mdpa
Active Member
For fun, I configured the BMW IX3 and it has some basic autonomy for $2,500. However, it has a disclaimer that it is only a 4 year subscription.
Sponsored
With any tech, one company spends years figuring it out and then others just do the method that ended up workingThe other problem is assuming Rivian can’t do a better job faster.
I wish it worked this well for me. I am at 80% FSD on my Cybertruck and wife is at even less in her 2025 Y. If we didn’t have to disengage and drive manually all the time for incorrect speed limits and incorrect lane choices I probably wouldn’t even consider switching brands.Using "Hey Grok", I can simply tell it a few destinations, such as go to my doctors, then the pharmacy, and then the grocery store, before taking me home
If they’re smart they won’t raise the price. Tesla can’t get more than 15% take rate and it’s because of price. FSD should be $49/mo. Rivian A+ in current state should be $29 IMOYes the price will go up which is why, if you keep the vehicle for 4+ years and value autonomy, it is worth the single payment.
Fine, but I'm not buying a Tesla. So that's meaningless for me.The ability for an elderly person to have a car take them to the doctor is essentially here today. While no manufacturer is rated for point to point today, the consumer version of Tesla FSD backed me out of the garage and drove over 100 miles to my friends house yesterday and parked me curbside as I requested. Never touched the wheel.
Using "Hey Grok", I can simply tell it a few destinations, such as go to my doctors, then the pharmacy, and then the grocery store, before taking me home. It is not perfect, and if these were set as favorites it would probably be more reliable.
I'm hoping Rivian will do a basic version of this with the current Autonomy + and one of the reasons I'm leaning towards the R2 instead of the Scout.
There is a continuum in capabilities even within Supervised FSD. I am at the point where I will let almost anyone with a license drive my car with FSD on. It is that good.
The Model Y with one additional camera washer is driving people around in test locations as a Robotaxi today.
Rivian develops their ADAS in serbia. Tesla develops their ADAS in the U.S., has 1200+ data labelers, 500+ AI engineers. Tesla pays $500K+ in AI engineer salaries because that's what it takes to hire competent talentThe other problem is assuming Rivian can’t do a better job faster.
Agree, don’t expect them to raise prices until early next year when they have achieved the goals for this year including point to point. Then raise it before the RAP1/lidar to capture some income from those who waited.With any tech, one company spends years figuring it out and then others just do the method that ended up working
I wish it worked this well for me. I am at 80% FSD on my Cybertruck and wife is at even less in her 2025 Y. If we didn’t have to disengage and drive manually all the time for incorrect speed limits and incorrect lane choices I probably wouldn’t even consider switching brands.
If they’re smart they won’t raise the price. Tesla can’t get more than 15% take rate and it’s because of price. FSD should be $49/mo. Rivian A+ in current state should be $29 IMO
Rivian showed a lot of their technology during autonomy day. It is very outdated. Even the driving they demonstrated was very poor.With any tech, one company spends years figuring it out and then others just do the method that ended up working
Tesla tried and failed for a decade before they developed the current method. Rivian will have to do a lot of work but not nearly as many years as Tesla.Rivian showed a lot of their technology during autonomy day. It is very outdated. Even the driving they demonstrated was very poor.
The car was designed to not turn right at red lights and it did anyways (despite being told not to). When it did it, it turned into oncoming traffic. They faked a driving demo by having the car only lane center/cruise control and only turn left at protected intersections with dotted lines. The car did not need to understand traffic/pedestrians, as it did worse than what a comma.ai system does now
Any time there was a pedestrian crossing, they took over because the car did not understand. Any time the car turned right on red, they took over if the car was going to hit cross traffic.
The car lane centered, kept distance with the cars in front and then turned at an intersection with a protected arrow and a dotted line. In a lidar mapped geofence. This could never go to consumers because after 2 minutes you'd figure out the "trick" and realize the system is barely doing anything and doesn't work.
