This goes back to my earlier comment -
Elon mentioned this in the meeting that FSD infra is going to be 10x larger in 2023 and 10x of that in 2024. Time will tell.
2 Likes
I don’t doubt that it is improving by leaps and bounds. But that doesn’t negate the fact that I think the current hardware (HW3 and HW4) is likely not able to handle all scenarios.
For a trivial example, I drive on a curved road multiple times every day, the speed limit is 30 mi/hr on that road (the entire length of it, about 0.7 mi)). That road had many turnoffs to shopping areas. Periodically, and repeatedly, the car will trigger on a vehicle in front of it, that has turned into a parking lot of a shopping area, and do an FCW (forward collision warning) AFTER that vehicle has completely cleared the road I am driving on. It colors that vehicle as red long after it’s off the road. This is for three reasons, one, the car doesn’t know the exact curvature of the road, and two, it doesn’t have “sideways” cameras with a wide enough angle, and three, it apparently doesn’t have the processing power to remember that scenario and apply it again later. I drive there at least 4 times a day, sometimes 6 or 8 or 10 or 12 times a day. The same applies to potholes, it will drive into the same pothole every day because it can’t remember and process “hey, there was a nasty pothole there yesterday and the say before yesterday and the day before that, so I should attempt to avoid it”.
Another issue I experienced last week. The car was driving on a road that split into two, the left lane became a feeder to a different highway, and the middle three lanes continued, and the two right lanes fed into an exit. The speed limit on the road is 70 mi/hr, and It took the left lane, immediately entered a VERY curved (it was doing a 270 degree change in direction in a tight radius) one lane road, and attempted to take it at 70 mi/hr. That’s because there was no sign indicating lower speed limit. And it also couldn’t see enough ahead to estimate the curvature. I had to grab the wheel, hit the brakes, and take over at that point.
The processor is doing so many things at once, and processing the inputs from many cameras (7) all the time, and running many multiple algorithms, that it can’t always keep up. It also has to decide on the fly which algorithms to run and which to not run. My experience tells me that it can’t keep up, and that it very likely won’t be able to keep up as new and more clever algorithms are added, and as new “edge case” scenarios have to be handled individually. All in real-time of course, because it can’t be any other way in this application.
6 Likes
Rhetorical questions - Will Tesla be able to figure it out ? How long will it take ? 3 years, 5 years, 10 years ?
My view is that in 5 yeas, the smaller car equipped with state of the art cameras, sensors inference engines and a great entertainment system will be at L5.
Time will tell.
1 Like
In the end, it won’t depend solely on technology. It’ll also depend on politics. That’s because, in 5 years (or 10), no software will be able to handle all the edge cases. And when they come close, someone will find a new edge case that isn’t handled. So, in essence politics will decide at what point it is sufficient.
People aren’t logical and regardless of mathematics, they care a lot more when an automated vehicle causes a death than they care about 100 human driven vehicles causing a death. Maybe even 1000 or 10000. So if automated vehicles crash once per X00,000,000 miles and human driven vehicles crash once per X,000,000 miles, they may reject the automated vehicles despite them doing 100 times better. In essence, politics.
2 Likes
I understand the general point you are making but I disagree. I don’t think regulators are bad people who play politics if there is conclusive evidence that FSD is safer but not perfect.
There are 40k deaths due to auto accidents in US. The job of the regulators is to keep the public safe. L5 autonomy will be a lot safer. In the end people will vote with their pocket books.
1 Like
I didn’t say anything about regulators! I am talking about the PEOPLE and the resultant politics based on their feelings. The regulators are usually quite rational, but are sometimes forced into doing things that the people want.
In the end we will surely have full self driving. In fact, human driving may eventually be outlawed on public roads. But “the end” is not next year, and it’s not in 5 years, it’s not even in 10 years.
Heck, we were supposed to have v2v by now, which would have made the roads far safer overall (40k deaths likely cut in half or even lower), but the automakers and the politics screwed it up big-time.
1 Like
People are rational. If technology saves lives and makes their lives easier, they will adopt. If it does not they will not. If there is an occasional accident and media highlights it, people understand.
EVs catch fire , Escalator accidents kills kids, Trains and Airplanes crash. People know how to sort all of this out.
2 Likes
I hope you’re right but I know quite a few people who can’t stand the thought of not being ‘in control’, however wrong headed that may be. Might take longer than I’d like to think for most people to see the advantages of autonomous vehicles.
