frame rate. I hope they fixed that because I'd expect the car from the future to be able to scroll a map smoothly.
Pary tell, in what way are they better? I've been in both a Model S and 3, and neither of them have a great fit and finish. My VW is far more polished than either. High level Mercs or BMW are way beyond.
I wonder if people saying stuff like this have ever been in a nice car?
Almost everyone else I talk to outside of software seems to pigeonhole AI as magic — either they’re believers who think it can pass the Turing Test, or they’re non-believers who refuse to accept it can do what is in 2020 the AI equivalent of Hello World.
So no, I don’t think most people do know; rather, I think most people either think it’s impossible or perfect already.
So at the very least, the authors of that article were unaware that the limitations of Tesla's autopilot make it effectively equal to the i3's assistance systems for everyday use.
Not sure if Tesla's approach to that is more customer friendly. Their cars are basically iPhones on wheels, they are locked down to the same extent. iirc you cannot even get replacements for the most car parts. With traditional cars however, you can basically go everywhere and get your car fixed. No matter if it is a brand dealer or third party repair shop.
edit: fixed typos
Haha. My buddy just waited for Tesla to finish repairs on his Model 3 for 4 months and in the mean time had to have a rental. The reason? They couldn't acquire certain bolts.
Give me "exploitative dealer networks" that can source parts properly to Tesla's outrageous incompetence of a service network.
The first mass market Tesla (Model S) came to market in post app store world. By that I mean, not only is it possible but it's obvious that vehicles can and should be diagnosed and firmware should be updated over the air. The only reason they were able to throw away the dealer model is because they could collect analytics, perform diagnostics and firmware flashes over the air. Without this advancement, Tesla would need significantly more physical presence akin to dealerships (and in fact, they kind of are building a light showroom/repair network as they grow so the dealer model isn't all that crazy in some respects).
Traditional automakers have been around since before OTA was possible or practical or anyone had even thought of it, so dealers were at least necessary at one point in time to perform these tasks. Now that it is possible, and many new vehicles even have wireless internet, very few traditional automakers actually perform remote diagnostics and even fewer update more than a couple ECUs (Electronic Control Units) over the air.
There are a number of reasons for and against this, but at the end of the day they were forced into a situation where they needed dealers and for that reason they also have the luxury of offloading many of these tasks onto the dealers. This luxury is actually very important because a major OEM has hundreds of models of vehicles that they need to support. Tesla obviously has far fewer so they can afford to take on some of those diagnostic tasks and of course they have to because they don't have franchises with technicians in them.
As for success, I believe Tesla is very successful for a variety of reasons (too many to list, but I'll mention some of the major ones):
1. They care about design. They're not the flashiest or most radical designs, but the designs are very thoughtful, tasteful, elegant and minimalistic -- all of these are qualities that are fashionable at the moment.
2. The cars are technology first... they come with a lot of useful technology that people want because it seems obvious to have these things in contemporary cars. Most other vehicles in the Model 3 price range are technology last -- the technology that seems obvious is a paid upgrade.
3. They are optimistic and progressive. Even when someone knows that some model 3s are built outside in a tent and there are mounting quality and repairability concerns, they don't care because they want to support the flier that just might make a difference in the world.
With Tesla, a lot of the parts are still standard, but you can't replace it with same or better part that isn't sourced from Tesla supply. Even if it's the same part.
What stops you from replacing standard parts?
There are quite a few off-brand repair shops here in Norway for things like wipers, brake calipers, regular service etc.
For example they are not allowed to order extra parts for stock (they have to "tie" them to a car in service) and they are not allowed to use aftermarket replacement parts even if compatible. I am specifically referring to car parts, not consumables. So break pads yes, break calipers no. Light bulb yes, headlight no. For example you will not find aftermarket headlights for Tesla that you can install in an authorized shop, get warranty, and ideally not discover that the car already logged the use of an AM part. In this case most (all?) aftermarket headlights begin their life as an OEM part that gets modified and this pretty much bypasses the restrictions, technical or administrative.
These restrictions may be lifted as Tesla becomes better at making cars and can dedicate resources to making more spare parts, and as they start to allow others to take a bite out of their profit by manufacturing compatible parts but we are not there yet.
Having watched Rich Rebuilds' attempt at buying a second-hand Tesla Model X directly from Tesla, I'm not convinced that a) dealerships are bad b) Tesla are the good guys.
Yeah, cars have had no innovation over time...wait, when's the last time you've been in one?
And for the most part dealers are independently owned, not some scandalous way to rip off customers. I have owned many new cars in my life (I like tech, so I lease). I've never had an issue with the dealer that wasn't pleasantly resolved. I realize that isn't the case for every person, but dealers are not universally bad, or any worse than any other business. In fact, my wife likes the brand dealer so much that she continues to purchase her cars from them based on that alone.
>The dinosars deserve to get their ass handed to them until they wake up and serve customers.
You should look around the Tesla forums and Twitter at the complaints. The 'dinosaurs' may be bad, but the idea that Tesla service is somehow better is laughable.
Who are the dinosaurs? Volkswagen sells 10 million vehicles a year. Toyota sells 10 million vehicles a year. Are you claiming that Volkswagen and Toyota are somehow not serving customers? If they're not serving customers then why are they selling so many cars?
It's not about the sales. It's about the brand.
I don't see that it's useful being emotional about it. They're car companies selling cars. You're better off being realistic.
Some people just want a generic econobox that brings them from point A to B, and others are swayed by other things, such as leather seats, ads that show a single car driving by the ocean or in the mountains, fast acceleration, an appealing paint color, high tech looking HUDs, cup holders, moon roof, etc.
well at least we used to have some diversity in design - these days it's mostly variations on the shapeless blob theme. Yes I know, safety regulations etc etc...