-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD fails to detect children in the road
- Thread starter 2101Guy
- Start date
Yeah, I really don't think Whole Mars Catalogue is joking or being sarcastic. I will try and look away - as you suggest. Hopefully this Jackass-style behavior goes away soon. I'm just grateful for now that there aren't any L4/L5 cars for sale out there in North America yet, oh how we will mess with those...
Jmolloy3
Member
hmmm, I need to line up some kids in my neighborhood, and see if I can replicate this claim (joking sort of, but not really). So far the FSD in my car is exceptional at identifying, and reacting to humans in the path.
Tam
Well-Known Member
Unlike Tesla, which uses unpaid owners, SuperVision has been tested by paid workers until it matured enough for a commercial release to the public last month. Unlike Tesla, which describes Autonomous features, MobilEye has been clear that it's an ADAS.
The release of SuperVision is more like GM Super Cruise that was tested first by GM workers before it was released to the public, not as a prototype but as a finished ADAS with stable AutoSteer that does not require hands-on the steering wheel.
Daniel in SD
(supervised)
It's not clear to me that there's much difference in safety between saying "this is a driver assistance system" and "this is a beta of robotaxi software" if the functionality is exactly the same. Also there is no simple correlation between how well the system performs and how safe it is. I think these "door to door" ADAS will get less safe the better they perform until you can eliminate the human entirely.
CyberGus
Not Just a Member
FSD beta users: 100,000
Children murdered by FSD: zero
TSLAQ shills losing all their money: priceless
Children murdered by FSD: zero
TSLAQ shills losing all their money: priceless
Tam
Well-Known Member
It sounds like you are talking about the safety issues of L2 in general rather than comparing an unfinished L2 prototype Tesla and a consumer-quality L2 product like SuperVision.
It turns out both GM and Zeekr are using MobilEye products.
GM does not specifically use "SuperVision," but it has been using MobilEye for its Super Cruise since 2017 and in the future Ultra Cruise.
Tam
Well-Known Member
Tesla autopilot killed 15-year-old Bay Area boy, says lawsuit
'There is a massive hole in the family.'
www.sfgate.com
Daniel in SD
(supervised)
Yes, I'm talking about what I think is the primary safety risk with these systems, automation complacency.
diamond.g
Active Member
That doesn't read like a FSD Beta accident. That reads like a "standard" autopilot accident.Tesla autopilot killed 15-year-old Bay Area boy, says lawsuit
'There is a massive hole in the family.'
This video analyses a number of, shall we say, points of concern with the "FSD kills kids" video:
thesmokingman
Active Member
diamond.g
Active Member
brainhouston
Active Member
another thing everyone forgets is how many Teslas vs Competition with FSD like software is on the road
its apples to oranges until we compare equal amount of "miles driven on autopilot" for every Software/Hardware
right now its all kinda biased cause Tesla has the most EVs on the road
im sure with 3 million cars from MobileEye there will be few glitches here n there
its apples to oranges until we compare equal amount of "miles driven on autopilot" for every Software/Hardware
right now its all kinda biased cause Tesla has the most EVs on the road
im sure with 3 million cars from MobileEye there will be few glitches here n there
Tam
Well-Known Member
Waymo is the one that doesn't believe in partial automation that still relies on humans.
On the other hand, MobilEye started as an ADAS supplier that wants to evolve to L5 eventually.
When I bought my Tesla in 2012, some Tesla owners also installed the add-on MobilEye lane departure alarm.
Daniel in SD
(supervised)
“It is not designed to cover all possible crash situations in a safe manner,” Amnon Shashua, who is also chief technology officer at the Israel-based maker of collision detection and driver assistance systems, told Reuters.
“No matter how you spin it, (Autopilot) is not designed for that. It is a driver assistance system and not a driverless system,” he said in an interview.
Will Supervision cover all possible crash situations in a safe manner? They literally say Supervision is "derived from AV development", it's their beta robotaxi software sold to consumers just like FSD beta.
Tam
Well-Known Member
MobileEye never liked the way Tesla explained Autopilot with a futuristic feature that MobilEye components in Tela AP1 hardware could not deliver: Your parking Tesla in your garage would come to greet you at the curb. That's why it broke off with Tesla.
When MobilEye describes its products, it accurately describes the expected capability.
SuperVision is a good L2 capacity expected to have fewer problems than Tesla has now (deaths and customer complaints).
To make SuperVision an L4, it has already added LIDAR to its testing fleet. In the future, it might use the next generation of radar--4D radar.
Tesla is the reverse of MobilEye: Instead of adding better hardware (LIDAR, 4D radar...), Tesla deletes existing hardware.
_Redshift_
Member
It's sad, but not what I'd call "Tesla murdering a child". It's a tragic combination of human error more than anything else; the driver was driving too fast for the conditions and not paying enough attention. the car that was hit pulled out of their lane in front of the speeding Tesla probably going at a low speed and couldn't move out of the way in time, and the teen that was killed was not wearing a seat belt, and was ejected from the vehicle. I know this freeway, and it's one I avoid whenever possible, because it's the truck route through parts of town with crazy drivers.
The Tesla tried to stop, but was not able to do it in time. But since Tesla has lots of money, they get sued whether it's their fault or not.
This whole thread is kind of stupid, as it's obviously a lot of bias both ways, and the original article was written by an idiot who hates Tesla for personal reasons, and so of course any results from them will be biased also, and contrived, with no analysis of what it means or how it compares to a human driver in a non-Tesla. But there's really no way to test this except by long-term statistics, most likely.
Similar threads
