No other car manufacturer will give you new software features for free either. New free features that keep my car feeling new is worth it over some minor complaints about the new UI.
No other car company has taken away features with updates either.
Agree with
@linux-works. You need one CEO for a startup; totally different for a mature company and very, very few of them could transition from one to the other. In fact, the VC has tremendous respect for those who step down and move on to other areas where they could deploy their skills the best.
Elon was amazing at creating the vision, essentially, creating a whole new market segment, proving that EV are viable, etc. For that, he has my gratitude!
However, opening a market is very different from scaling a company. Scaling requires a totally different skill set and that is where most startups fail.
@TexasTezla still drinks the cool aid and he is right that Teslas are selling like hot cakes. However, there are some troubling indicators - quality is going down (it was not great to begin with); service is abysmal; lack of focus (EV? FSD? Mass manufacturer? Truck or car? Van maybe?); ignoring customers; spatting with regulators; simply doing stupid stuff. Again, startup CEOs are so powerful at the early stage because they challenge everything and one or two things pan out. You cannot do that when you have a bigger company.
Elon clearly does not have the experience, state of mind, or maybe even desire to run a large company. He should have stepped down _before_ the ramp up and let someone else who have done similar thing before do it for Tesla. Elon would have been much more successful and beneficial for the society if he had fully re-deployed his talent in SpaceX and similar projects. I still maintain that Tesla is making a mistake bundling EV and FSD; spin-off FSD would be a great place for him to run.
Tesla shorters’ saw all of that but they did not factor in that nowadays it takes longer for those things to unfold because of the fan base phenomenon. See what happens with GameStop.
I know that we are getting off-topic but V11 is actually a symptom of a deeper, structural problem at Tesla. Once a critical mass of people realize that the events will unravel very, very quickly. It pains me to write that because I love the idea, the vision and the mission. But if I, as a lame person, can see those issue, I am sure that many people in other car manufacturers strategy departments see them as well.
GameStop is a garbage stock propped up by fan enthusiasm. Tesla actually has valuable intellectual property that will continue to keep the company going, even if they do make serious mistakes. The company is still way ahead of the competition in many areas. As screwed up as Tesla is right now, most of the competition is much worse. Sandy Munro knows the car business very well and he's pointed out the established companies are moving too slow and have no way to catch up as consumer demand grows.
I do think FSD is a dead end. I just don't see the dream of completely driverless cars happening anytime soon. Even if the technology gets there (maybe, but they aren't there yet), regulators are going to be cautious about approving it, especially if there are some serious accidents as a result of FSD tech.
As a driver assist aid I think FSD tech will be valuable, but it will never be capable of taking over 100% of the time.
Exactly, their UI team outpaces the FSD team. They act as if the FSD is way more mature than it actually is.
I am questioning if FSD, in its current form, is even possible. If we look at the aircraft, where they have FSD for decades, the environment is much simpler to model and, due to the regulator, all of the manufacturers move at the same pace.
With cars we have a much more complex, almost chaotic, environment. You have pedestrians jumping on the road, debris (Tesla still does not know how to avoid those on the highway), potholes, bicycles, emergency vehicles, road closures … While there are rules on the road, there are a lot of informal agreements that, sometimes, go against the rules so that the traffic goes smoother. For example, how many times at a 4 way stop you just look at the other driver and silently agree who would go first? In a lot of states the rules is that the person on the left should give way; in reality, whoever comes first goes first. How many times you look at the person on the sidewalk for visual cues if they intent to cross the street? There is so much judgement and human interaction that I doubt they could model it in AI anytime soon. They cannot fully solve the highway yet (fantom braking anyone?), let alone city driving.
Had it been like airplanes and the cars could “talk” to each other that would have been a much easier and possible undertaking. While we have 2022 Tesla and 1964 Mustang on the road the FSD remains a driver’s aid, nothing more.
And I am not even touching the legal and moral aspect of FSD. Even if they solve the technical issues I am not sure if I want an FSD car.
I was at Boeing when TCAS came along. That's the system that allows aircraft to talk to one another and avoid collisions. That was around 1989, it's now standard on all commercial aircraft. Back then there were jokes that the latest flight deck consisted of one pilot and one dog. The pilot's job was to feed the dog and the dog's job was t bite the pilot if they touched anything. That's how automated aircraft were by 1990. Today all commercial aircraft over a certain size (bigger than a small commuter plane) must have two pilots.
It's a more regulated industry than the automotive industry, but I don't see regulators going for allowing FSD without a driver anytime son.
Interestingly, I used the FSD as driving aid last two winters and I absolutely loved it! There were occasions where I could not see the road - thick fog, complete whiteout, covered signs and marking - and this thing was showing even the lane markings. I guess they are reflective for the LiDAR.
I loved to be able to “see” all that but to let it drive - no way in hell! Just like a teenager - the moment I tried it it went up to the speed limit
Tesla doesn't use LiDAR. They used to have radar, but got rid of that. It's 100% visual. Which I also think is a mistake.
Exactly. It is fundamentally true that machine learning and current AI techniques have never been used to successfully solve a problem like this. Will it work? Nobody knows. The fanbois will say 'of course it will' and we are all secret oil company shills for saying otherwise. But they are wrong. I don't know if it will work. They can't know if it will work. Tesla does (or should) not know if it will work. It might work, but just as likely it will not.
There is also the reality of any software development that the first 90% of the 'finished' product takes about 10% of the development time. Software development progress does not accelerate the longer you work on it. It slows down. Hard.
But the possible future existence of real FSD is a horrible excuse for bad design decisions today.
Not just software development, all engineering works on this 90/10 -> 10/90 rule. I first heard of it back in the 1970s with high end stereo equipment like Macintosh. Companies like Sony and such were able to get to 90% perfect sound reproduction at an affordable price, but to get that last 10% out of the system took jumping from a $1000 stereo system to a $25 K+ system.
They probably are over 90% with FSD, but that last 10% is proving to be a nightmare. I've said it before, the big advantage humans have over computers is our ability to quickly filter out irrelevant things. Many accidents are caused by humans not filtering out things that they should have, but we do a great job of it 99.9% of the time. Computers have to filter everything manually. They can do it faster than humans can, but they still need to do it.
There is a test of an AI called the Turing Test that was originally proposed in the 40s or 50s. The concept is a true AI can't be detected by a human over a period of a few minutes. Some AIs are pretty good short term, but somebody can confuse them by doing something unexpected.
Human drivers still have advantages over AIs. If our field of view is bad, we can move the sensory unit (our eyes) to a different spot. I once drove in fog so thick I rolled down the window and listened for other cars at a stop sign. I only went ahead when I was sure there were no cars coming. An AI can't do that because they don't listen to the environment.
Sun glare can disable cameras, especially this time of year when the sun is low in the sky. Humans have visors and sun glasses to help manage.
There are also malevolent actors who could mess with vehicle AIs. They could do various things to jam sensors in ways that will make FSD vehicles just stop and sit there confused, but would allow human drivers to at least get off the road.
Human drivers are fallible, and some human drivers are very poor. AIs driving can probably manage most problem areas than humans, but there are still edge cases where I'm not sure FSD AIs will even be able to manage properly. During the long transition period (decades because non-FSD cars will be around that long) FSD cars will need to deal with the idiot humans driving other cars as well as natural events and humans and animals outside of cars.