What's new
By:

Robotaxis prediction

G

gaskiller

Active Member

Joined
Nov 16, 2023
Messages
1,583
Reaction score
1,524
nvidianews.nvidia.com

European Mobility Leaders Build Safe Vehicles and Robotaxi Fleets on NVIDIA Hyperion

NVIDIA today announced that leading automakers in Europe and mobility providers, including Bolt with Lucid, Mercedes-Benz, Volkswagen Commercial Vehicles and Einride, are showcasing selected autonomous-mobility use cases on the NVIDIA Hyperion computing platform, demonstrating potential...
nvidianews.nvidia.com nvidianews.nvidia.com

I could see Waymo being the costly option with how they incorporate sensors.

Tesla will hit a brick wall not using Lidar

Robotaxis using Nvidia software will leapfrog Tesla eventually, and be cheaper to manufacture than Waymo.
 
gaskiller said:
nvidianews.nvidia.com

European Mobility Leaders Build Safe Vehicles and Robotaxi Fleets on NVIDIA Hyperion

NVIDIA today announced that leading automakers in Europe and mobility providers, including Bolt with Lucid, Mercedes-Benz, Volkswagen Commercial Vehicles and Einride, are showcasing selected autonomous-mobility use cases on the NVIDIA Hyperion computing platform, demonstrating potential...
nvidianews.nvidia.com nvidianews.nvidia.com

I could see Waymo being the costly option with how they incorporate sensors.

Tesla will hit a brick wall not using Lidar

Robotaxis using Nvidia software will leapfrog Tesla eventually, and be cheaper to manufacture than Waymo.
Click to expand...
I will always remember this scene
www.themarysue.com

Let’s Break Down That Eerie Tesla Scene in 'Leave the World Behind'

'Leave the World Behind's Tesla scene has stirred some conversation among viewers. Here's what the scene means.
www.themarysue.com www.themarysue.com
 
The cyber cab thing looks like a joke. Who wants a two seater, two door cab?!? Sounds miserable. If I saw my Uber pull up in that I would pass.
 
gaskiller said:
nvidianews.nvidia.com

European Mobility Leaders Build Safe Vehicles and Robotaxi Fleets on NVIDIA Hyperion

NVIDIA today announced that leading automakers in Europe and mobility providers, including Bolt with Lucid, Mercedes-Benz, Volkswagen Commercial Vehicles and Einride, are showcasing selected autonomous-mobility use cases on the NVIDIA Hyperion computing platform, demonstrating potential...
nvidianews.nvidia.com nvidianews.nvidia.com

I could see Waymo being the costly option with how they incorporate sensors.

Tesla will hit a brick wall not using Lidar

Robotaxis using Nvidia software will leapfrog Tesla eventually, and be cheaper to manufacture than Waymo.
Click to expand...
Agreed. More and more people are choosing Nvidia's hardware and software stacks. That will just make Nvidia develop both even quicker. This is certainly a business that benefits from economies of scale and right now Nvidia seems like the clear winner if we're assuming Tesla can't make it work with vision-only.
 
The_Don said:
Agreed. More and more people are choosing Nvidia's hardware and software stacks. That will just make Nvidia develop both even quicker. This is certainly a business that benefits from economies of scale and right now Nvidia seems like the clear winner if we're assuming Tesla can't make it work with vision-only.
Click to expand...
The one problem with using Nvidia for the software portion on their own hardware is nividia really has no reason to make the software efficient since doing so would mean less hardware sales.
 
gaskiller said:
nvidianews.nvidia.com

European Mobility Leaders Build Safe Vehicles and Robotaxi Fleets on NVIDIA Hyperion

NVIDIA today announced that leading automakers in Europe and mobility providers, including Bolt with Lucid, Mercedes-Benz, Volkswagen Commercial Vehicles and Einride, are showcasing selected autonomous-mobility use cases on the NVIDIA Hyperion computing platform, demonstrating potential...
nvidianews.nvidia.com nvidianews.nvidia.com

I could see Waymo being the costly option with how they incorporate sensors.

Tesla will hit a brick wall not using Lidar

Robotaxis using Nvidia software will leapfrog Tesla eventually, and be cheaper to manufacture than Waymo.
Click to expand...
As a resident of Ariona, I've watched the development and maturation of Waymo over the past decade (and more). I am also a frequent Waymo customer when I go to/from my house to the airport.

Like you, I am skeptical of the safety of autonomous vehicles without Lidar.

Overall, I think Waymo's safety record is pretty good. But there are still many real-life challenges that even Lidar might not solve. Witness the rainstorm the week before last Christmas in the San Francisco area. I happened to be in the Bay area and en-route from Marin County to SFO to pick up my nephew. The traffic lights were out over a large part of SF, and the sea of Waymos stranded on the flooded SF streets was quite a sight to behold!

I am all for above L2 "autonomous" cars. It is one thing to have hands off "autonomous" driving on a freeway for long trips and commutes. It is a different challenge in city streets, inclement weather, children, dogs and cats etc..

It will be a good challenge for the industry.
 
ke2 said:
The one problem with using Nvidia for the software portion on their own hardware is nividia really has no reason to make the software efficient since doing so would mean less hardware sales.
Click to expand...
Mmmm that’s not what I would expect from that company.
 
