UBER CFO buys R2 after Autonomy+ test drive

[savethemanual]

Sure. It's also possible this never happened.

As a marketing stunt, it wasn't very good. What they should have done is livestreamed it, because it's 2026 and these things are possible. No facts/data/evidence to support the claim made by Uber's CFO so I treat it as such.

Also quite possible there are reasons why (1) it wasn't livestreamed and (2) it wasn't recorded end to end and shared unedited.

Sure man, we're just living in the Matrix . Everything is fake and generated...

Sponsored

 

[Doogl]

I get that Rivian needs partnerships/sales as much as possible, but I personally dislike Uber quite a lot as a company, and this collaboration makes me a little bit uneasy.

 

[DCFC]

I get that Rivian needs partnerships/sales as much as possible, but I personally dislike Uber quite a lot as a company, and this collaboration makes me a little bit uneasy.

I greatly dislike Uber. I avoid at all costs and use Lyft. Only places I've had to use Uber has been in other countries.

 

[UnsungZero_OldTimeAdMan]

Uber is placing bets wherever it can (not just Rivian) to increase its odds of winning the driverless car race. It's like playing roulette in Vegas, if you were allowed to place as many bets on the table as you want.

https://www.wsj.com/video/series/ws...car-race/10F91546-7884-4404-8B65-E252B6C514A3

 

[shandering]

This statement is completely fabricated.

False. RJ said that he used his system in a loop around the bay area and the car drove for several hours. He implied without intervention. This is not possible.

In the demos they showed during autonomy day, they claimed the car was trained not to turn right on red. In one of the demos, it accidentally started to turn right on red. right into very obvious traffic. It's because the car literally did not understand other cars and their trajectory. That's why they had to train the car to not turn right on red. To hide this deficiency.

The car turned left by merging into a protected turning lane. Effectively you don't need to understand other cars because the turning lane is a free entrance if you go in right away.

So they faked the appearance of driving by making the car turn left at dotted intersections and turn right on green. Comma.ai systems can already do this. What they did was no more than what you can already do with a beta build of openpilot. It's basically a glorified lane centering and cruise system

That's why it takes 1 year for them to release a beta to consumers. Because they don't actually have a system that can drive or even close to it. If you actually put in an address and tried to drive there, you'd likely have an intervention after 10-20s

 

[shandering]

This video showing a demo drive with autonomy+ from 7 months ago. It did pretty well in my opinion. There was a glitch where it slowed down for a speed hump at a pedestrian crossing but did not speed back up. Early development days and it's certainly been worked on extensively since.

Watch every single video during autonomy day and you can get a picture of what the car really did. It was a "trick routing


In the demos they showed during autonomy day, they claimed the car was trained not to turn right on red. In one of the demos, it accidentally started to turn right on red. right into very obvious traffic. It's because the car literally did not understand other cars and their trajectory. That's why they had to train the car to not turn right on red. To hide this deficiency. If you can't do this, you can't even perform basic driving. Even the first version of FSD could do this, although badly

you can only figure out the trick they did after watching all the videos.

 

[cwq93r]

False. RJ said that he used his system in a loop around the bay area and the car drove for several hours. He implied without intervention. This is not possible.

In the demos they showed during autonomy day, they claimed the car was trained not to turn right on red. In one of the demos, it accidentally started to turn right on red. right into very obvious traffic. It's because the car literally did not understand other cars and their trajectory. That's why they had to train the car to not turn right on red. To hide this deficiency.

The car turned left by merging into a protected turning lane. Effectively you don't need to understand other cars because the turning lane is a free entrance if you go in right away.

So they faked the appearance of driving by making the car turn left at dotted intersections and turn right on green. Comma.ai systems can already do this. What they did was no more than what you can already do with a beta build of openpilot. It's basically a glorified lane centering and cruise system

That's why it takes 1 year for them to release a beta to consumers. Because they don't actually have a system that can drive or even close to it. If you actually put in an address and tried to drive there, you'd likely have an intervention after 10-20s

Isn’t that how progress is made? One step at a time?

 

[shandering]

Isn’t that how progress is made? One step at a time?

In this context, no. They built a "fake" system that really doesn't scale to what they are intending to do

They did it to hide their lack of progress

The chinese are behind tesla, and rivian has shown no indication they can even match the performance of the chinese



Here is a system performing on 33w which is trained on 1000 hours of driving, and it outperforms what rivian showed. Made by effectively one person

Sponsored