Rivian May Be Much Closer to FSD Than You Think

[sunydrm]

China cars are making rapid progress so hopefully Rivian will be as fast as them at least.

These companies have thousands of employees in china (lesser paid) working on this technology. They are also spending billions of dollars like Tesla is. They also have many more vehicles than rivian does. They are not even close to unsupervised even in small scale like FSD is.

That would make me think Rivian is further than we think.

My belief is that rivian will show universal hands free in december and it will be like any early version of FSD (maybe on the surface better because of E2E head start). When it finally gets to consumers you will see all of the issues with poor map data and speed limit data rearing its ugly head. The idea that Rivian is not beta testing on consumers is false. This is purely a skill issue.

What people don't realize is tesla has equally incorrect map and speed limit data. They are also restricted by a reading road signs patent which Rivian will have the same issue. All cars in europe have speed limit sign reading and tesla might be the only one using their own method. The difference is that tesla has years of smoothing out the driving and working around these incorrect maps.

I would bet by the time UHF is no longer a gimmick, will be mid 2027. The issue with rivian pursuing level 3 (eyes-off) is it is an entirely different technology path with very little reward over level 2++ (too restricted and waste of money to develop) and requires rivian to certify autonomous operation in all states and take liability.

Sponsored

 

[Gen(R3)Xer]

In the medium/long run it won’t matter. I think it’s a mistake car OEMs will be able to monetize driving assistance software.

If anything, level 4/5 driving will make car ownership obsolete for many people.

Waymo will likely soon provide a couple hundred dollars/mo subscription for 1,000 mile/mo of travel. That will cover the cost of energy, wear/tear, insurance, etc. And the more people who subscribe will make it cheaper for everyone.

Car OEMs like Rivian will have a tough sell if services like Waymo take off.

Waymo is now allowed to operate in greater Bay Area, LA, and San Diego.

In regard to Waymo and other ride sharing services. I know people have been saying this for years and it does appear that younger generations have less desire to own a vehicle mainly due to the expense and that they do most of their socializing online, BUT there’s no way that any company will convince the majority of drivers in the US to give up their vehicles unless they give away the service for free and even that might not be enough.

There are way too many cases where people want to keep their stuff in their vehicles (sports equipment, stuff for work, etc.), go on long trips and not worry about mileage, not wait for a ride share service to arrive (especially if they live in a rural area), hauling their pets without worrying about potential damage to the vehicle, dealing with last-minute emergencies related to children or the elderly, and/or wanting to customize their vehicles.

If it’s in an urban area where having a vehicle doesn’t make sense because of decent public transport and the expensive of ownership (payments, insurance, fuel, and parking) then I think autonomous ride-share services will be more successful overall.

 

[sunydrm]

People forget that eventually UHF and these level 4 systems are not driving assistance software. They are driving replacement software which allows you to own your own car and road trip while you are sleeping.

Even if you prefer to drive a car consumers won't buy a car without this. And the problem is it costs billions a year to enable this solution. Rivian is in a tough spot where they do not have both the money for R2 and to enable self driving. it is one or the other

 

[ATLRivvy]

People forget that eventually UHF and these level 4 systems are not driving assistance software. They are driving replacement software which allows you to own your own car and road trip while you are sleeping.

Even if you prefer to drive a car consumers won't buy a car without this. And the problem is it costs billions a year to enable this solution. Rivian is in a tough spot where they do not have both the money for R2 and to enable self driving. it is one or the other

Im really starting to think people that post on these forums don’t actually own Rivians.

I just drove one 400 miles today. Rivian doesn’t have a market competitive L2 ADAS system right now. They are not remotely near L3/L4. RJ is turning into Elon with the over promising.

The R1S and R2 will both be either sunset or on entirely new generations before they have autonomy in market. Rivian currently isn’t even on par with Ford and GM were doing in 2022 with autonomy

 

[Viet658]

Im really starting to think people that post on these forums don’t actually own Rivians.

I just drove one 400 miles today. Rivian doesn’t have a market competitive L2 ADAS system right now. They are not remotely near L3/L4. RJ is turning into Elon with the over promising.