Aka-none said:
Mmmm that’s not what I would expect from that company.
Click to expand...
Why not? The engineering side might want to make the code as efficient as possible but sales may say "hey your are making this software too good so we'll miss sales targets of our latest bigger and better widget. Stop that now"
 
ke2 said:
Why not? The engineering side might want to make the code as efficient as possible but sales may say "hey your are making this software too good so we'll miss sales targets of our latest bigger and better widget. Stop that now"
Click to expand...
I think we are many country miles away from that romantic scenario!
 
ke2 said:
Why not? The engineering side might want to make the code as efficient as possible but sales may say "hey your are making this software too good so we'll miss sales targets of our latest bigger and better widget. Stop that now"
Click to expand...
Having worked for a sw company that was the most cutthroat ever even they didn’t do that. This isn’t 1970
 
Aka-none said:
Mmmm that’s not what I would expect from that company.
Click to expand...
As a gamer who has suffered Nvidia for ~30 years, this is exactly what I would expect from that company.
 
The_Don said:
Agreed. More and more people are choosing Nvidia's hardware and software stacks. That will just make Nvidia develop both even quicker. This is certainly a business that benefits from economies of scale and right now Nvidia seems like the clear winner if we're assuming Tesla can't make it work with vision-only.
Click to expand...
It seems a bit premature to predict that Nvidia will be the clear winner. While they obviously have aspirations for autonomous driving, they really haven't demonstrated much actual success in that field. They're certainly a company with the deep pockets needed to invest in autonomous driving. But as the last decade has demonstrated, it's not a problem for the faint of heart.

On the hardware side, they're obviously extremely competitive. I don't question that they'll sell a lot of in-vehicle computing hardware. But on the software side, their performance has been a lot less consistent. I can think of three companies right off the top of my head that have concretely demonstrated more robust autonomy software than anything from Nvidia.
 
gaskiller said:
Tesla will hit a brick wall not using Lidar
Click to expand...

That does not appear to be the case. Tesla latest released data from tesla.com/robotaxi/safety

shows they are running 800+ vehicles and the entire texas registered fleet is being used

They seem to be very close to getting freeway driving to a level they will unlock as well as 24 hr service. At that point and with tesla continuing to scale at the rate they are, they could surpass waymo in 2027.

Right now waymo shuts down in light rain (or no rain but in forecast) because of getting stuck in floodwaters. Whatever remote support tech they use does not allow them to move the car in floodwaters and the car freezes up. They also refuse to use their remote driving feature so they always have to rescue the car.

It seems to be the limiting factor for tesla was fleet management as in the early days of unsupervised, the cars would go to the depot and circle for 20 mins trying to park. Whatever technical improvements they made with FSD v15 AI parking seems to have allowed them to run hundreds of vehicles in one city.

Every release of FSD so far has made big leaps in areas and had fundamental rewrites to core functionalities.

Mobileye at one point said tesla would hit a glass ceiling and they never did.
 
The thing I don't see anyone mentioning when discussing the idea that Tesla is so "great" or catching up with Waymo, or taking about things Waymo doesn't do ... Is the difference in safety mindset between the two companies.

In my mind, Waymo is erring on the side of safety, while Tesla seems to be balls to the wall and damn the consequences in a lot of things.
 
doubletwist said:
The thing I don't see anyone mentioning when discussing the idea that Tesla is so "great" or catching up with Waymo, or taking about things Waymo doesn't do ... Is the difference in safety mindset between the two companies.

In my mind, Waymo is erring on the side of safety, while Tesla seems to be balls to the wall and damn the consequences in a lot of things.
Click to expand...
This is 100% not the case.

Waymo has an issue where their cars get stuck in flooding waters. They've had this issue for more than 10 years, yet some how they managed to hide it by geofencing out the areas of failure. Right now waymo pretty much does not operate in rain anymore. Waymo is just good at hiding any deficiencies they have and it becomes ugly when they come out.

It's funny because every city waymo operates has pretty bad flooding issues.

Tesla does not cross train tracks. There was a failure video of a waymo the other day where the waymo straight up wanted to pass a downed gate arm and drive right onto the train tracks.

https://twitter.com/x/status/2112651514754007371

here is waymo literally stuck on a train track

1792814385794.webp


And this is waymo literally inbetween the gate arms with a train coming

1792814413461.webp



Tesla on the other hand seems to improve at a faster rate.

look at zoox. They have been doing unsupervised much longer than tesla and their driving does not appear to have gotten any better. The CEO said that tesla doesn't have a technology that works, and now tesla is dancing all over them.

people look at the news and they just simply don't understand how quick FSD reaction times are or the edge case safety that tesla really has. Tesla knows how good their safety is because they know what issues of consumer FSD are caused by people having dirty cameras (robotaxi has camera cleaning) or there is some mapping issue which you can obviously fix in robotaxi areas. With all the mechanical issues that consumer cars can have as well, it does show consumer self driving is not an easy or cheap thing to enable.
 
Last edited:
The way I see it is E2E neural networks gives you good looking driving quickly, but does not give you a safety level to do unsupervised. There seems to be some ridiculous hurdle that tesla has crossed.

on the other hand, using HD maps and 360 lidar gives you the ability to do driverless very quickly but the driving takes longer to improve and improve ODD. Just look at how badly zoox drives, yet they are able to do unsupervised. They also seem to not be improving or increasing ODD
 
Let's not forget that Tesla has something nobody else has...millions of beta test units on the street all the time.
 
mobada said:
Right now waymo shuts down in light rain (or no rain but in forecast) because of getting stuck in floodwaters. Whatever remote support tech they use does not allow them to move the car in floodwaters and the car freezes up. They also refuse to use their remote driving feature so they always have to rescue the car.
Click to expand...
I am not sure what your source for this is but I see Waymos operating in the rain in Phoenix all the time. Not that we get that much rain here. Halting service during flash floods actually seems prudent.
 
You must log in or register to reply here.
Back
Top