If you look at the reaction time of FSD vs chinese systems and edge case accident avoidance, it really shows that these systems are not remotely the same. Presumably they use very different methods and even if you develop a system for 5 years it shows that it's not always possible to get the same result.
You would also assume that tesla does not use any technology from the early days and that is definitely not the case. Or that tesla has zero proprietary technology.Tesla tried and failed for a decade before they developed the current method. Rivian will have to do a lot of work but not nearly as many years as Tesla.
I don’t expect Rivian to match FSD either of my Cybertruck or Model Y can do right now this year or next but it also won’t take a decade as long as the R2 fleet gets them enough data. I’ve used FSD since before it was public beta so I guess if I get R2 I’ll just be a beta tester again![]()
No they don't. This is not for mapping. Rivian when they were doing their demo they were running custom mapping and navigation to make the car basically appear like it was driving. Whereas you'd see a fail within 10s if a consumer was given this software.btw, Tesla basically uses a lidar geofence for driverless robotaxi now. They send a fleet of lidar ground truth vehicles to travel around an area before they launch driverless.
I’m guessing you aren’t familiar with the fact that Tesla did a ton of “fake it till you make it” stuff with FSD too?You would also assume that tesla does not use any technology from the early days and that is definitely not the case. Or that tesla has zero proprietary technology.
The methods are complex enough that you can't just copy everything even if you had all of the relevant information. Tesla has 100s of thousands of nvidia gpus and more than 50K of the car GPUs that they train on. Presumably the methods Rae not easy to copy even if you know everything they do.
No they don't. This is not for mapping. Rivian when they were doing their demo they were running custom mapping and navigation to make the car basically appear like it was driving. Whereas you'd see a fail within 10s if a consumer was given this software.
Tesla uses consumer level maps for robotaxi and the only thing they do is geofence where the car drives.
The only limitations are around freeway driving/max speed, avoiding extremely hard unprotected turns, and not crossing a train track.
They did. Other than the fake video at the beginning and fixing independent issues with mapping, they have stopped doing that for years. it has been 6 years probably since tesla did anything like that.I’m guessing you aren’t familiar with the fact that Tesla did a ton of “fake it till you make it” stuff with FSD too?
This x10000FSD should be $49/mo. Rivian A+ in current state should be $29 IMO
Tesla was faking a lot of things in FSD. Even today it doesn’t read signs. It “recognizes” speed limit signs (because of big numbers) and shows them on screen. Has no ability to read or understand context if the speed is “trucks only” or anything like that. Also completely ignores those signs and uses whatever is in the map data, which in my town has suddenly gotten terrible in the past few months. Passes 20mph signs, shows them on screen and still displays 35 mph as the current speed limit. Since all we have are speed profiles it thinks sloth should drive 35 mph. If the map data here was great I would never have figured this out. But it’s not at allThey did. Other than the fake video at the beginning and fixing independent issues with mapping, they have stopped doing that for years. it has been 6 years probably since tesla did anything like that.
Tesla only updates maps a few times a year and even then, the only reason tesla drives as well as it does is because in real-time the car has fast reaction time and fixes issues without you noticing
Waymo doesn't read road signs. It's locked under patent by mobileye. If you own a car with a speed sign reading (using mobileye patent) you'll find it doesn't work. The car doesn't read a lot of non standard looking speed limit signs and it also misreads the miniumum speed limit 40 signs in GA. By comparison, whatever method tesla invented seems to work better.Tesla was faking a lot of things in FSD. Even today it doesn’t read signs. It “recognizes” speed limit signs (because of big numbers) and shows them on screen. Has no ability to read or understand context if the speed is “trucks only” or anything like that. Also completely ignores those signs and uses whatever is in the map data, which in my town has suddenly gotten terrible in the past few months. Passes 20mph signs, shows them on screen and still displays 35 mph as the current speed limit. Since all we have are speed profiles it thinks sloth should drive 35 mph. If the map data here was great I would never have figured this out. But it’s not at all![]()
This has already happened...I am still waiting for the coast to coast FSD drive that was promised nine to ten years ago