4 ms·

It operates under such specific conditions, it’s not very useful right now. Unless you spend a large part of your commute on clogged California highways, there’

by ra7 2y ago

It operates under such specific conditions, it’s not very useful right now. Unless you spend a large part of your commute on clogged California highways, there’s not a big use case. But Mercedes would be hoping to expand that operational design domain (ODD). Ultimately, I don’t think their L3 technology is the groundbreaking part (I think they use Nvidia Drive). It’s that they are willing to take liability inside their ODD. This will be a good experiment to see how liability shakes out for an automaker putting it in the hands of customers to be used conditionally vs someone like Waymo who doesn’t want anyone in the driver’s seat at all.

pavlov 2y ago
> “commute on clogged California highways” It’s embarrassing for Tesla because they pretended they had this solved and practically ready to ship eight years ago, started taking customer money; still take customer money on promises of “Real Soon Now”; and yet it’s a bunch of staid old Germans who actually ship it.
- heyoni 2y ago
  I’m so glad I didn’t fall for it. Nearly bought one, then took it for a test drive at the dealership and it couldn’t even make a left turn correctly. Then I rented one on another trip because I wanted to test it more extensively and the car locked up because I activated some stupid easter egg that started playing a song with an animated gif all over the screen…even getting out of the car and locking it didn’t stop the damn thing and I literally just had to sit on the side of the road and wait it out. The weird thing is I haven’t been able to find that easter egg mentioned anywhere.
- kmacdough 2y ago
  I'd hold off on the complements, we haven't seen the Mercedes on the road yet.
  - drivingmenuts 2y ago
    Even if the self-driving bit sucks, you're still getting a pretty decent car, which is more than Tesla should be allowed to claim.
  - cycomanic 2y ago
    We have seen those cars on the road for a while (this has been over the news for quite a bit already), they just didn't market the tech as L3 due to legal reasons. Also the liability cover clearly shows a level of confidence above that of other car makers. Yes it is a very limited, but it is still quite a bit ahead of everyone else and if you're thinking we will get full selfdriving on any roads under any conditions in the near future than I think you're kidding yourself. The failure modes I have observed in driver assist (which is much easier than full self driving) with several car makes, really shows the tech is nowhere close yet.
- freetinker 2y ago
  I don’t quite understand Tesla’s insistence on vision-only. I don’t believe vision-only “FSD” will ever be good enough for mass adoption – even if it performs at human-level. Autonomous systems need to be better than humans to gain acceptance, and that’s simply not possible with a vision-only let’s-mimic-humans approach.
BoorishBears 2y ago
I work on AVs and this is a) incredibly useful b) the future of self-driving L3 is 100% the groundbreaking part because it's when the manufacturer has to take liability and it's where people can actually afford it. I believe in the technology "trickling up" for L3 much more than I believe tech in our million dollar robotaxis trickling down to consumer vehicles.
- RaftPeople 2y ago
  I agree with you, it seems like it's the most valuable first step. Just eliminating the daily rear-end collisions during commute would be a huge step forward.
  - ac29 2y ago
    Automated emergency breaking systems can prevent a lot of those already. That feature is available in Level 0 systems.
- ra7 2y ago
  I think it’s a good first step, but it’s still limited in usability in its current iteration.
  - RichardHesketh 2y ago
    This. It matters how you achieve level three. MBs solution requires a lot of legwork to build the datasets describing the specific roads where the L3 capability can then operate. There’s a massive amount to do in this approach to even come close to supporting a useful subset of all roads in their key markets. To my mind, Tesla is playing a long game betting on achieving a more generalised solution that doesn’t rely on having a detailed database encompassing every inch of many roads in vast amounts of detail that have to be captured in a costly, resource intensive effort. That, or their vast fleet of vehicles is gathering the data for them to leapfrog the likes of MB to higher levels of autonomy using a fused approach in the future. No other manufacturer yet has a comparable on the road data gathering capability. It’s a matter of opinion right now which of these stands a better chance of succeeding at higher levels of autonomy alone - the best chance may be a synthesis of both, or something else entirely. Fanboy holy wars aren’t going to resolve any of this.

by ra7 2y ago

pavlov 2y ago

> “commute on clogged California highways” It’s embarrassing for Tesla because they pretended they had this solved and practically ready to ship eight years ago, started taking customer money; still take customer money on promises of “Real Soon Now”; and yet it’s a bunch of staid old Germans who actually ship it.

heyoni 2y ago
I’m so glad I didn’t fall for it. Nearly bought one, then took it for a test drive at the dealership and it couldn’t even make a left turn correctly. Then I rented one on another trip because I wanted to test it more extensively and the car locked up because I activated some stupid easter egg that started playing a song with an animated gif all over the screen…even getting out of the car and locking it didn’t stop the damn thing and I literally just had to sit on the side of the road and wait it out. The weird thing is I haven’t been able to find that easter egg mentioned anywhere.
kmacdough 2y ago
I'd hold off on the complements, we haven't seen the Mercedes on the road yet.
- drivingmenuts 2y ago
  Even if the self-driving bit sucks, you're still getting a pretty decent car, which is more than Tesla should be allowed to claim.
- cycomanic 2y ago
  We have seen those cars on the road for a while (this has been over the news for quite a bit already), they just didn't market the tech as L3 due to legal reasons. Also the liability cover clearly shows a level of confidence above that of other car makers. Yes it is a very limited, but it is still quite a bit ahead of everyone else and if you're thinking we will get full selfdriving on any roads under any conditions in the near future than I think you're kidding yourself. The failure modes I have observed in driver assist (which is much easier than full self driving) with several car makes, really shows the tech is nowhere close yet.
freetinker 2y ago
I don’t quite understand Tesla’s insistence on vision-only. I don’t believe vision-only “FSD” will ever be good enough for mass adoption – even if it performs at human-level. Autonomous systems need to be better than humans to gain acceptance, and that’s simply not possible with a vision-only let’s-mimic-humans approach.

BoorishBears 2y ago

I work on AVs and this is a) incredibly useful b) the future of self-driving L3 is 100% the groundbreaking part because it's when the manufacturer has to take liability and it's where people can actually afford it. I believe in the technology "trickling up" for L3 much more than I believe tech in our million dollar robotaxis trickling down to consumer vehicles.

RaftPeople 2y ago
I agree with you, it seems like it's the most valuable first step. Just eliminating the daily rear-end collisions during commute would be a huge step forward.
- ac29 2y ago
  Automated emergency breaking systems can prevent a lot of those already. That feature is available in Level 0 systems.
ra7 2y ago
I think it’s a good first step, but it’s still limited in usability in its current iteration.
- RichardHesketh 2y ago
  This. It matters how you achieve level three. MBs solution requires a lot of legwork to build the datasets describing the specific roads where the L3 capability can then operate. There’s a massive amount to do in this approach to even come close to supporting a useful subset of all roads in their key markets. To my mind, Tesla is playing a long game betting on achieving a more generalised solution that doesn’t rely on having a detailed database encompassing every inch of many roads in vast amounts of detail that have to be captured in a costly, resource intensive effort. That, or their vast fleet of vehicles is gathering the data for them to leapfrog the likes of MB to higher levels of autonomy using a fused approach in the future. No other manufacturer yet has a comparable on the road data gathering capability. It’s a matter of opinion right now which of these stands a better chance of succeeding at higher levels of autonomy alone - the best chance may be a synthesis of both, or something else entirely. Fanboy holy wars aren’t going to resolve any of this.