2 Likes
Rented a Model 3 this week on a business trip. Took a couple of days to get into it, then I ended up liking it a lot. Autopilot is very good. Very good. Navigation was excellent. The acceleration is addictive.
Quite a few squeaks a rattles in a 2022 car with only 10k miles on it. A lot of noise on the highway. The Malibu I’m renting while my Bolt EV is in the shop (thanks, GM) is quieter on the highway and has a smoother ride. Rather have the 3, though.
4 Likes
That is wierd for a model Y or any Tesla, they are usually butter smooth and have only wind and tire noise (but plenty of that).
I am aware of Model Y that was delivered with some super squeaky wheel covers, some wierdness with the plastic tabs that held it on. I don’t recall what the fix was, but maybe on a rental that never gets taken care of?
Wierd.
The wind noise is very weird to me. Since the car is specifically designed to be very aerodynamic, why would there be so much wind noise? I would expect the opposite … that is would “slide” through the wind as noiselessly as possible. The tire noise I can kind of understand, they use wide flat tires that are filled to a relatively high pressure (42 psi recommended in my model 3), and the vehicle is somewhat heavy.
I had an issue with one of my wheel covers (loose and kept falling off), and Tesla replaced it the last time I was at a service center for an unrelated recall issue. Most service, though, has been done in my driveway by a mobile Tesla technician.
Not wind noise. Road noise. Took a group out to lunch - they all wanted to ride in a Tesla. A couple of folks remarked on the highway noise and stiff ride.
Had the bigger sport wheels. Maybe that causes it?
Anyway, I liked it a lot. I’ll definitely rent them again. Considering buying a Y later this year.
2 Likes
Perfectly good question. The answer is a few fold.
First, noise reduction is done in cars by insulating the car against noise getting in. Cars that are quiet inside are stuffed with noise insulation. The noise insulation weighs a fair amount. Think of the lowly race car. OBVIOUSLY going to be more aerodynamic than the family luxury sedan. OBVIOUSLY going to be lighter, also. And is it quiet inside? No.
Second, the tires Tesla uses have about half the rolling friction of “regular” tires. This is to get more mileage from the battery. The thing about Physics is the only way to make a tire more efficient is to make it stiffer. Rolling energy loss in a tire comes from the deformation of the tire as it rolls. This is why railroads (SUPER efficient by the way) use steel “tires,” how much do you think those deform as they roll, compared to balloony rubber tires? Not much.
The stiffer tire on a Tesla produces two things 1) a stiffer ride, which is good for performance handling, but which feels to the passengers, well, stiffer and 2) road noise. Inflate your Tesla tires 50% higher pressur than you would inflate a “regular” car tire, and it is stiffer and makes more noise rolling.
The good part of not carrying the weight of noise insulation and using noisier, higher efficiency tires is you wind up with a higher performing car: more miles per kWh in the battery (essenentially higher mpg), better more controllable handling at higher speeds. But with performance comes stiff rides and noise.
R:
3 Likes
Elon: The answer was much simpler than we thought. We were too dumb to realize how simple the answer was
Most of the media missed the Elon statement that v12 is very good. He stated that the FSD problem in the end was much simpler than they thought. Tesla is also increasing their DOJO capacity by 10x this year and now have an insurmountable lead with data, compute and storage.
L5 in 5 years for FSD seems likely in my opinion.
3 Likes
People are anything BUT rational especially as it pertains to risk and fear.
You can even take investing for example. There are people that will NEVER buy stock because of fear. Certainly they risk a substantially lower standard of living but the risk of loss - and more importantly - their inability to avoid or influence that loss, overrides all reason.
FSD will be the same for many people. The fear that keeps people away from it or otherwise results in them opposing it is the same fear that drives much of our collective irrational behavior and fears (fear of “others”, the government, deep state, big cities, investing, etc.).
2 Likes
People still drive. They are able to sort out the cost benefit analysis. They don’t sit at home due to fear.
1 Like
Yes, they drive (and will continue to do so in many cases) because they have a false sense of control over their environment. They lose that with FSD. No one thinks they are a below average driver. We all think we have the skills to avoid an accident. Reality doesn’t matter when it comes to fear avoidance.
What date do you expect it to be introduced to the market? I still think 2025 is not going to happen and 2026 is a stretch and very unlikely … I’ve been using FSD on both a HW3 model (mine) and a HW4 model (one of my kids), and while it is terrific, and it is VERY impressive, it still isn’t really F or S yet.
2 Likes
v13 reviews are good.
I believe is that Robotaxi starts in 2025 end in few cities in CA and TX. It is heavily trained on data in those cities.
1 Like