It had a driver so I wasn't too scared. Mirrors? We ain't got no stinkin' mirrors! He's wack.
RoboTaxi in the wild, around Palo Alto...
- Thread starter JeffKeryk
- Start date
Not enough LIDAR. 
- Joined
- Nov 23, 2020
- Messages
- 784
This Reuters report is pretty critical. I would say more FUD, but it's not controversial to say that Tesla has not hit its Robotaxi expansion goals. And based on this report, it's a good thing it hasn't.
It ain't gonna be perfect. But sure; the question is, can it be any worse than human drivers?
- Joined
- Nov 23, 2020
- Messages
- 784
Unfortunately we can't really tell because Tesla is manipulating the stats. I don't have an issue with FSD per se. I have an issue with how Tesla is advertising it and reporting its performance. They're trying to get have their cake and eat it too by pushing a narrative about how capable it is while simultaneously making false comparisons. Even the name is a misnomer. It can easily lull people into a false sense of security that results in death or serious injury. A Tesla driver can decide to take that risk, but other motorists shouldn't be subjected to that.
Ya mean like drunk drivers?
Yeah if Elon wasn't stubborn about no lidar they would have been out on the street 2 years ago. Waymo was right. Lidar on KFC bucket is the right way to do it.Not enough LIDAR.
- Joined
- Nov 23, 2020
- Messages
- 784
Yes. I hope you don't mean to imply that because some people drive drunk a potentially dangerous driver assistance feature should be permitted? That sounds like whataboutism, which we all know is a complete logic fail.
Waymo has been all over for at least 2 years now. It is rare to see a Waymo with a driver now actually. I think I see on average 5-10 a day on my way to work everyday.
The only way I trust the statistic is when an insurance company outside of their own will write a policy with their own money at risk.
If they sold these for $25K with a steering wheel and pedals and the charge port in the normal location... sign me up!
- Joined
- Nov 23, 2020
- Messages
- 784
The thing I don't get is why it's not on the Cybercab. But I guess it would undermine the entire "cameras only" narrative that Elon pushed with FSD. There's no way Cybercab will be rolled out widely without safety drivers using only cameras. IMO, it won't work, no matter how many hardware upgrades he throws at it. The only real question left is how long it will take for investors to start questioning some of these narratives, which underlie the company's valuation.
It MAY work, but when your competitor is already 2 years ahead of you can you wait till it works (with probably GPU 2 generations into the future) or should you first start with lidar then gradually remove them?
- Joined
- Nov 23, 2020
- Messages
- 784
Tesla won't even do that. Despite all its claims regarding FSD capability, the driver is still considered at fault the event of an accident. Now, contrast that with Waymo, which operates with no safety driver and accepts liability for any collisions in which its vehicle is at fault.
You are behind, I have been seeing those things for months around Austin. Saw one today in fact. I'm to the point it rates a ho hum.
I even saw one in my neighborhood a few weeks ago and I live out on the edge of the suburbs, way outside the Cybertaxi service zone. I guessed that I must have a neighbor that works for Tesla.
I even saw one in my neighborhood a few weeks ago and I live out on the edge of the suburbs, way outside the Cybertaxi service zone. I guessed that I must have a neighbor that works for Tesla.
But LIDAR is not the limitation... the stuff at the very beginning and end typically is.
Almost all my FSD disengagements after using it for like 10,000 miles are convenience or navigation issues. Not safety stuff. Like, when I get to work, I have to park on the side of the building, not in front of it... FSD always tries to park in front of it. And when I get home, I need to back in to make the charger reach.
If I bought a Cybercab and couldn't disengage self-driving features this stuff would be an issue.
This IMO is the main reason why future FSD is subscription only instead of permanent sale. They could yank the support if they found problem that need expensive upgrade in a subscription, but if you bought it and they discontinue it they have to refund you and face class action lawsuit, or give you free upgrade. Too big of a liability.
- Joined
- Nov 23, 2020
- Messages
- 784
The issue isn't even GPU processing. The video is captured for labeling and training for FSD. X.ai has surplus compute that it is leasing to Anthropic. The issue is that the system can't account for every scenario that humans can with the same proficiency. There are so many FSD fails videos online illustrating the system making dumb mistakes. If it was a GPU limitation, the Cybercab could have more hardware on board to facilitate faster processing. It still needs human supervision for safety.
I used to work with some robotics optics. It is very hard for optics to be always accurate, and human eyes have way better dynamic range than cameras already.
Waymo's strength is not only lidar but that they don't give a hoot their cars are ugly with those silly KFC buckets on top and the spinning cameras all 4 corners. When you are trying to sell cars you have a reputation to keep your car sexy. When you don't sell car and and people are just sitting in it for a few minutes nobody freaking care how ugly it looks.
Elon wouldn't do that to his own car like Kanye West and Jay Z did to Maybach. Heck they wouldn't either if they are shareholder of Maybach.
https://supercarblondie.com/remembe...-jay-z-destroyed-a-maybach-for-a-music-video/