The R1S and R2 will both be either sunset or on entirely new generations before they have autonomy in market. Rivian currently isn’t even on par with Ford and GM were doing in 2022 with autonomy

There are plenty who know how bad Rivian ADAS is currently. Hence the whole market for comma. I have a AI4 Tesla and would love for rivian to get to that level in the next couple of years. Really curious what rivian announces on 12/11.

 

[sunydrm]

Im really starting to think people that post on these forums don’t actually own Rivians.

I just drove one 400 miles today. Rivian doesn’t have a market competitive L2 ADAS system right now. They are not remotely near L3/L4. RJ is turning into Elon with the over promising.

The R1S and R2 will both be either sunset or on entirely new generations before they have autonomy in market. Rivian currently isn’t even on par with Ford and GM were doing in 2022 with autonomy

I think the issue is people blame Tesla for beta testing on customers and deaths using FSD (find me them?) across 6.5 billion miles.

And somehow Rivian is taking the higher ground in delaying the release of half-assed software purely due to a skill issue.

When FSD was bad back in 2020, you had to have a safety score of 100 to get access. RIvian does not have any sort of safety score software at all. They will beta test on consumers. Mark my words.

 

[portdirect]

Either too much turkey, or there’s somthing going on above the windscreen camera cluster…

https://www.facebook.com/share/v/1LS8dEzXZ7

(not visible in car in 1st two cuts, but then there’s something odd about the camo after that, and reflections that remind me of the ‘taxi’ lump…)

 

[sunydrm]

Either too much turkey, or there’s somthing going on above the windscreen camera cluster…

https://www.facebook.com/share/v/1LS8dEzXZ7

(not visible in car in 1st two cuts, but then there’s something odd about the camo after that, and reflections that remind me of the ‘taxi’ lump…)

this is lidar. Which we will need to see how hidden it is.

lidar is 100% not needed for L2++ and cameras are more than acceptable for level 3 (eyes-off) driving only

Lots of manufacturers are removing lidar because they are both not using it and it makes no sense for level 2. Only for level 4 which rivian would need to deliver

 

[portdirect]

this is lidar. Which we will need to see how hidden it is.

lidar is 100% not needed for L2++ and cameras are more than acceptable for level 3 (eyes-off) driving only

Lots of manufacturers are removing lidar because they are both not using it and it makes no sense for level 2. Only for level 4 which rivian would need to deliver

I don’t agree with your assertions re required or not - thats an implementation choice. However I‘ve previously offered back in February to eat my hat (or head ware of another’s choosing) if lidars not present in r2.

 

[sunydrm]

I don’t agree with your assertions re required or not - thats an implementation choice. However I‘ve previously offered back in February to eat my hat (or head ware of another’s choosing) if lidars not present in r2.

L2++ is driver supervised. You can achieve 99.99+% reliability with cameras only. It is unknown whether lidar really adds something that you can't do with cameras until you test at scale. Rivian would have to surpass likely 100K miles per safety critical intervention (years away) before they can even prove that lidar adds any safety over tesla's implementation. Which assumes FSD never gets better.

L3 typically requires good weather and a lead car which 100% does not require lidar in any scenario.

Many chinese cars with lidar do not even use them. For level 4 it is even unknown whether lidar would provide anything useful that you can't get from HD radar.

 

[portdirect]

L2++ is driver supervised. You can achieve 99.99+% reliability with cameras only. It is unknown whether lidar really adds something that you can't do with cameras until you test at scale.

L3 typically requires good weather and a lead car which 100% does not require lidar in any scenario.

Many chinese cars with lidar do not even use them. For level 4 it is even unknown whether lidar would provide anything useful that you can't get from HD radar.

Again it’s implementation choice - for example the first certified L3 vehicle in the USA uses lidar, so clearly Mercedes Benz thought it was useful.

 

[sunydrm]

Again it’s implementation choice - for example the first certified L3 vehicle in the USA uses lidar, so clearly Mercedes Benz thought it was useful.

it's because they don't have the skills to do monocular depth estimation. If you are doing backbone vision driving (which rivian will be doing with L2++ end to end) then doing level 3 with vision only is trivial

Mercedes' lidar placement is low to the ground. it is not optimal for anything other than level 3. It would not help level 4. The technology that enables mercedes to have level 3 (a gimmick) would not help with l2++ or leve 4 driving.

Also I would bet most of these lidar systems in the car are purely for marketing. They are low resolution, low FOV and do not have cleaning. These are nothing like the 360, high resolution spinning lidars that a standard AV uses. Those are such high quality lidars that they are classifying objects directly from the lidar images


I bet any amount of money that rivian advertises heavily that R2 has lidar on autonomy day. And talks about how lidar will allow them to deliver self driving. Tesla would not have self driving any sooner if they had lidar.

In fact, all Tesla has to do is remove the safety monitor and then you will see RJ making more claims about autonomy. Why would you buy one car when a competitor's technology will be already here, even if at small scale

 

[portdirect]

it's because they don't have the skills to do monocular depth estimation. If you are doing backbone vision driving (which rivian will be doing with L2++ end to end) then doing level 3 with vision only is trivial

Mercedes' lidar placement is low to the ground. it is not optimal for anything other than level 3. It would not help level 4. The technology that enables mercedes to have level 3 (a gimmick) would not help with l2++ or leve 4 driving.

Also I would bet most of these lidar systems in the car are purely for marketing. They are low resolution, low FOV and do not have cleaning. These are nothing like the 360, high resolution spinning lidars that a standard AV uses. Those are such high quality lidars that they are classifying objects directly from the lidar images


I bet any amount of money that rivian advertises heavily that R2 has lidar on autonomy day. And talks about how lidar will allow them to deliver self driving. Tesla would not have self driving any sooner if they had lidar.

In fact, all Tesla has to do is remove the safety monitor and then you will see RJ making more claims about autonomy. Because why would you buy one car when a competitor's technology will be already here, even if at small scale

what SAE L3 system exists today with monocular vision based depth perception? I had not realized this was a trivial problem, and that companies were able to do this with a single sensor stack and no fusion? You are also chopping and changing the arguments - so I’m gonna dab out for a bit while the tryptophan works its way through the system.

 

[sunydrm]

what SAE L3 system exists today with monocular vision based depth perception? I had not realized this was a trivial problem, and that companies were able to do this with a single sensor stack and no fusion? You are also chopping and changing the arguments - so I’m gonna dab out for a bit while the tryptophan works its way through the system.

Level 3 driving (mercedes) requires a lead car, below 45mph,premapped/validated roads, sunny weather only, no wet roads, perfect markings, no construction, no emergency vehicles, and no driving during sunrise/sunset/night.

Under those conditions cameras can perform with 100% reliability.

In fact under those conditions it is impossible for a system like tesla FSD to get into a crash that they would be at-fault. Tesla could build a level 3 system already. They could have had this technology 3 years ago.

Level 3 systems are not attention-free systems in that the human supervisor is responsible for mechanical failures of the car even if they are not looking at the road. Therefore the only accidents you would ever get into with such a level 3 system are rear-endings, cut-ins, or mechanical failures.

I also don't think you realize that these L2++ end to end systems (even if not FSD) are more than 99% using vision only. Wayve AI is #2 in this field and they are doing vision only, even though their technology is sensor agnostic.

People blanketly state that you need these sensors for safety, but this is not guaranteed. There are scenarios where camera do not perform as well as other sensors. In those scenarios you simply drive differently than you would. Camera sensors surpass human vision in most respects already.

 

[mkhuffman]

...
If it’s in an urban area where having a vehicle doesn’t make sense because of decent public transport and the expensive of ownership (payments, insurance, fuel, and parking) then I think autonomous ride-share services will be more successful overall.

Where is there "decent" public transportation?

...
Rivian is in a tough spot where they do not have both the money for R2 and to enable self driving. it is one or the other

Do you have any data that supports this assertion?

Sponsored